The NEXT LAST R600 Rumours & Speculation Thread [Archive] - Page 4

icecold1983

18-Mar-2007, 18:42

If you expect 30% "across the board" then you better get real. 30% in OGL games where NV has traditionally had a significant advantage to begin with? 30% in non-bw limited situations where 512-bit won't help? 30% from framebuffer size when 768MB isn't even being used fully yet? Those are unreasonable expectations. I really have a bad reaction to "across the board", as by defintion you've set up "failure" as one benchmark somewhere at 25%. That's the action of someone who has set up for failure before the game begins. If you said, for instance, "I'll be disappointed if it doesn't average 30% better in high-res bandwidth limited situations" that'd be different.

if a few tests were only 25% obv it wouldnt be a failure. but the card is launching at least 6 months after g80. also if any of the rumors pan out to be true(32 rops, almost twice the g80s flop throughput, ridiculous clocks) it should be at LEAST 30% faster than g80. i dont expect nvidias refresh to be noticeably faster because the factory overclocked cards cant go far past 625/2000.

btw 30% in the tests that matter, aka high res with aa+af in games that matter, aka not games like half life 2. so 30% across the board in stressful benches.

nicolasb

18-Mar-2007, 18:45

While there have been plenty of references to RV6xx having audio capability which may not have mentioned R600, have there been any explicit references to R600 not having audio? Absence of evidence is not necessarily evidence of absence.

trinibwoy

18-Mar-2007, 18:49

Why is he going to be disappoined if R600 isn't 30% faster "across the board", and why wouldn't those same reasons apply to expectations for an NV Ultra part?

Ah, you're focusing on "across the board". That didn't even register with me. Thought you were saying that the same things that are feeding the expectation of a significant performance advantage for R600 should apply to the Ultra as well. And I don't think those expectations have anything to do with timing in this case - more with the 512-bit and process advantage hype. Well actually, yes timing does play a role. It's reasonable to expect ATi to aim considerably higher than current GTX performance given how long it's been around. If you look at scenarios that actually count we already have 10% increases just from overclocking the GTX...so why not 30%?

But the Ultra supposedly would have no advantage over the GTX except for clocks. So 30% performance increase would mean +30% clocks.. So I still don't see how expectations for R600 (many unknowns) have anything to do with expectations for the Ultra......

willardjuice

18-Mar-2007, 18:57

RVxx

RV generally refers to mid range cards.

Yeah I'm aware that one of the slides said RV, but from this slide:

http://img465.imageshack.us/img465/6059/amdpkr6004se1.th.jpg (http://img465.imageshack.us/my.php?image=amdpkr6004se1.jpg)

ATi Radeon DX10 Graphics Line Up

Best-in class HDMI / HDCP implementation
- Fully integrated Solution for superior compatibility
- Supports complete DRM specifications (protected HD audio)

I took that to mean all the cards (including R600).

Kaotik

18-Mar-2007, 19:27

Source?

The AMD slides, one is titled "RV6xx HDMI / HD-Audio Implementation", if it was R600 too, it would surely say "R6xx" or "R/RV6xx" or something similar, in the graph on the slide, it also says "RV6xx" under the chip

Geo

18-Mar-2007, 19:35

Ah, you're focusing on "across the board"...

Yah. Take a look at the link I added upstream in my mini-rant. R300 vs GF4 was a disappointment by that standard (GF4 being released 8 months previously, btw). Now 1920x1200 8xaa/16xaf. . . if R600 doesn't win there comfortably on average, then I'll be disappointed too. . .

icecold1983

18-Mar-2007, 20:19

at least post aa/af benches geo.

Geo

18-Mar-2007, 20:25

at least post aa/af benches geo.

hence my objection to the standard you started with!

willardjuice

18-Mar-2007, 20:37

So can anyone definitely say either way if R600 supports the audio controller?

Kaotik

18-Mar-2007, 20:44

So can anyone definitely say either way if R600 supports the audio controller?

Well one slide clearly talks only about RV6xx's, but one other slide, which talks of the DX10 range of cards overall, might indicate that R600 has it too

neliz

18-Mar-2007, 20:47

If R600 supports audio out, it would be nice to see a HDMI connector on the back of the latest Rx6xx shots.

Slowly we see the whole card picture to picture :lol:
http://resources.vr-zone.com/newspics/Mar07/17/R600-vrzone1.jpg

I don't think any of the cards we've seen have Audio in (on the board itself)
Just look at w0mbat's pics.
http://forums.vr-zone.com/showthread.php?t=136208

nothing...

Tim Murray

18-Mar-2007, 20:59

If R600 supports audio out, it would be nice to see a HDMI connector on the back of the latest R600 shots.
Well, remember what the components of an HDMI connection actually are.

Ailuros

18-Mar-2007, 20:59

if a few tests were only 25% obv it wouldnt be a failure. but the card is launching at least 6 months after g80. also if any of the rumors pan out to be true(32 rops, almost twice the g80s flop throughput, ridiculous clocks) it should be at LEAST 30% faster than g80. i dont expect nvidias refresh to be noticeably faster because the factory overclocked cards cant go far past 625/2000.

Next time I read about R600 it'll have 64 ROPs and a core frequency over 1.5GHz. Of course is the sarcasm intended since I'm scratching my head as well over the supposed 32ROPs, as the almost twice the GFLOP throughput, not to mention the "ridiculous clocks".

It might very well be that R600 arrives after all with a higher frequency than the rumoured 800MHz, but all the other stuff isn't magically increasing from one day to the other.

btw 30% in the tests that matter, aka high res with aa+af in games that matter, aka not games like half life 2. so 30% across the board in stressful benches.

Granted HL2 is so CPU bound these days that it's practically worthless for such investigations, but how sure are you that you're not suggesting cherry picking specific games in order to reach the supposed 30%? Where it'll get tricky is if someone else cherry picks another bunch of games and shows a lower than that difference. If you'd be saying "up to" I'd say that even 30% might be estimated conservative.

neliz

18-Mar-2007, 21:04

Well, remember what the components of an HDMI connection actually are.

Haven't found anything wrt. audio yet

http://www.pcgameshardware.de/screenshots/medium/Imgp5544.jpg
http://www.pcgameshardware.de/screenshots/medium/Imgp5543.jpg

icecold1983

18-Mar-2007, 21:14

Next time I read about R600 it'll have 64 ROPs and a core frequency over 1.5GHz. Of course is the sarcasm intended since I'm scratching my head as well over the supposed 32ROPs, as the almost twice the GFLOP throughput, not to mention the "ridiculous clocks".

It might very well be that R600 arrives after all with a higher frequency than the rumoured 800MHz, but all the other stuff isn't magically increasing from one day to the other.

Granted HL2 is so CPU bound these days that it's practically worthless for such investigations, but how sure are you that you're not suggesting cherry picking specific games in order to reach the supposed 30%? Where it'll get tricky is if someone else cherry picks another bunch of games and shows a lower than that difference. If you'd be saying "up to" I'd say that even 30% might be estimated conservative.

any game that is gpu limited. oblivion, r6 vegas, tm united, 3d mark type benchmarks, coh, test drive unlimited, coj, stalker etc. those are the games that matter because they stress the gpu

Geo

18-Mar-2007, 21:59

Well, remember what the components of an HDMI connection actually are.

Dongle ftw?

INKster

18-Mar-2007, 22:13

What about analog audio ?

Is this a full blown Audio "Controller", or just an Audio "Codec" ?

jamis

18-Mar-2007, 22:20

What about analog audio ?

Is this a full blown Audio "Controller", or just an Audio "Codec" ?
Furthermore, Bergman gave details about the HDMI digital output interface that is supported by the R600. AMD has chosen to integrate an HD-Audio controller into the GPU, and not into the south bridge chipset as rival Nvidia has announced so far. The essential HDCP encryption keys for reproduction of HD commercial video titles are stored in to the GPU, allowing both HD audio and video signal to be driven to the HDMI output, after the essential synchronization. Note that video and audio signals will be merged only when the HDMI interface is in use.
http://www.cdrinfo.com/Sections/News/Details.aspx?NewsId=20005

AMD has also integrated an audio controller on its low and mid-range RV6xx range of graphics cards, as the GPUs natively support HDMI.

Since HDMI carries both video and audio streams over one cable, you might want to take advantage of this, especially in HTPCs. With an internal audio controller, this means the entire sound loop back process doesn’t require external cables and is synced up using graphics software before the frames are sent out.

This is a far superior way of doing things for two reasons: 1) no messy cables going out from your sound card, back in your case on your video card and 2) sync issues. Since graphics can take longer to process than audio, especially if you’re upscaling or doing a lot of post processing effects. Thus, you want the software to manage the audio and sync it with the processed video before it hits your screen.

This audio syncing technology isn’t wholly new; it’s essentially the same as we’ve previously seen in the AMD 690G intergrated graphics chipset last month.
http://www.bit-tech.net/news/2007/03/16/r600_family_has_independent_video_processor/

Kaotik

18-Mar-2007, 22:26

What about analog audio ?

Is this a full blown Audio "Controller", or just an Audio "Codec" ?

Nope, no analog, and you get only out from the card with HDMI

INKster

18-Mar-2007, 22:31

Well, Nvidia did the GPU + Audio "thing" twice in the past (NV1, and the Xbox 1 chipset).
They even had the Soundstorm solution on Nforce/Nforce 2.
It was deemed as a "superfluous feature for the core target", perhaps.

So, if having digital audio+video back on a single discrete card is back, it can only mean one thing:

The ICT start date is closer than we think... ;)

willardjuice

18-Mar-2007, 22:45

Thank you Jamis for confirming that the audio controller is indeed supported on the R600! :smile:

What about analog audio ?

Pfft, who needs analog support? :wink: Get a receiver with a HDMI input and then get real speakers! :cool:

The R6xx series could be the best sound cards I have ever purchased (if the driver support is there, which at this point I wonder about AMD's commitment toward the audio portion of this card). Who cares if it's only 29.9% faster than the G80, it's a great sound card!

INKster

18-Mar-2007, 22:58

Pfft, who needs analog support? :wink: Get a receiver with a HDMI input and then get real speakers! :cool:

The R6xx series could be the best sound cards I have ever purchased (if the driver support is there, which at this point I wonder about AMD's commitment toward the audio portion of this card). Who cares if it's only 29.9% faster than the G80, it's a great sound card!

I'm a purist.
Analog FTW. ;)

Besides, with the huge amount of electrical noise on a graphics card, the quality will never be quite the same than by using a discrete, dedicated soundcard.

The thing that makes this possible is the "over-simplification" that Microsoft did on Vista's audio sub-system.
Software-only is a world of pain and low quality sound so far.

willardjuice

18-Mar-2007, 23:08

Besides, with the huge amount of electrical noise on a graphics card, the quality will never be quite the same than by using a discrete, dedicated soundcard.

Huh? It's a digital signal. What noise are you talking about?

Also my receiver, without question, has better sound quality (ie: better DAC's) than my Elite Pro. I don't see how a dedicated soundcard (unless sound cards start coming out with HDMI outputs) would be better than R600's intergrated audio solution. Finally we have a sound card that supports sending multi-channel (up to 7.1 @ 24/192) uncompressed linear PCM out of a computer digitally. Ironically, it's a video card. :lol:

NocturnDragon

18-Mar-2007, 23:11

Besides, with the huge amount of electrical noise on a graphics card, the quality will never be quite the same than by using a discrete, dedicated soundcard.

If the electrical noise gets a problem for the digital audio, you will be pretty f**ked up also for the rest of the data transferring on the graphic card, won't you?

Cuthalu

18-Mar-2007, 23:44

Huh? It's a digital signal. What noise are you talking about?

Also my receiver, without question, has better sound quality (ie: better DAC's) than my Elite Pro. I don't see how a dedicated soundcard (unless sound cards start coming out with HDMI outputs) would be better than R600's intergrated audio solution. Finally we have a sound card that supports sending multi-channel (up to 7.1 @ 24/192) uncompressed linear PCM out of a computer digitally. Ironically, it's a video card. :lol:

Wireless lan uses digital signal and it doesn't have constant quality? When using toslink and even a good quality integrated audio controller compared to X-Fi, X-Fi has better audio quality, can be heard without numerical tests. And there's also tests which show clear sound quality differences between differend audio solutions when using S/PDIF, so it's not completely "subjective".

Btw, X-Fi supports multi-channel 24/192 PCM already.

mao5

19-Mar-2007, 00:02

Ailuros

19-Mar-2007, 00:06

any game that is gpu limited. oblivion, r6 vegas, tm united, 3d mark type benchmarks, coh, test drive unlimited, coj, stalker etc. those are the games that matter because they stress the gpu

What exactly did you just answer with the above? Read again.

icecold1983

19-Mar-2007, 00:08

What exactly did you just answer with the above? Read again.

u saying im going to cherry pick games

Ailuros

19-Mar-2007, 00:16

It's not worth debating anything if you can't or aren't willing to understand what others are saying. Drop the one-liner tactic and put a bit more effort in your replies.

I never said you're going to cherry pick games, my wording suggested something entirely different and was more like a question than any sort of weird accusation. Read back and after you're done with that, riddle me where you get the exaggerated unit counts/specs from.

Geo made a couple of very accurate points, but instead of trying to reason with them you decided to ask him for performance numbers which he obviously hasn't.

Kaotik

19-Mar-2007, 00:40

breaking news:

http://www.digitimes.com.tw/n/article.asp?id=0000044185_A3P5N6091U9DLGF4EIXKH

Here's a better quality translation (imo anyway):

Affects Vista PC to change planes tide brand-new cartography specification DirectX 10, draws a chart chip double male NVIDIA and AMD-ATI original predetermined in 2007 the second season promotes the correlation product, but because is connected plays is unable to follow promptly, the market universal anticipated DirectX 10 generations fear only then treat in 2007 next half or in 2008 can approach truly, however the game entrepreneur disclosed, AMD-ATI has in the near future planned hand in hand many cartography card factory, purchases 2 section DirectX 10 games jointly, makes a tie-in sale to promote higher order product R600 the detention to May in, encourages the DirectX 10 generations to accelerate to approach, so as to leads PC to change planes the tide, and strengthens the cartography card market sale strength.

NVIDIA takes the lead in November, 2006 to issue industry the jieshou funds support the DirectX 10 higher order cartography product, the core code number G80's Geforce 8800 families, the DirectX 10 generations enter the seed stage officially, because although has not had the DirectX 10 games to be taken to the threshing ground the support, is unable to display the excellent characteristic, but has actually raised the mighty waves in the cartography card entrepreneur. Promotes brand-new cartography core R600 as for the AMD-ATI original estimate in at the end of March, 2007 to plan, expected catches up with the match product to promote the time interval, because however considers the DirectX 10 benefits and GDDR4 memories is out of stock seriously, retards promotes R600 product to May in.

The game entrepreneur disclosed that, AMD-ATI in faces the partner to appeal promotes the product as soon as possible, and appraised the competitor rushes forward the DirectX 10 domains, but does not have the DirectX 10 games to match, the hero does not have the opportunity, then on a large scale plans and many cartography card entrepreneur purchases 2 section 9 to change the DirectX 10 game writings jointly by DirectX, including by the Techland manufacture, the Ubisoft release western cowboy fires at the game “the wilderness double flood dragon (Call of Juarez)”, the Eidos release “the justifiable defense (Just Cause)”, makes a tie-in sale in May to promote R600 cartography card, hardly fires the movement to urge the DirectX 10 generations softly to approach truly.

The cartography card entrepreneur pointed out that, at present AMD-ATI is being at backwardness stage, the continual three generation of cartography chip product in researches and develops and promotes the time interval to fall behind NVIDIA, lets the partner burning with impatience, in addition, as a result of the AMD-ATI higher order cartography card always all by EMS a factory Celestica generation of labor, pastes the sign sale by the cartography card entrepreneur, therefore, this AMD-ATI plans makes a tie-in sale 2 section DirectX 10 game software, has purchased the cost including in R600, will be unifies the cartography card partner union purchase to promote, impels the DirectX for 10 times to approach with joint forces.

It is noteworthy that, although NVIDIA is in the lead promotes the DirectX 10 products, but because of the non-game software coordination, is unable to display its characteristic, the market anticipated generally the third season distributes “Crysis” in 2007 after American commercial skill electricity (EA) and so on the DirectX 10 games to promote, changes planes the effect only then to be able to increase obviously warm, however offers a sacrifice to the DirectX 10 soft hard synchronizations in AMD-ATI after May to go on the market the plan, the attempt reverse declining tendency, the entire aspect greatly for will have a new look.

The cartography card entrepreneur will believe, because the consumer can true construct on Vista PC, experiences the DirectX 10 formidable cartography functions for the first time, further will stimulate the high consumption ability to swim plays the family to adopt changes planes or the promotion movement, and will lead changes planes the tide to approach ahead of time, this will be extremely important regarding Microsoft (Microsoft) new generation of work system Vista and the global PC industry.

Translation provided by http://www.systransoft.com/

19-Mar-2007, 00:40

breaking news:

According to Taiwan's manufacturers revealed that Although Germany Crytek developed before the "other Crysis> announced that the first DirectX 10 games can be listed third quarter, but AMD and ATI AIB partners. In mid-May has decided to allow listed DirectX 10 hardware and software. AMD ATI has joined AIB manufacturers, game manufacturers to jointly purchase 2 DirectX 10 games. 2 so that the game launched in mid-May and the flagship R600 bundled products listed DirectX 10 games for popularity. Taiwan's manufacturers revealed that R600 delayed due to DirectX 10 games and have yet to see GDDR4 Xiancun stock. So AMD ATI R600 graphics chip will be postponed to mid-May. Industry revealed AMD ATI AIB joint procurement of this two game manufacturers are fighting patch upgrade to DirectX 9 Dir ectX 10 games. including "Of Jauze> CAll (West fanaticism) and the" Jast Cause> (self-defense ), which has already cost 2 games into R600 chip cost, will be inserted in mid-May listed among the R600 chip packaging. In addition, the industry also revealed that R600 chip will be under the Celestica OEM EMS, AIB and then shipped to manufacturers note card sales.

http://www.digitimes.com.tw/n/article.asp?id=0000044185_A3P5N6091U9DLGF4EIXKH

Can someone translate this into proper English? Especially the first part?

Does it say that AMD will bundle R600 with Crysis (if it's released in Q3 probably through vouchers like HL2?)? And does it also say that the devs of Just Cause and Call of Juarez are working on DX10 patches for their games so it can be bundled by AIBs with the R600? If it does... that's a very smart move.

Edit: that longer translation is a bit clearer, but still... "retards promotes R600 product to May in". Lol. So no Crysis bundled with R600, but either Just Cause and/or Call of Juarez?

Kaotik

19-Mar-2007, 00:45

The way I understood it, DX10 versions of Call of Juarez and Just Cause

Skrying

19-Mar-2007, 00:50

That would be mighty interesting. Both of those games are very beautiful in their own ways.

Geo

19-Mar-2007, 00:55

If true, that's one more indication that ATI thinks that DX10 is to their advantage. That's been the subliminal message for many months, really. I think it can't just be Nvidia's driver issues, as AMD must know as well as we do that those are transient.

It will be very, very interesting to see what a "DX10 patch" to an existing DX9 game can bring to the party, IQ-wise. I can already hear some of our graybeards sneering that a game must be built from the ground-up over several years, etc. :smile: (Yes, Ail, I'm looking at you! :razz: )

Skrying

19-Mar-2007, 00:58

Well, if Just Cause is getting a patch I would be extremely interested in what they would add. The games biggest draw back with its visual was its textures and the lack of AA (would I guess could change), but the towns when up close were lacking in detail (texture and geometry). Not something I was aware a DirectX 10 patch could fix. Its not like they could improve the draw distance much (they want you to see clear detail on the moon then?) as its by far the best I've seen on a game yet.

Razor1

19-Mar-2007, 01:27

Farid

19-Mar-2007, 01:38

From now on, we should refer to AIBs as cartography card entrepreneurs.

Anyway, I was expecting something else, than DX10 versions of their last games, from Techland and Avalanche Studios good relations with AMD.
I knew they were amongst the first, for quite a while, studios with R600 samples, but I was expecting something such as (tech) demos of their next games to be shown at R600 launch, or something of that sort.

With that said, that article is the only one which states that it's merely DX10 patches to old games (It's written in Chinese and on internet, though). Avanlanche Studios and Techland might have more surprise up their sleeve for AMD DX10 class cards.
If true, that's one more indication that ATI thinks that DX10 is to their advantage.
Yeah, I see this the same way. They probably want DX10 software to be available ASAP for this very same reason.

Suflex

19-Mar-2007, 01:56

Forget about R600, those translations made my day, I'll now go "draw a chart chip double male NVIDIA..."

willardjuice

19-Mar-2007, 02:09

Wireless lan uses digital signal and it doesn't have constant quality? When using toslink and even a good quality integrated audio controller compared to X-Fi, X-Fi has better audio quality, can be heard without numerical tests. And there's also tests which show clear sound quality differences between differend audio solutions when using S/PDIF, so it's not completely "subjective".

Btw, X-Fi supports multi-channel 24/192 PCM already.

Uh what are you smoking?

First, S/PDIF does not support uncompressed mult-channel PCM. It only supports uncompressed stereo PCM. It uses Dolby Digital and DTS codecs for multichannel, both of which are compressed (not to mention that Dolby Digital Live and DTS Connect, both real-time encoders, are limited to 16/48 on the PC). So you are wrong; the X-Fi only supports uncompressed PCM with analog output. The R600 does it digitally.

Second, digital is digital. The digital output of the X-Fi will sound equally the same as the digital output on any other sound card. It's scientifically impossible for them to be different.

So uh, could you pass whatever you're smoking done to me? I've seem to ran out. :razz:

bloodbob

19-Mar-2007, 02:19

Second, digital is digital. The digital output of the X-Fi will sound equally the same as the digital output on any other sound card. It's scientifically impossible for them to be different.
Unless your using the sound card to do some of the processing.

That would be like saying a Geforce and a Raedon should produce an identical image on digital display.

Geo

19-Mar-2007, 02:20

Yeah, I see this the same way. They probably want DX10 software to be available ASAP for this very same reason.

Then the question becomes "Why and how?". Will these patches provide new levels or patch up the game across existing levels? And what will it bring IQ-wise? If we look at DX10 and where AMD might think it would have an advantage where it matters, then you'd think something to do with either Geometry Shader or physics would be the answer. Most likely Geometry Shader. And if we look at GS, then geometry amplification would stand in the dock as the most likely suspect, wouldn't it? How much of a tri budget increase would it take to make a little :shock: with a DX10 patch?

Or is there some other area of possibility I'm missing?

trinibwoy

19-Mar-2007, 02:27

Unless your using the sound card to do some of the processing.

That would be like saying a Geforce and a Raedon should produce an identical image on digital display.

Uh no...the Geforce and Radeon actually produce the output. In this case the digital information is already created - it just needs to be transmitted.

Cuthalu

19-Mar-2007, 03:12

Uh what are you smoking?

First, S/PDIF does not support uncompressed mult-channel PCM. It only supports uncompressed stereo PCM. It uses Dolby Digital and DTS codecs for multichannel, both of which are compressed (not to mention that Dolby Digital Live and DTS Connect, both real-time encoders, are limited to 16/48 on the PC). So you are wrong; the X-Fi only supports uncompressed PCM with analog output. The R600 does it digitally.

Second, digital is digital. The digital output of the X-Fi will sound equally the same as the digital output on any other sound card. It's scientifically impossible for them to be different.

So uh, could you pass whatever you're smoking done to me? I've seem to ran out. :razz:

Yes, I was wrong on the first issue. The second matter, I thought like that several years ago aswell, but you might wan't to check your facts about digital output in practice and do some listening; does it really look like identical to you: http://www.digit-life.com/articles2/creative-audigy2-zs/

There's probably lots of things which the sound card can screw up.

trinibwoy

19-Mar-2007, 03:29

Everything that makes Creative's products special has to do with digital to analog conversion and vice versa. My receiver has a mode that tries to improve compressed digital audio (mp3) but I'm not aware of sound cards doing anything similiar. And of course that sort of processing would be irrrelevant for a dolby encoded or uncompressed signal. Passing along a digital signal is as straightforward as it gets. No conversions or processing to screw up.

Kaotik

19-Mar-2007, 03:29

http://cgi.ebay.ca/ws/eBayISAPI.dll?ViewItem&item=160095063471

Scammer or someone got hold of R600 and is really selling it? :shock:

Skrying

19-Mar-2007, 03:30

The X-Fi have the "Crystalizer" that is suppose to enhance the sound quality of MP3, sure it changes it but its up to the user if that's better or not.

trinibwoy

19-Mar-2007, 03:34

http://cgi.ebay.ca/ws/eBayISAPI.dll?ViewItem&item=160095063471

Scammer or someone got hold of R600 and is really selling it? :shock:

Hmmm well the seller has decent and considerable (> 400) feedback. But it's very curious how someone could be in a position to obtain an R600 sample and not have the "conditions to test this type of card" :roll:

Probably jacked it off a truck in Cebit's parking lot :lol:

Skrying

19-Mar-2007, 03:40

http://cgi.ebay.ca/ws/eBayISAPI.dll?ViewItem&item=160095063471

Scammer or someone got hold of R600 and is really selling it? :shock:

How the? The pictures are ones I have not seen before, the seller has considerable feedback yet is completely clueless to what they have. That's going to spread on the net really quick, its going to be VERY interesting how how the bids will go, I'm willing to bet it will be a large chunk if someone has the guts to test the waters.

trinibwoy

19-Mar-2007, 03:45

I hope Mr. Clueless has drivers for that badboy too :lol:

{Sniping}Waste

19-Mar-2007, 03:49

http://cgi.ebay.ca/ws/eBayISAPI.dll?ViewItem&item=160095063471

Scammer or someone got hold of R600 and is really selling it? :shock:
What would suck is winning the bid of the card and not having drivers for it. Card is useless without the drivers.

willardjuice

19-Mar-2007, 03:57

Yes, I was wrong on the first issue. The second matter, I thought like that several years ago aswell, but you might wan't to check your facts about digital output in practice and do some listening; does it really look like identical to you: http://www.digit-life.com/articles2/creative-audigy2-zs/

There's probably lots of things which the sound card can screw up.

That's with "extra DSP processing" enabled. I use my X-Fi right now in bit-perfect playback mode with all DSP effects off. I know this works because I can pass 44.1kHz DTS or 44.1kHz DD files perfectly to my receiver (if the data is changed even the slightest it will ruin the DTS/DD file and you will only hear loud static; if the audio data is not changed at all the receiver will be able to decode the DTS or DD file fine). So for me, who never enables DSP effects, all digital outputs are the same. Your right, if you add extra effects, obviously you are changing the output. But that doesn't mean one digital output is any different from the next digital output, just the DSP's (and their effects) are different.

Cuthalu

19-Mar-2007, 04:04

That's with "extra DSP processing" enabled. I use my X-Fi right now in bit-perfect playback mode with all DSP effects off.

I do so aswell, but that's one way how the sound card can screw up things, and those 3 other sound cards aren't identical either.

On the topic, this is interesting video about some old stuff, power consumption etc.: http://www.gearlive.com/news/article/159-amd-press-qa

rwolf

19-Mar-2007, 04:53

http://cgi.ebay.ca/ws/eBayISAPI.dll?ViewItem&item=160095063471

Scammer or someone got hold of R600 and is really selling it? :shock:

The picture has a little web site watermark on it.

weeds

19-Mar-2007, 05:03

The picture has a little web site watermark on it.

The watermark is a small camera, all of his items and previous items have the same watermark.

EDIT: Looks like the auction has been ended.

Sobek

19-Mar-2007, 05:17

Looking through the other items the guy offers, the pictures match and it seems legit...

But that's just too wierd :???:

Pete

19-Mar-2007, 05:27

Someone help me read Tridam's mind (http://www.behardware.com/news/8667/cebit-radeon-x2000-discrete-but-present.html):

Finally, according to a partner of AMD, the ultra high end version of the R600 equipped with 1GB of GDDR4 might not be released, at least for now. The 512MB is, however, scheduled to be released end of April at 500€ approximately. To finish, we have noted the distance of ATI's partners with the manufacturer official position. They say that the R600 is running perfectly and that the additional delay is only for strategic purposes…

"They" in the last sentence refers to AMD, not ATI's partners, and the trailing "..." implies skepticism, right? So, is he skeptical of the official explanation? Does that mesh with the rest of the paragraph and the Digitimes word of bundled DX10 games (firm launch date, price, even bundle)? It's starting to sound like insufficient GDDR4 supplies and maybe possibly (gaspity) a delay of game in those DX10 patches are the real reasons for the launch delay, rather than core problems (yield, clocks, power).

Damien, feel free to help me out yourself. :)

Skrying

19-Mar-2007, 05:49

Cuthalu

19-Mar-2007, 05:52

Yea, he seems to skeptical about the official explanation. At least now the reason can't be "full line-up launch". Perhaps AMD want's to see nVidias counter to a 'lesser' R600, and if the 512mb version is enough good to beat 8800gtx, then it might make sense.

Ailuros

19-Mar-2007, 05:54

It will be very, very interesting to see what a "DX10 patch" to an existing DX9 game can bring to the party, IQ-wise. I can already hear some of our graybeards sneering that a game must be built from the ground-up over several years, etc. :smile: (Yes, Ail, I'm looking at you! :razz: )

I can't imagine any fundamental IQ improvements through a D3D10 path, rather mostly some minor performance increases (but I'd love to be convinced of the opposite). One exception (and to that a mighty important one for me) could be the combination of anything deferred (or "semi-deferred") and MSAA in a D3D10 path.

Arty

19-Mar-2007, 05:55

neliz

19-Mar-2007, 10:17

"They" in the last sentence refers to AMD, not ATI's partners, and the trailing "..." implies skepticism, right? So, is he skeptical of the official explanation? Does that mesh with the rest of the paragraph and the Digitimes word of bundled DX10 games (firm launch date, price, even bundle)? It's starting to sound like insufficient GDDR4 supplies and maybe possibly (gaspity) a delay of game in those DX10 patches are the real reasons for the launch delay, rather than core problems (yield, clocks, power).

It's like AMD's current statements, right? We will be surprised. The delays are only for marketing reasons etc. Even if it dissipates 200Watts of heat, all that power must go somewhere.

I think fuad or inq also said that the cooler will be 8800GTX like, just as I said in October/November. the fact that the cooling was ready then and it's an AMD assisted design leads me to believer that heat still is not the "problem." An issue maybe, but not a problem. heck, I'd even expect something like the 8600 to have abysmal performance when they finally launch it.. and AMD now knows they have to compete with nVidia's full-range launch in mid-April. Somehow they feel they can afford to launch R600 a month after the 8800Ultra and still best it.

_xxx_

19-Mar-2007, 10:23

The X-Fi have the "Crystalizer" that is suppose to enhance the sound quality of MP3, sure it changes it but its up to the user if that's better or not.

Well it's nothing but a software equalizer/compressor of some sort, so I'd call it a "feature for the lazy people" ;)

icecold1983

19-Mar-2007, 10:30

400,000 from the last CC.

do high end video cards normally sell this much? i was always under the impression very few people bought them.

no-X

19-Mar-2007, 10:55

http://cgi.ebay.ca/ws/eBayISAPI.dll?ViewItem&item=160095063471

Scammer or someone got hold of R600 and is really selling it? :shock:
He sells one more ATi graphics card - it looks like X1800/X1900 sample - no stickers, no AIB's signs...

http://cgi.ebay.ca/ATI-RADEON-VIDEO-CARD-PCI-EXPRESS-DUAL-DVI_W0QQitemZ160095058127QQcategoryZ40158QQrdZ1QQc mdZViewItem

Bouncing Zabaglione Bros.

19-Mar-2007, 10:59

The picture has a little web site watermark on it.

That's just the watermark you get when you host your pictures at Ebay.

neliz

19-Mar-2007, 11:12

He sells one more ATi graphics card - it looks like X1800/X1900 sample - no stickers, no AIB's signs...

http://cgi.ebay.ca/ATI-RADEON-VIDEO-CARD-PCI-EXPRESS-DUAL-DVI_W0QQitemZ160095058127QQcategoryZ40158QQrdZ1QQc mdZViewItem

With A11 written on top of it..

nicolasb

19-Mar-2007, 11:22

Just to chime in on the sound-card discussion: it is very, very easy for two digital audio outputs to produce sound with different qualities, despite both being bit-perfect. This is nothing whatever to do with noise, though, it's to do with jitter. Jitter is what pretty much entirely determines the quality of a digital audio source (such as a CD transport).

What "jitter" means in this context is variation in the timing of the pulses. The pulses themselves are almost always reliable, in so far as the actual 1s and 0s will always get through intact. But the interval between pulses can vary, and unfortunately most audio DACs depend on the timing of the digital signal itself to control the timing of the D/A conversion. If a pulse arrives slightly too soon or slightly too late this introduces distortion into the eventual analogue signal.

If it is the case that the Creative X-Fi's analogue output is superior to its digital output, this is not an indication that it has superior DACs; it is more likely to indicate that its S/PDIF output has very bad jitter. (This could, of course, be corrected by using a reclocking DAC, but most DACs don't do this).

As someone else has already pointed out, there are also significant advantages to having digital audio sent over HDMI rather than S/PDIF because the available bandwidth is so much higher: you can send 8 channels of uncompressed digital audio over HDMI, but no more than two over S/PDIF. You can send Dolby Digital or DTS over S/PDIF, but those are compressed, and are lossy compression formats - the equivalent of converting CD audio into an mp3 before outputing it. Converting multiple sound channels to DD or DTS on the fly is also a bit tricky. It can be done, but doing without introducing perceptible lag is harder.

The downside of using an R600 as a sound-card is that it won't have all the fancy EAX processing that a dedicated soundcard possesses. But I don't think R600/RV6xx audio output is really designed for games, I think it's designed for HTPC setups where you're using the PC as a BluRay or HD-DVD disc player, and feeding the soundtrack to an external surround-sound processor for decoding and conversion.

Sobek

19-Mar-2007, 12:00

With A11 written on top of it..

Which would suggest?...Surely not. :???:

no-X

19-Mar-2007, 12:10

A11 revision of R520/R580? Why not?

Sobek

19-Mar-2007, 12:16

Right! :wink:

Wonder why it was ended. Either the joke did it's purpose, he realised what he had, or found someone to sell it to privately...

Snyder

19-Mar-2007, 12:27

If it is the case that the Creative X-Fi's analogue output is superior to its digital output, this is not an indication that it has superior DACs; it is more likely to indicate that its S/PDIF output has very bad jitter. (This could, of course, be corrected by using a reclocking DAC, but most DACs don't do this).

The DACs themselves maybe don't, but shouldn't reclocking happen in any receiver as soon as the signal path passes some kind of dsp? (Even if no effects are in use - disregarding a "direct path" setting?)

Razor1

19-Mar-2007, 12:34

What are 8800 sales numbers like? Starting to get the impression that AMD has not felt any pressure to truly push the series out and instead is focused on doing it "right". Such a strategy would also suggest comfort in the performance of R600. I can't say I blame them if the actual sales numbers suggest no need to rush.

do high end video cards normally sell this much? i was always under the impression very few people bought them.

sales numbers are actually decent and yeah higher then normal for the g80 possibly due to lack of competition, 400k as serenity pointed out, that would be minimum around 2%-3% very possibly higher, don't have exact figures, but 57 million total gpu's desktop discrete and IGP were sold, so after taking IGP out, the figure has to be much less, then out of that figure 53% of that is nV's discrete section.
but for now lets take 53% of 57 million and 2% of that is around ~600k

Now if we are talking about 2% of gross profits from discrete sales well then that 2-3% very possibly higher would be even higher still.

seahawk

19-Mar-2007, 12:35

It's like AMD's current statements, right? We will be surprised. The delays are only for marketing reasons etc. Even if it dissipates 200Watts of heat, all that power must go somewhere.

I think fuad or inq also said that the cooler will be 8800GTX like, just as I said in October/November. the fact that the cooling was ready then and it's an AMD assisted design leads me to believer that heat still is not the "problem." An issue maybe, but not a problem. heck, I'd even expect something like the 8600 to have abysmal performance when they finally launch it.. and AMD now knows they have to compete with nVidia's full-range launch in mid-April. Somehow they feel they can afford to launch R600 a month after the 8800Ultra and still best it.

Or NV can´t afford to hold back the Ultra any longer. ;) And that is not because of whatever AMD/ATI is having or doing.

_xxx_

19-Mar-2007, 12:48

Right! :wink:

Wonder why it was ended. Either the joke did it's purpose, he realised what he had, or found someone to sell it to privately...

Or someone told him to remove it :)

_xxx_

19-Mar-2007, 12:49

Or NV can´t afford to hold back the Ultra any longer. ;) And that is not because of whatever AMD/ATI is having or doing.

Or it's just time for a new release since it'll be good 6 months since G80.

nicolasb

19-Mar-2007, 12:49

The DACs themselves maybe don't, but shouldn't reclocking happen in any receiver as soon as the signal path passes some kind of dsp? (Even if no effects are in use - disregarding a "direct path" setting?)Most audio DACs don't actually have a built-in clock circuit at all. The only time-reference available to them is when each pulse actually arrives at the S/PDIF input. (When I say "DAC" in that sentence I don't mean a DAC chip, I mean an actual DAC device that plugs into the mains and feeds an amplifier)

Not all DACs work like this. Some higher-quality ones (things like the Chord DAC64) actually buffer the signal in local memory for anything up to several seconds, then reclock it according to an onboard clock circuit. A not-quite-as-good solution is to reclock the signal on the fly by upsampling at a rate that is asynchronous with the original signal (for example, 44.1kHz to 96kHz). The fact that the signal is now 96kHz makes no difference whatever to the quality of the sound, but the fact that it has been reclocked does. This is why (for example) the Benchmark DAC1 sounds as good as it does.

Most audio processing (including things like Dolby Digital decoding) does not actually involve the signal being reclocked: the time data still comes purely from the timing of the original input pulses. This is why there is an audible difference between using an ultra-cheap DVD player and a more expensive one even if all they are doing is producing a digital bitstream which the same external device is decoding and converting.

Some AV processors/receivers these days can reclock the signal (and are thus almost immune to the quality of the transport) but they're still in the minority.

seahawk

19-Mar-2007, 13:28

Or it's just time for a new release since it'll be good 6 months since G80.

Kind of thinking around the same logic. If 8800 Ultra still is G80, then they musr release it 3-4 months before the refresh arrives in the stores. otherwise hardware partners and buyers would be very disappointed by the short time the 8800U enjoyed on top of the NV line-up.

17th April launch of 8800U would be very fitting for the refresh to arrive just in time for the back-to-school cycle.

Snyder

19-Mar-2007, 14:06

Most audio processing (including things like Dolby Digital decoding) does not actually involve the signal being reclocked: the time data still comes purely from the timing of the original input pulses.

Say what?! Looking at the AC3 algorithm, to even achieve that on purpose would be no small feat...

(BTW: sorry to all for the OT. :) )

Rangers

19-Mar-2007, 14:13

How the? The pictures are ones I have not seen before, the seller has considerable feedback yet is completely clueless to what they have. That's going to spread on the net really quick, its going to be VERY interesting how how the bids will go, I'm willing to bet it will be a large chunk if someone has the guts to test the waters.

Umm, it didn't get any bids.

For like 50 bucks Beyond 3d forum could have had it's own R600 to test to it's hearts content..

http://cgi.ebay.ca/ws/eBayISAPI.dll?ViewItem&item=160095063471

neliz

19-Mar-2007, 14:25

He IS from Canada after all.. has anyone tried to contact the guy yet?

Geo

19-Mar-2007, 14:56

Umm, it didn't get any bids.

For like 50 bucks Beyond 3d forum could have had it's own R600 to test to it's hearts content..

http://cgi.ebay.ca/ws/eBayISAPI.dll?ViewItem&item=160095063471

Click on the "0 bids" and you see a different story.

Sobek

19-Mar-2007, 14:56

He IS from Canada after all.. has anyone tried to contact the guy yet?

I considered it, but in the event it's a consiprical joke, I decided to reserve my dignity and leave the "Oh...hi. Uh...is that card real? Can you bla bla bla" for someone else :wink:

mao5

19-Mar-2007, 15:05

breaking news, did you see the R600 DX10 Patch for "Just Cause"? http://www.chiphell.com/images/smilies/lol.gif

http://bbs.souyo.com/uploadimage/igdbol_2007319_2776.jpg

Suflex

19-Mar-2007, 15:13

Shouldn't R600 cards have 2x6 or 1x6 + 1x8 power inputs? This Ebay card has only 1x6 AFAIC.

neliz

19-Mar-2007, 15:16

I considered it, but in the event it's a consiprical joke, I decided to reserve my dignity and leave the "Oh...hi. Uh...is that card real? Can you bla bla bla" for someone else :wink:

http://forum.beyond3d.com/images/smilies/icon_wink.gif

It seems to be a hardware retailer behind it (canadian and serbian) with a strangely defunct site

vertex_shader

19-Mar-2007, 16:34

Someone help me read Tridam's mind (http://www.behardware.com/news/8667/cebit-radeon-x2000-discrete-but-present.html)

This means r600 512mb gddr3 version will be the top dog? than the "1gb gddr4 vram, super BW" version hyped for nothing.

Hard to belive this card will be enough to beat nv refresh card.

Evildeus

19-Mar-2007, 16:46

This means r600 512mb gddr3 version will be the top dog? than the "1gb gddr4 vram, super BW" version hyped for nothing.

Hard to belive this card will be enough to beat nv refresh card.Well, in the french version, he says that:

- For 1 partner, the X2900 XTX 1 GB DDR4 will be available in the middle of may
- but for another one, and considering their last information, the card could never be produced

So take your bid ;)

Sound_Card

19-Mar-2007, 16:58

R6xx to be a sound card too:?:

http://www.beyond3d.com/content/news/127

Sorry if already pointed out, but I really like the idea. I just hope the solution is at least better than on board.

trinibwoy

19-Mar-2007, 17:12

Maybe AMD realized, like many people suspect, that 128GB/s of bandwidth to 512MB of RAM is enough for now? :roll: :grin:

Kaotik

19-Mar-2007, 17:13

R6xx to be a sound card too:?:

http://www.beyond3d.com/content/news/127

Sorry if already pointed out, but I really like the idea. I just hope the solution is at least better than on board.

Yes, it's been pointed and discussed, mainly in the RV6xx thread though, as it might be only RV6xx feature.
It's only there to pass through digital signal through HDMI, you won't get analog audio out of it.

Sound_Card

19-Mar-2007, 17:16

vertex_shader

19-Mar-2007, 17:31

Maybe AMD realized, like many people suspect, that 128GB/s of bandwidth to 512MB of RAM is enough for now? :roll: :grin:

Maybe, or AMD not want sell cards with 799$ price tag :wink:
Many user will be dissapointed, in the last few months most of the rumors come from the 1gb gddr4 version aspect, and x1950xtx already using gddr4 512mb memory and nv top dog has 768mb.

Fastest gddr3 memory in production is Qimonda (Infineon) 0.83ns (1200mhz), available 2007 Q2 :smile:

Kaotik

19-Mar-2007, 18:13

DemoCoder

19-Mar-2007, 18:22

If true, that's one more indication that ATI thinks that DX10 is to their advantage. That's been the subliminal message for many months, really.

Does this mean that because NVidia was heavily promoting HDR and PS3.0 with devs in their TWIMTBP program, that NVidia secretly thought they had a huge advantage here? HDR and PS3.0 on the NV4x weren't really that useful due to performance and non-orthogonality. I think rather, ATI's push for DX10 is simply that they need some showcase devs onboard.

It will be very, very interesting to see what a "DX10 patch" to an existing DX9 game can bring to the party, IQ-wise. I can already hear some of our graybeards sneering that a game must be built from the ground-up over several years, etc. :smile: (Yes, Ail, I'm looking at you! :razz: )

I voiced the same complaints when Half-Life 2 was announced, and I coined the term "piece-meal DX9" meaning an engine not designed from the *ground up* around DX9, but merely, "here and there". Sort of like when DX8 arrived and we had DX7 games that looked completely ordinary and unimpressive until you hit a scene with Water, and low and behold, you had this wierd "one of these things is not like the other" experiences. Not artistically a good design IMHO.

I do not think a "DX10 patch" to a DX9 came can showcase anything more than a few trivial effects anymore than the PS3.0 patch to FarCry really changed much. What are the major new extensions to DX10 that don't exist in DX9? Well, you have GS. But GS can really perform any real magic unless the art assets (geometry inputs) are designed for it. So what's the likely "GS patch"? Cruft.

What's "cruft"? I define this as taking existing primitives and adding "cruft" to them. :) Bumps, debris, growing branches, limbs, particles, etc. It's trivial to do, and doesn't require redesigning any geometry inputs specifically for high quality ampliciation.

Maybe, just maybe, we might see some "cube map in one pass" attempts, but it won't visually change anything, and probably won't even help with performance.

What's the other big DX10 feature? Stream out. If you're going to create geometry, you're going to use GS (streamout creating geometry from a pixel shader would just be like R2VB and can even be done under DX9). Therefore, what other uses for stream out could they show in a "patch"? Physics? Unlikely, because the major physics middlewares Havok/Novodex/et al I believe will be customized for CUDA/CTM/SMP/etc and not written to high level DX10 streamout.

Ah...water. Maybe there will be a killer DX10 GS/Streamout water. :)

John Reynolds

19-Mar-2007, 18:29

I voiced the same complaints when Half-Life 2 was announced, and I coined the term "piece-meal DX9" meaning an engine not designed from the *ground up* around DX9, but merely, "here and there". Sort of like when DX8 arrived and we had DX7 games that looked completely ordinary and unimpressive until you hit a scene with Water, and low and behold, you had this wierd "one of these things is not like the other" experiences. Not artistically a good design IMHO.

Couldn't agree more. That habit has been a really bad way of creating very incongruent looking games.

19-Mar-2007, 18:30

I would like some imput on this...

I have been thinking, it would be a great idea if ATi stuck the Xilleon chip on thier graphics boards instead of the Rage theater pro?

http://ati.amd.com/products/dtv/revolutionary.html

All X1K cards that support AVIVO already use Xilleon technology inside. They all have the same TV Encoder as Xilleon: http://ati.amd.com/technology/avivo/pdf/Avivo_Display_Engine_Whitepaper_v2_final.pdf

Kaotik

19-Mar-2007, 19:40

All X1K cards that support AVIVO already use Xilleon technology inside. They all have the same TV Encoder as Xilleon: http://ati.amd.com/technology/avivo/pdf/Avivo_Display_Engine_Whitepaper_v2_final.pdf
But the rest of the video engine is using the 3D part of the chip, in the R/RV6xx's, there's separate Xilleon based video engine

caffeinated

19-Mar-2007, 20:25

Frank

19-Mar-2007, 20:34

Ah...water. Maybe there will be a killer DX10 GS/Streamout water. :)
Fluids, particles and destructible environments always go well. And bounching boobies, of course.

;)

Jawed

19-Mar-2007, 21:06

And bounching boobies, of course.
Sadly Ruby: Whiteout doesn't appear to offer evidence that the D3D10 generation will suffice :sad:

Jawed

Kaotik

19-Mar-2007, 21:36

Sadly Ruby: Whiteout doesn't appear to offer evidence that the D3D10 generation will suffice :sad:

Jawed

Maybe you will need 2nd R600 to do the physics for the boobies? :eek:

vertex_shader

19-Mar-2007, 21:54

What a surprise vrzone writing from r600 and bad yields :lol:
CeBIT : 65nm R600 Suffers Poor Yield
Over at CeBIT, we overheard some production problems with R600. The yield at 65nm is very low at this point of time. It is projected by middle of April, AMD is only able to churn out some 20,000 Radeon X2900 XTX graphics cards and XT SKU might not even make it for the launch in May.

Vrzone never give up writing BS for ATi and now for AMD, i bet before this will be the next one from vrzone, so i won :lol:

Kaotik

19-Mar-2007, 22:05

What a surprise vrzone writing from r600 and bad yields :lol:

Vrzone never give up writing BS for ATi and now for AMD, i bet before this will be the next one from vrzone, so i won :lol:

Regardless of the numbers, they're damn right they have "poor yields" with 65nm R600 - I mean, what else can you expect when you're doing 80nm R600's :wink:

DemoCoder

19-Mar-2007, 22:16

Sadly Ruby: Whiteout doesn't appear to offer evidence that the D3D10 generation will suffice :sad:

Jawed

Well, even worse, Ruby's skin shader in those pictures looks even worse than the DX9 generation, and Ruby looks more plastic than ever in the cockpit scene.

Bouncing Zabaglione Bros.

19-Mar-2007, 22:18

Well, even worse, Ruby's skin shader in those pictures looks even worse than the DX9 generation, and Ruby looks more plastic than ever in the cockpit scene.

Are you saying that based on the terrible quality Youtube video? If there's any subsurface scattering, would we be able to even see it on that clip linked upstream?

bloodbob

19-Mar-2007, 22:32

Uh no...the Geforce and Radeon actually produce the output. In this case the digital information is already created - it just needs to be transmitted.
And what about in all the cases where the sound card PRODUCES the output have you heard of a mixer? Or what about the game do they do all the stuff in software and there is no such thing as hardware accelerated D3DSound or OpenGl and I guess there is no hardware MIDI either?

Razor1

19-Mar-2007, 22:33

Well, even worse, Ruby's skin shader in those pictures looks even worse than the DX9 generation, and Ruby looks more plastic than ever in the cockpit scene.

The actually demo looks much better, Ruby's animations (facial are especially good, now thats pure artwork nothin to do with Dx API, but speical attention has been paid to final touches). Ruby's skin is pretty good as good as Adriana's yeah hard to tell in that low quality video though I agree, now the eye shaders don't look that great, but the eye movement is much more releastic then we saw with Adrianna, her eyes looked like a total freak!

neliz

19-Mar-2007, 23:37

Her mouth and all muscle animation in the face looks pretty amazing actually..

Tim Murray

19-Mar-2007, 23:40

Speaking of Ruby... (http://www.hexus.tv/show.php?show=66)

Sound_Card

19-Mar-2007, 23:42

Speaking of Ruby... (http://www.hexus.tv/show.php?show=66)

Hexus hates opera.

Sound_Card

19-Mar-2007, 23:49

ok just watched it.

looks even more impressive now. :wink:

Doomtrooper

20-Mar-2007, 00:01

Interesting...

willardjuice

20-Mar-2007, 00:07

Frank

20-Mar-2007, 00:09

Speaking of Ruby... (http://www.hexus.tv/show.php?show=66)
I don't think Ruby has any sex appeal. Which should be the main thing of using a girl for the demo, right?

Tim Murray

20-Mar-2007, 00:26

For what it's worth, the demo is running in real-time on an X2900 XT.

Razor1

20-Mar-2007, 00:40

I don't think Ruby has any sex appeal. Which should be the main thing of using a girl for the demo, right?

I think that russain female voice with the masculine undertone is what killed the sex appeal :wink:

Ike Turner

20-Mar-2007, 00:47

Speaking of Ruby... (http://www.hexus.tv/show.php?show=66)

Fap fap fapfap

INKster

20-Mar-2007, 00:58

Fap fap fapfap

Huh ?
------------

I agree, the Russian voice sounds too masculine.
I want the voice of a nice, curvaceous, air headed woman, like a Paris Hilton or Pamela Anderson. :D

Raqia

20-Mar-2007, 01:03

It would have been funny if the sniper had fairy wings...

Geo

20-Mar-2007, 01:07

What a surprise vrzone writing from r600 and bad yields :lol:

Vrzone never give up writing BS for ATi and now for AMD, i bet before this will be the next one from vrzone, so i won :lol:

2% yield on 24-pipe R520s I think they reported wasn't it? :lol:

Geo

20-Mar-2007, 01:10

Interesting...

Whoa. He's alive! :wink:

I think that demo stomps Adrienne pretty well. Tho I thought Ruby's head was a bit narrow and long in a few scenes.

It's hard to argue that ATI doesn't put more effort into these things than NV does.

Edit: Here, I'll be a little provocative. I'd estimate based on action, scenes, art assets, and length that the number of person hours directly involved in making "Whiteout" (directly involved --no hours from the driver team) were more than Froggy, Smoke, Cascades, and Adrienne combined. Agree/disagree?

rwolf

20-Mar-2007, 01:17

2% yield on 24-pipe R520s I think they reported wasn't it? :lol:

And how long was it until the R520 was replaced. :wink:

Geo

20-Mar-2007, 01:26

And how long was it until the R520 was replaced. :wink: Relevance to how far up their ass VR Zone's head has historically been regarding ATI yields?

INKster

20-Mar-2007, 01:27

Edit: Here, I'll be a little provocative. I'd estimate based on action, scenes, art assets, and length that the number of person hours directly involved in making "Whiteout" (directly involved --no hours from the driver team) were more than Froggy, Smoke, Cascades, and Adrienne combined. Agree/disagree?

Too bad they didn't use those hours getting R600 out the door sooner instead.

flippin_waffles

20-Mar-2007, 01:29

Dang, I was sorta expecting to see Jade poke her head out in the demo. Is there another demo, or is Jade still-born? :( Or was it just a rumor!.... hmm

[edit]
Impressive as is though! :)

Geo

20-Mar-2007, 01:30

Too bad they didn't use those hours getting R600 out the door sooner instead.

Think those are fungible resources do you? You must be right at home in "all the employees are just widgets" corporate-land.

icecold1983

20-Mar-2007, 01:31

the mountains look unimpressive and very out of place in the ruby demo.

INKster

20-Mar-2007, 01:36

Think those are fungible resources do you? You must be right at home in "all the employees are just widgets" corporate-land.

I'm certain that they are far more at will than me in that department. :wink:
And ATI now has more resources/"widgets" than Nvidia, so, no excuses for them.

BTW, i'm still holding out on my decision to purchase a 8800 GTS 640MB, just to see what the "cheaper" R600 (if there's any) has got to woo me away from Nvidia this time.

Geo

20-Mar-2007, 01:42

Every week I like my decision to buy an 8800 GTX the week it was released a bit better. YMMV. :smile:

INKster

20-Mar-2007, 01:46

Every week I like my decision to buy an 8800 GTX the week it was released a bit better. YMMV. :smile:

Hey, is that a hint that the 8800 GTS will not be worth it soon ?
Thanks... ;)

Andrew Lauritzen

20-Mar-2007, 01:54

I think that demo stomps Adrienne pretty well.
Agreed, although Adrienne was particularly unimpressive IMHO. However, the new Ruby demo seems fairly dull as well, although I'll wait until I see it in high resolution to pronounce final judgement,...

To be honest, Cascades has been the coolest demo technically so far. It actually uses a lot of new technology to do things that were not possible on previous hardware... sure it's not a terribly complicated idea in and of itself, but it does a lot of cool things at once, and looks pretty decent as well.

I'm the first to admit that art is primarily what people perceive when they comment on the quality of a modern demo like these, and Ruby probably has the best art of the new demos (and arguably Toy Shop for last generation). Games like HL2 and Max Payne show this effect very clearly.

That said, I still prefer the "tech demos" to really show off the technology primarily, the art being somewhat secondary. With that in mind, I would consider all of NVIDIA's recent demos to be tech demos, whereas Ruby seems to be targeting a completely different "market".

I guess my point is that while Ruby may be the coolest to the average end user, I still find Cascades (and Smoke really) to be a more impressive achievement, and more useful to developers - so don't be too hard on NVIDIA ;)

Geo

20-Mar-2007, 01:54

Hey, is that a hint that the 8800 GTS will not be worth it soon ?
Thanks... ;)

Hmm. Actually, as my sig might suggest, I'm not a big fan of GTS right now at the current price. Unless you have an urgent need. . .

INKster

20-Mar-2007, 01:59

Hmm. Actually, as my sig might suggest, I'm not a big fan of GTS right now at the current price. Unless you have an urgent need. . .

Maybe they'll be cheaper come April 17th, who knows... :wink:

Geo

20-Mar-2007, 02:01

Agreed, although Adrienne was particularly unimpressive IMHO. However, although I'll wait until I see it in high resolution to pronounce final judgement, the new Ruby demo seems fairly unimpressive as well..

To be honest, Cascades has been the coolest demo technically so far. It actually uses a lot of new technology to do things that were not possible on previous hardware... sure it's not a terribly complicated idea in and of itself, but it does a lot of cool things at once, and looks pretty decent as well.

I'm the first to admit that art is primarily what people perceive when they comment on the quality of a modern demo like these, and Ruby probably has the best art of the new demos (and arguably Toy Shop for last generation). Games like HL2 and Max Payne show this effect very clearly.

That said, I still prefer the "tech demos" to really show off the technology primarily, the art being somewhat secondary. With that in mind, I would consider all of NVIDIA's recent demos to be tech demos, whereas Ruby seems to be targeting a completely different "market".

I guess my point is that while Ruby may be the coolest to the average end user, I still find Cascades (and Smoke really) to be a more impressive achievement, and more useful to developers - so don't be too hard on NVIDIA ;)

Ooh, I didn't mean to be "hard" on nvidia. It's about choices, really. Every dog should have his day where he spends his resources. Certainly it's generally been pretty clear that NV spends more on devrel, for instance.

I like Cascades a lot, really. And Smoke (lots of aliasing tho). And Froggy. And. . . err, no I can't say it. :sad: I don't like Adrienne. My gallant reflex makes me feel bad for the lady (rather than nvidia).

But having said all of that. . . do you agree or disagree with my estimation of person hours? :razz:

Geeforcer

20-Mar-2007, 02:04

Hmm. Actually, as my sig might suggest, I'm not a big fan of GTS right now at the current price. Unless you have an urgent need. . .

IDK - 320MB version seem like pretty good deal... and the cheapest 640MB version is below the cheapest X1950 XTX. :shock:

Andrew Lauritzen

20-Mar-2007, 02:05

I like Cascades a lot, really. And Smoke (lots of aliasing tho). And Froggy. And. . . err, no I can't say it. :sad: I don't like Adrienne.
We're certainly in agreement then :)

But having said all of that. . . do you agree or disagree with my estimation of person hours? :razz:
You're probably right, considering that it seems that most of the NVIDIA demos (Cascades and Smoke at least) were written primarily by one person each. With Ruby I suspect there was a decently large art team and much more work on stuff like motion capture, etc.

nutball

20-Mar-2007, 02:09

Edit: Here, I'll be a little provocative. I'd estimate based on action, scenes, art assets, and length that the number of person hours directly involved in making "Whiteout" (directly involved --no hours from the driver team) were more than Froggy, Smoke, Cascades, and Adrienne combined. Agree/disagree?

I'll be a bit provocative back: they've had six months longer to work on it.

Geo

20-Mar-2007, 02:13

I'll be a bit provocative back: they've had six months longer to work on it.

Hah. Care to check that release date on the Cascades demo?

Geo

20-Mar-2007, 02:16

IDK - 320MB version seem like pretty good deal... and the cheapest 640MB version is below the cheapest X1950 XTX. :shock:

If your need is urgent, sure. :smile: I don't have price info for X2900 XT. But I still have a suspicion it's going to be a much better price/performance value than current GTS 640 pricing, and a bit more future proof (framebuffer size-wise) than GTS 320. At any rate, I suspect it might drive down GTS 640 prices. But make no mistake --that's speculative.

INKster

20-Mar-2007, 02:25

If your need is urgent, sure. :smile: I don't have price info for X2900 XT. But I still have a suspicion it's going to be a much better price/performance value than current GTS 640 pricing, and a bit more future proof (framebuffer size-wise) than GTS 320. At any rate, I suspect it might drive down GTS 640 prices. But make no mistake --that's speculative.

If NV were to release a smaller GTX (like the QuadroFX 4600 is doing, basically a G80 GTX in a shorter G80 GTS PCB), do you think it would be better value against the price/performance R600 counterpart (according to current public knowledge, we can expect a R600 "XL" or R600 "XT" to fill that segment) ?

R300King!

20-Mar-2007, 02:28

Hexus hates opera.

No. Works fine here on my Opera 9.10 build 8666 :)

Much nicer video than the YouTube one. I bet watching it in real time is even better.

Here is a quote from the poster.

ATi Whiteout... a video of the fantastic new 'Ruby' technology demo which AMD is showcasing the power and technology of its much anticipated next-gen flagship graphics accelerator 'R600'.

Watch it and weep NV fan-bois... the HD version is utterly awesome and even more so when seen rendered in real-time!

Highlights to us are the highly realistic textures on Ruby's leather jump-suit, the motion-blur and other cinematic effects plus the final, massively detailed frame of everyone's favourite red-head smiling wryly at the camera.

Whilst woefully late, the ATi 'R600' looks like it could well be worth the wait and we can only imagine how cool the ATi Whiteout demo would look on an R600 CrossFire rig at even higher resolutions!

DemoCoder

20-Mar-2007, 02:39

Even on poor youtube quality videos you can tell if the skinshader looks sucky or not. Besides, going back to the leaked presskit snapshots of the closeup of Ruby's lips/face, it still doesn't look as good as some other fake-SSS shots we've seen to date. I don't think Ruby is an example of technical excellence, I think its an example of good artistry.

Geo

20-Mar-2007, 02:46

If NV were to release a smaller GTX (like the QuadroFX 4600 is doing, basically a G80 GTX in a shorter G80 GTS PCB), do you think it would be better value against the price/performance R600 counterpart (according to current public knowledge, we can expect a R600 "XL" or R600 "XT" to fill that segment) ?

Well, if Tridam is right and XT is 512MB, then you've got a bit of asymmetry going on in cost structure and possibly performance profile implications thereof. 1GB vs 768MB vs 640MB vs 512MB vs 320MB. Add to that that NV's architecture and cost structure explicitly ties (while sticking the hot knife to unlockers!) memory size to activated units.

I look at that situation and wonder if we could see a GTX 384 MB? That would help on the cost side, certainly. . . but would it hurt significantly on performance vs a 512MB XT in stuff like 8XQ AA?

trinibwoy

20-Mar-2007, 03:08

And what about in all the cases where the sound card PRODUCES the output have you heard of a mixer? Or what about the game do they do all the stuff in software and there is no such thing as hardware accelerated D3DSound or OpenGl and I guess there is no hardware MIDI either?

I never said sound cards don't produce sound. Just that RV6xx's digital passthrough doesn't. You seem to be ignoring the fact that the sound signal is not being produced or manipulated by the hardware in this case. If R600 was capable of any of those functions then they would be relevant but at this time they aren't.

Rangers

20-Mar-2007, 03:15

Add to that that NV's architecture and cost structure explicitly ties (while sticking the hot knife to unlockers!) memory size to activated units.

Unlocking quads has been a thing of the past for some time. Something about they're laser cut now.

Geo

20-Mar-2007, 03:21

Unlocking quads has been a thing of the past for some time. Something about they're laser cut now.

Even so. It's likely to be cheaper and quicker to control such by how many memory chips you put on the pcb than going back to laser cut healthy chips. . . Lower price = higher volume. So that's almost always going to mean a healthy add-in of "undamaged" chips to unit-deactivated skus. At least in the first 1/2 of an architecture's life-span. In the last 1/2 you start getting skus that are designed only to clear out the cardboard boxes in the CEO's garage at a healthy profit. . .

trinibwoy

20-Mar-2007, 03:22

Hmm. Actually, as my sig might suggest, I'm not a big fan of GTS right now at the current price. Unless you have an urgent need. . .

Oh lol, that's what that means? Have been trying to figure that out for a while now :) Yeah, I've always thought there was a big gap to fill between the GTS and GTX. I'm still giving Nvidia a chance to play their full hand to see if that sheds some light on that thinking.

With regard to your comment about the number of units being correlated with memory size...exactly which units do you mean? AFAIK the number of ROPs determines bus-width and this is all decoupled from the shading/texturing core.

EDIT:

e.g If I were Nvidia I'd be working on the following 80nm refresh SKU for ~ $350 (the 320MB GTS is cool but performance falls off a cliff in situations where it should not):

16 ROPs / 256-bit
625 Mhz core / 1500 shader
5 clusters / 80 shaders / 20 TMUs
512MB 2Ghz GDDR3

I think that would be a nice card for the price and probably cost less to manufacture overall. The clocks shouldnt be a big deal considering what they're hitting on 90nm at the moment.

Geo

20-Mar-2007, 03:29

With regard to your comment about the number of units being correlated with memory size...exactly which units do you mean? AFAIK the number of ROPs determines bus-width and this is all decoupled from the shading/texturing core.

It certainly is the case that ROPs are correlated to # of memory chips. As to whether shader/texture units are. . .the data points on that are pretty thin on the ground, wouldn't you say? But what we do have. . .

Geo

20-Mar-2007, 03:37

Oh lol, that's what that means? Have been trying to figure that out for a while now :) Yeah, I've always thought there was a big gap to fill between the GTS and GTX. I'm still giving Nvidia a chance to play their full hand to see if that sheds some light on that thinking.

Well, that's more or less where I was going with my comment about a GTX 384MB. Surely by the time the 8800 line was released they knew that ATI was not breathing down their neck. Why wouldn't you go with a GTS in those circumstances when you've got a damn big chip to begin with? Very similar to the 7800 situation, right? We shall see. I have no information that a GTX 384MB is coming, but if one shows up you can be sure I'll point at this series of posts. :cool:

trinibwoy

20-Mar-2007, 03:41

It certainly is the case that ROPs are correlated to # of memory chips. As to whether shader/texture units are. . .the data points on that are pretty thin on the ground, wouldn't you say? But what we do have. . .

Yeah but that's not unique to G80 - AFAIK that's the case for all architectures.....? And it determines the number of chips - not memory size (320MB/640MB GTS?) And if you're targeting a certain performance level you're not necessarily stuck with a certain number of ROPs - see my suggestion for my version of the GTS above (well maybe a GT given Nvidia's current naming scheme).

Geo

20-Mar-2007, 03:45

Yeah but that's not unique to G80 - AFAIK that's the case for all architectures.....? And it determines the number of chips - not memory size (320MB/640MB GTS?) And if you're targeting a certain performance level you're not necessarily stuck with a certain number of ROPs - see my suggestion for my version of the GTS above.

Certainly your suggested sku makes a lot of sense at some point. That might be a Wacky Jen-Hsun special in 2008 tho. . or at least late 2007.

trinibwoy

20-Mar-2007, 03:53

Not sure if this was posted - a shot from an apparent R600 physics demo ( looks like a real-life pic to me for some reason :oops: )

http://www.hexus.net/content/item.php?item=8013

And Guru3D with more bad news - http://www.guru3d.com/article/article/422/3/

AMD decided to pull the plug on these presentations as they really do not have anything to show. The way it is right now is that the press-briefings have been delayed towards likely the end of April. I still believe we'll see four models of the R600 based chipset. I think one of the biggest issues right now is the low yield levels due to data leakage (successful working products). It was rumored at the CeBIT that at the initial launch only 30,000 products will be available "World-wide"! That's both retail and OEM (which takes easily half of the products).

icecold1983

20-Mar-2007, 04:16

wow, keeps getting worse, how long till it leaks out that its been delayed till at least june?

Geo

20-Mar-2007, 04:33

wow, keeps getting worse, how long till it leaks out that its been delayed till at least june?

I think I'm up for bets no-later-than May 31st for at least one R600 sku availablity. You up for that? :smile:

silent_guy

20-Mar-2007, 04:39

Unlocking quads has been a thing of the past for some time. Something about they're laser cut now.

Just a side note: laser tuning was (is?) something that was done to calibrate high precision analog stuff, such as ultra-accurate A/D converters. Internal chip configuration management is typically done with fuses. It's typically used for on-chip RAM configuration (replacement of failing rows or columns) after test (see here (http://www.viragelogic.com/render/content.asp?id=191)), but it can be used for pretty much anything that needs to be enabled or disabled permanently in a chip.

You can blow a fuse by applying an excess voltage, so it's possible to program things after packaging (though that's typically not what's done for memory repair.) And contrary to laser tuning (expensive!), it virtually comes for free.

Edit:
Even so. It's likely to be cheaper and quicker to control such by how many memory chips you put on the pcb than going back to laser cut healthy chips.
How do you prevent rogue hardware vendors from buying lower grade chips and put them on high level PCB's? Maybe not a problem of GPU's where, I suppose, the vendor knows the OEM quite well, but a real issue for more off-the-shelve high volume chips. (cfr rebranding of microprocessors.)

INKster

20-Mar-2007, 04:47

I think I'm up for bets no-later-than May 31st for at least one R600 sku availablity. You up for that? :smile:

Surely you wouldn't define 15000 retail units worldwide as "availability" ?
:sad:

Geo

20-Mar-2007, 05:17

Surely you wouldn't define 15000 retail units worldwide as "availability" ?
:sad:

It's an interesting point. I'm pretty self-centered on that standard, really, and determine these things on my ability to order one online in less than an hour of looking for one in the first few days from "the usual suspects" in North America. Depending on popularity, it's not unusual to have initial availaibilty, then a lull, then more availability. This would be why I ordered my 8800 GTX pretty quickly. . .

icecold1983

20-Mar-2007, 05:50

geo how much $$ are you looking to bet? ill take the bet on the following terms. the fastest r600 having widespread availability before/after may 31st. if its before u win, if its after i win.

Geo

20-Mar-2007, 05:59

geo how much $$ are you looking to bet? ill take the bet on the following terms. the fastest r600 having widespread availability before/after may 31st. if its before u win, if its after i win.

Nah. By that standard "the fastest" 7800 wasn't available until November of 2005, even tho the 7800 was launched in June.

icecold1983

20-Mar-2007, 06:02

Nah. By that standard "the fastest" 7800 wasn't available until November of 2005, even tho the 7800 was launched in June.

uhhh the 7800 gtx wasnt available in june??? unless u mean the gtx 512 which obviously doesnt apply. when i say fastest i mean fastest of the original model line. for example if were were betting on the r520, the fastest would have been the x1800xt. g80? 8800gtx. im obv not going to change the terms to include a refresh part.

Shtal

20-Mar-2007, 07:26

The whole thing about R600 delay over delay and delay and another delay, who cares anyway anymore....
As lone as Nvidia Geforce 8900GTX 65nm-G81/NV55 does not hit the store shelves anything between June - July 2007. It be alright
Geforce 8800Ultra - I believe is not going to be a real treat/enemy to R600.

The way I look at this 65nm version of R600. I believe that ATI started working 80nm and 65nm about same time frame for their first development; and 65nm was probably reserved for R650 refresh. So it could mean that R600 is to big for 80nm and ATI has no choice but to skip 80nm due to the low yields.

Gelanin

20-Mar-2007, 07:46

More R600 rumors and less G80 chatter please :D

rwolf

20-Mar-2007, 07:49

Relevance to how far up their ass VR Zone's head has historically been regarding ATI yields?

Perhaps it had a quick demise because of yields. Just speculation of course.

IbaneZ

20-Mar-2007, 08:26

And Guru3D with more bad news - http://www.guru3d.com/article/article/422/3/

$2000 on Ebay. I better start saving. :wink:

volt

20-Mar-2007, 12:07

I can confirm, AMD didn't let their partners show the board. Raymen was pretty pissed about it I recon.

trinibwoy

20-Mar-2007, 13:05

I can confirm, AMD didn't let their partners show the board. Raymen was pretty pissed about it I recon.

I would love to know the strategic reason for that kinda move.....

vertex_shader

20-Mar-2007, 14:36

I would love to know the strategic reason for that kinda move.....

Launch not coming anytime soon (2-3weeks), this is the reason, AIB's only show the cards for the biggest business partners , the card is exist and working :smile: (not duke nukem forever style anymore :grin: )

CarstenS

20-Mar-2007, 15:32

I would love to know the strategic reason for that kinda move.....
To strategically piss of your AIBs? Well, less competition keeps prices up (or makes customers go to the "evil green").

INKster

20-Mar-2007, 15:33

To strategically piss of your AIBs? Well, less competition keeps prices up (or makes customers go to the "evil green").

Both "greens" are now too similar to distinguish. :D

trinibwoy

20-Mar-2007, 15:53

To strategically piss of your AIBs? Well, less competition keeps prices up (or makes customers go to the "evil green").

I don't think the AIB's are too concerned about keeping prices up in the niche markets. They are much more interested in having something new and shiny to flog to OEM's and the mainstream retail market. Fudo is even claiming that Nvidia postponed the G8x releases because of AMD's delays. I don't see the point in that either....

jamesnmandy

20-Mar-2007, 16:33

you guys missed your chance to get your hands on a prototype for $50 + shipping

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&item=160095063471&ru=http%3A%2F%2Fsearch.ebay.com%3A80%2Fsearch%2Fse arch.dll%3Fsofocus%3Dbs%26sbrftog%3D1%26from%3DR10 %26submitsearch%3DSearch%26satitle%3D160095063471% 26sacat%3D-1%2526catref%253DC6%26sargn%3D-1%2526saslc%253D2%26sadis%3D200%26fpos%3DZIP%252FP ostal%26ftrt%3D1%26ftrv%3D1%26saprclo%3D%26saprchi %3D%26fpos%3DZIP%252FPostal%26lsot%3D%26fsop%3D1%2 526fsoo%253D1%26coaction%3Dcompare%26copagenum%3D1 %26coentrypage%3Dsearch%26fgtp%3D%26fpos%3DZIP%252 FPostal%26fvi%3D1

reported by the Inq :oops:

Geeforcer

20-Mar-2007, 16:46

jamesnmandy

20-Mar-2007, 16:50

I can't tell if you are being sarcastic or not... in any case: Link (http://forum.beyond3d.com/showpost.php?p=949738&postcount=795)

not at all......i first read about it this morning at the Inq, which is why i mentioned it, so i checked the last two pages of posts in this thread and didn't see mention of it, so for what it's worth, i linked to it

http://www.theinquirer.net/default.aspx?article=38360

alexsok

20-Mar-2007, 16:58

I think it's safe to say now that all the demos of G80 & R600 are equally unimpressive.
Guess they wasted all their energy on the cards themselves - at least we should be thankful for that :)

Sobek

20-Mar-2007, 17:28

Man those Inq reporters are MORONS!

Closed because of no interest? Did he check the bid history? How does he know it's one of the first revisions? How does he know it has exactly 1gb of memory? How does he know for sure it's an 80nm core (granted that's a given, but fancy-pants Theo cites it like fact).

Argh, when will they die.

Ike Turner

20-Mar-2007, 17:35

Argh, when will they die.

When people stop clicking links pointing to their site?
I think I've been there a total of 1! time in a year...I fell clean

INKster

20-Mar-2007, 17:39

Razor1

20-Mar-2007, 17:47

well early drivers but not like the g80 has any performance drivers come out yet either.

INKster

20-Mar-2007, 17:50

well early drivers but not like the g80 has any performance drivers come out yet either.

That is not a credible excuse, given the following:
http://www.extremetech.com/article2/0,1697,2103631,00.asp

They say their "Vista Drivers are top-notch" and they are "not struggling with performance issues".
Since they've been banging on Nvidia's head because of their slow and buggy drivers since Vista came out, we shouldn't be excusing them for doing the same thing now with their own DX10 driver.
Both IHV's had plenty of time to work with Microsoft on the issue.
A large part of DX10 is what it is because of the IHV's. Why should they (any of them) be forgiven now ?

To me, the major loss seems to be that the news about cheaper versions (XT, XL) are not that amusing.

overclocked_enthusiasm

20-Mar-2007, 18:24

So we are looking at a 2-3% difference in these leaked benchmarks with AMD's drivers arguably more mature? I have to say that IF these scores are indicative of what we can expect out of R600 vs G80 than we can officially post the following job opening at AMD:

Executive Vice President of Visual and Media Businesses

A 65nm R6xx might be needed to meet or beat the G80 refresh as it seems the 80 nm R600 is incapable of little more than parity with G80. If it wasn't so damn sad it would laughable at this point. Where on earth did that R3xx team go?

Razor1

20-Mar-2007, 18:34

That is not a credible excuse, given the following:
http://www.extremetech.com/article2/0,1697,2103631,00.asp

They say their "Vista Drivers are top-notch" and they are "not struggling with performance issues".
Since they've been banging on Nvidia's head because of their slow and buggy drivers since Vista came out, we shouldn't be excusing them for doing the same thing now with their own DX10 driver.
Both IHV's had plenty of time to work with Microsoft on the issue.
A large part of DX10 is what it is because of the IHV's. Why should they (any of them) be forgiven now ?

To me, the major loss seems to be that the news about cheaper versions (XT, XL) are not that amusing.

I don't understand the vista driver performance now thing at all, lots of people came down on nV's g80 yes, but when xp came out there were a butt load of issues too, this isn't something out of the ordinary at all. Its going to take time for maturity especial with graphics drivers since Vista is so dependent on graphics cards, IHV's have to be much more cautious I agree but underlying problems with the shift to Vista is more of the cause of performance issues then IHV's not paying enough time to create drivers. Its pretty obvious ATi's drivers for previous generations are good as with nV's drivers pre g80, but again, new GPU's new OS, many new issues, think Geo mentioned this earlier in another thread.

Kaotik

20-Mar-2007, 18:39

I don't understand the vista driver performance now thing at all, lots of people came down on nV's g80 yes, but when xp came out there were a butt load of issues too, this isn't something out of the ordinary at all. Its going to take time for maturity especial with graphics drivers since Vista is so dependent on graphics cards, IHV's have to be much more cautious I agree but underlying problems with the shift to Vista is more of the cause of performance issues then IHV's not paying enough time to create drivers. Its pretty obvious ATi's drivers for previous generations are good as with nV's drivers pre g80, but again, new GPU's new OS, many new issues, think Geo mentioned this earlier in another thread.

Actually ATIs Vista drivers for DX9 cards have been better than nV's for DX9 cards, at least before latest few betas from nV, dunno for sure about them.
It's a bit dated, but already with Cat 7.1 ATI was actually faster in few games on Vista compared to XP, while nVidia with the drivers out then was slower in everything.

20-Mar-2007, 18:41

If you listen to what was said at AMDs latest Q&A press conference (http://www.gearlive.com/news/article/159-amd-press-qa) it doesn't seem that they're too worried about R600s performance compared to G80 nor do they give any signs of posting new job openings anytime soon. ;) They're even talking about R650 and a silent multi GPU/CPU system that's gonna impress us all.

There's been a lot of misinformation at CeBit. It's a pattern that happens every time a new product is about to launch and I've got a feeling that it's gotten worse in the last few years.

overclocked_enthusiasm

20-Mar-2007, 18:57

The problem with using AMD's Q&A at either the recent CC in January or the Analyst day in December as proof of anything is that they were proven wrong both times on many counts. AMD is recently lacking credibility in terms of launch dates, market share, ASPs, Gross Margins and revenue forcasts. Why should we therefore have confidence in their performance estimates?

I hate to keep bringing AMD's financial/managerial issues to light in this thread but they are just so attached to the product itself in terms of launch dates, performance estimates and actual ability to deliver.

*EDIT Added gross margins

trinibwoy

20-Mar-2007, 19:09

Gotta agree, I'm not sure AMD's bullish attitude on anything related to R600 is of any relevance nowadays.

vertex_shader

20-Mar-2007, 19:09

Here are some more R600 pics (and benchmarks, finally):

http://www.vr-zone.com/?i=4830

As i expected, the difference from the 8800 GTX is not that big.
That's surprising, given the 800MHz core clock, 512bit bus, and power consumption on a smaller 80nm process...

The pictures very bad photoshoped, some visible watermark left in the card, one AIB ass will be kicked.
AMD not give out for any AIB partner performance drivers.

icecold1983

20-Mar-2007, 19:15

no-X

20-Mar-2007, 19:15

So we are looking at a 2-3% difference in these leaked benchmarks with AMD's drivers arguably more mature? I have to say that IF these scores are indicative of what we can expect out of R600 vs G80 than we can officially post the following job opening at AMD:

Executive Vice President of Visual and Media Businesses
3DM06 was always quite NV-friendly. 7800GTX/X1800XT have almost the same score, but performace in games is somewhere else. Btw. 3DMark doesn't use FSAA, so R600's 512bit bus cannot show it's power.

20-Mar-2007, 19:16

The problem with using AMD's Q&A at either the recent CC in January or the Analyst day in December as proof of anything is that they were proven wrong both times on many counts.

That's why I was linking to the Q&A at the beginning of this month when they presented their Teraflop-in-a-Box, which was held after they cancelled the R600 Editors Day in Amsterdam. So it's fairly recent and thus imo much more trustworthy.

leoneazzurro

20-Mar-2007, 19:16

That is not a credible excuse, given the following:
http://www.extremetech.com/article2/0,1697,2103631,00.asp

They say their "Vista Drivers are top-notch" and they are "not struggling with performance issues".
Since they've been banging on Nvidia's head because of their slow and buggy drivers since Vista came out, we shouldn't be excusing them for doing the same thing now with their own DX10 driver.
Both IHV's had plenty of time to work with Microsoft on the issue.
A large part of DX10 is what it is because of the IHV's. Why should they (any of them) be forgiven now ?

To me, the major loss seems to be that the news about cheaper versions (XT, XL) are not that amusing.

VR Zone did not say the test was running on Vista, and anyway 3Dmark score of 95xx for G80 is on XP. So maybe Vista drivers are "OK", XP drivers are lacking.
Moreover, we have no hints on the relative performance in AA-AF modes (R600 has higher bandwidth ) and IQ settings. So, IMHO it's better to wait the tests on the shipping parts.

Natoma

20-Mar-2007, 19:18

Here are some more R600 pics (and benchmarks, finally):

http://www.vr-zone.com/?i=4830

As i expected, the difference from the 8800 GTX is not that big.
That's surprising, given the 800MHz core clock, 512bit bus, and power consumption on a smaller 80nm process...

Meh, useless benchmark. No AA/AF settings mentioned. Vista? XP? Bleh.

Never make proclamations on unverifiable #s lest you get burned. :wink:

3dilettante

20-Mar-2007, 19:18

That's why I was linking to the Q&A at the beginning of this month when they presented their Teraflop-in-a-Box, which was held after they cancelled the R600 Editors Day in Amsterdam. So it's fairly recent and thus imo much more trustworthy.

Just because it's later than their earlier misleading statements doesn't mean it has to be more trustworthy.

Lies and misdirection haven't gotten any harder to do since the last fiscal quarter.

Shtal

20-Mar-2007, 19:33

Next, we got hold of some preliminary benchmarks figures of the R600 XTX card with core clock at 800MHz vs a GeForce 8800 GTX card. Using a Core 2 Extreme 2.93GHz processor on an Intel 975X board, the 3DMark06 score at 1600x1200 resolution is 97xx on the R600XTX compared to 95xx on the 8800GTX.

I still think R600 has 16 ROP's.

no-X

20-Mar-2007, 19:38

do you remember? :-)

What do you see as being the balance of bandwidth utilisation between textures and ROP's at the moment?

[Eric Demers] Right now, they seem to balance at 1:1 (TEX:ROP), but the trend is towards lowering ROPs, in general. The reality is that shading per pixel is increasing, which usually means many ALUs and many textures per pixel, as well as many cycles per pixel. Since we need only 1 ROP per cycle per pixel, effectively, the ROP throughput requirement is going down on new apps. An RV530 is a prime example � It doesn't have more ROP than the R515, but having triple the shading and double the Z, it's around 2x the speed of the R515 in a lot of cases. Finally, with HDR becoming more popular, the BW requirements of these pixels is high, so that the ROP throughput is possibly going down, even though each operation is 2x wider.

http://web.archive.org/web/20060208001436/http://www.beyond3d.com/reviews/ati/r580/int/index.php?p=02

Take a read of the interview again, carefully, and bear in mind what items are likely to be on his mind when he's replying (given that R580 is a historic item to him at that point).
http://forum.beyond3d.com/showpost.php?p=767373&postcount=118

Razor1

20-Mar-2007, 19:42

Just because it's later than their earlier misleading statements doesn't mean it has to be more trustworthy.

Lies and misdirection haven't gotten any harder to do since the last fiscal quarter.

True I dont' remember nV ever stating the nv30 was going to be a shit box compared to the r300 :wink: they actually tried to exclude the r300 and performance comparisions and if I remember correctly nV mentioned it was going to faster much faster then the gf4 in at least one occassion.

David Kirk said this before the nv30 launch.
that NV30 will have more than two times the performance of GeForce4, with 51 billion floating point operations per second (51 gigaflops) in the pixel shader alone. The chip has 125 million transistors, which is three times the number on the Pentium 4. Nvidia, through its manufacturing partner Taiwan Semiconductor Manufacturing Co., is using a 0.13-micron manufacturing process to reduce the physical size of the chip and achieve better efficiency. By comparison, the top-end Radeon chip has 110 million transistors, and the GeForce4 Ti 4600 has 63 million.

Geo

20-Mar-2007, 20:20

hope that bench is fake

So much for insisting on AA/AF numbers, eh?

Geo

20-Mar-2007, 20:22

David Kirk said this before the nv30 launch.

Well, let's not get into R300 vs NV30 yet again. But it was (and I think still is) traditional for Nvidia to compare to their own previous product at launch when doing those kind of comparisons.

no-X

20-Mar-2007, 20:32

R600 is 80 nanometre - Various sources confirms

We spoke with dozen of contacts and we can now confirm that R600 is a 80 nanometre chip. It is not 60 nanometre and it won't be 65 nanometre as going from 80 nanometre to 65 requires complete redesign of chip.

ATI doesn't have the time for it and we don't believe that AMD / ATI could get away deceiving dozens of its partners and OEM's with news that R600 is 80 nanometre, until it suits them.

R600 was and is 80 nanometre and when it launches it will be 80 nanometre. Remember G80, Geforce 8800 series is 90 nanometre and it still runs great. So if you do you homework well even the big chip works well.
Sorry to disappoint you but R600 will be re branded and probably cheaper than everyone expects but its very unlikely that we are waiting for G80 killer.

http://www.fudzilla.com/index.php?option=com_content&task=view&id=221&Itemid=34

trinibwoy

20-Mar-2007, 20:40

I'm shocked and amazed....

Frank

20-Mar-2007, 20:52

R600 is 80 nanometre - Various sources confirms

http://www.fudzilla.com/index.php?option=com_content&task=view&id=221&Itemid=34
Then again, AMD did do a straight shrink with their CPUs, so they do have the software and macros to make that happen relatively painless.

3dilettante

20-Mar-2007, 20:58

Then again, AMD did do a straight shrink with their CPUs, so they do have the software and macros to make that happen relatively painless.

What AMD does with its CPU lines has little to do with its GPU lines. There is no commonality in engineering, design, or fab process.

It's also hard to say that AMD's much-delayed 65nm process that still clocks lower than the 90nm top end allowed for a painless shrink.

icecold1983

20-Mar-2007, 20:58

So much for insisting on AA/AF numbers, eh?

why do u assume im expecting this score to be 30% higher? i dont expect r600 to be 30% faster with no aa/af, but i expect it to be faster than this comparison shows.

vertex_shader

20-Mar-2007, 21:02

I think I'm up for bets no-later-than May 31st for at least one R600 sku availablity. You up for that? :smile:

What about April 24th? :wink:

Frank

20-Mar-2007, 21:09

What AMD does with its CPU lines has little to do with its GPU lines. There is no commonality in engineering, design, or fab process.

It's also hard to say that AMD's much-delayed 65nm process that still clocks lower than the 90nm top end allowed for a painless shrink.
Good points.

20-Mar-2007, 21:29

Before we know it, Fudzilla or The Inq will probably post that the R600 is a 55nm chip. ;)

Some other info about the roadmap of the mobile parts.

M76XT, the highend 65nm CrossFire capable DX10 chip with UVD, will be replaced by M88 in the first half of 2008. M88 is made on the 55nm process and supports DX10.1 and PCIe v2.

In the midrange the M76, 65nm CF capable DX10 chip with UVD will be replaced by M86, which is also made on 55nm and supports DX10.1 and PCIe v2.

The lowend will see M72/M74 chip being replaced by M82, which is also made on 55nm and has displayport support, DX10.1 support and PCIe v2.

Arty

20-Mar-2007, 23:26

M76XT = R600 on 65nm?

Geo

20-Mar-2007, 23:58

why do u assume im expecting this score to be 30% higher? i dont expect r600 to be 30% faster with no aa/af, but i expect it to be faster than this comparison shows.

Why? What advantage do you expect R600 to have that would apply to this specific app at that specific res with no aa and no af to make it a disappointment for it not to be signficantly faster than G80? How much faster were you expecting, if you weren't expecting the 30% you'd reserved for AA/AF intensive scenarios? You're getting to some pretty fine gradations in your expectations for an unreleased architecture that's never been performance profiled in public.

turtle

21-Mar-2007, 00:01

M76XT = R600 on 65nm?

Doubtful.

M*6 = Mobile midrange (M56 was X1600M and M66 was X1700 respectfully). I imagine it's just a high-clocked mobile Rv630. The regular M66 is probably M66-P, if the past is any indication.

High-end parts translated to the mobile sector are M*8 (M58 was X1800 Mobile, IIRC M68 was X1900 Mobile).

I think, in other words what he's trying to say is we should not be expecting a laptop R600. Too much heat, too much power, and that points as another indication the 65nm R600 part not happening, or if it does (as R650 or whatever) it will be too close to next gen (or not with good enough yields) to produce a mobile counterpart.

To put a happy face on it though, it looks like we can expect a Mobile R700 :). I was kind of hoping next-gen (2008,R7xx) would be 45nm as ATi announced their lineup as being on that process in 2008, but if CJ is correct, it looks like it's headed to 55nm...perhaps with a 45nm refresh being as likely as seeing a R600 refresh on 65nm. All that being said, the situation is also more-than-likely still fluid.

Russell

21-Mar-2007, 00:17

R600 is 80 nanometre - Various sources confirms
http://www.fudzilla.com/index.php?option=com_content&task=view&id=221&Itemid=34

The most interesting part from that article is:

"Sorry to disappoint you but R600 will be re branded and probably cheaper than everyone expects but its very unlikely that we are waiting for G80 killer."

Cheaper is good. I'm sure it will still be fast.

icecold1983

21-Mar-2007, 00:21

Why? What advantage do you expect R600 to have that would apply to this specific app at that specific res with no aa and no af to make it a disappointment for it not to be signficantly faster than G80? How much faster were you expecting, if you weren't expecting the 30% you'd reserved for AA/AF intensive scenarios? You're getting to some pretty fine gradations in your expectations for an unreleased architecture that's never been performance profiled in public.

i expect it to be a faster part in general. 2% is meaningless. while i dont expect it to pull ahead to the extent i think it should in high res with aa/af, it should still show noticeable increases. 80nm, at least 6 months later launch time, high high clocks, 500+ flops v 350ish etc etc.

Creig

21-Mar-2007, 00:24

Please stop quoting Inq/Fuad.

Just... please.

INKster

21-Mar-2007, 00:30

Please stop quoting Inq/Fuad.

Just... please.

Well, since they started (http://www.theinquirer.net/default.aspx?article=38300) the rumor in the first place, it's only fitting they should be the ones to bury it. ;)

Kaotik

21-Mar-2007, 00:32

Well, since they started (http://www.theinquirer.net/default.aspx?article=38300) the rumor in the first place, it's only fitting they should be the ones to bury it. ;)

It has been buried by other sources as well, even if it wasn't in such big manner as another news "article
but anyway, wasn't it k-hardware or something that said first it would be 65nm, not inquirer?

Razor1

21-Mar-2007, 00:49

INKster

21-Mar-2007, 00:58

It has been buried by other sources as well, even if it wasn't in such big manner as another news "article
but anyway, wasn't it k-hardware or something that said first it would be 65nm, not inquirer?

K-hardware's google-auto-translation was very ambiguous, to say the least.
So i'm willing to give them the benefit of the doubt. ;)

Doomtrooper

21-Mar-2007, 00:59

Whoa. He's alive! :wink:

I think that demo stomps Adrienne pretty well. Tho I thought Ruby's head was a bit narrow and long in a few scenes.

It's hard to argue that ATI doesn't put more effort into these things than NV does.

Edit: Here, I'll be a little provocative. I'd estimate based on action, scenes, art assets, and length that the number of person hours directly involved in making "Whiteout" (directly involved --no hours from the driver team) were more than Froggy, Smoke, Cascades, and Adrienne combined. Agree/disagree?

Oh yes still alive ;)
Agreed, although I personally wish some of these demos had a screensaver mode. I also believe that both the IHV heavyweights are going to have a hard time 'impressing the consumer' just do to technological equality.

Some of the leaked benchmarks mean nothing.

overclocked_enthusiasm

21-Mar-2007, 01:56

Kaotik

21-Mar-2007, 02:02

It's not 65nm, #1 + mysterious reason why they originally pushed it forward so much and then got the idea about "all-in-one" launch

Geo

21-Mar-2007, 02:21

It's not 65nm, #1 + mysterious reason why they originally pushed it forward so much and then got the idea about "all-in-one" launch

Looking at Tridam's report, it's not real clear just how many R600 skus they are launching with, so arguably they may have still rushed it. Personally, I scratched the liklihood of it being a clearly dominant part when we got a look at G80. Then when 512-bit became clear I upped that to being possibly dominant in some narrow selected scenarios (some of which might be quite tasty, however). That's still pretty much where I am. I've just never believed in putting that level of expectations on a product in advance. Competency and competitiveness we all have a right to expect; greatness should be judged after you've actually seen it.

R300King!

21-Mar-2007, 02:30

My head hurts...so WTF is going on with R600?

1. R600 is 80 nm and they waited to launch it with the rest of the R600 family and/or early Barcelona production samples?

2. R600 is actually R650 and is 65 nm as R600 at 80 nm was just unworkable and all the AIB partners are in the dark?

3. R600 is going to be targeted into a lower price point (I am assuming because of yield issues) and R650 at 65 nm will then compete against the G80 flagship refresh in the fall leaving the enthusiast segment to Nvidia unopposed until then?

4. Fuad and Inq are so utterly and hopelessly confused that none of the above is true.

I am putting my money on #1 with a slight chance of #2 and sprinkled with some #4. It will be funny to see what other explanations we get in the next month or so.

I say #2 & #3 with the first part of #4 ;)

Fuad's doin his disinformation campaign. Not lying but telling us R600 is 80nm and not 65nm. Yes, this we all know. If you change from 80nm to 65nm more than likely you'd need to do a whole redesign and that chip would be called something else...most likely R650 or R680.
But I think we'll see some R650s and the OEMs will see the R600s first. When 65nm yields for the R650 improve then we'll see wider availability for everyone and the OEMs will get the R650s too.

icecold1983

21-Mar-2007, 03:01

Looking at Tridam's report, it's not real clear just how many R600 skus they are launching with, so arguably they may have still rushed it. Personally, I scratched the liklihood of it being a clearly dominant part when we got a look at G80. Then when 512-bit became clear I upped that to being possibly dominant in some narrow selected scenarios (some of which might be quite tasty, however). That's still pretty much where I am. I've just never believed in putting that level of expectations on a product in advance. Competency and competitiveness we all have a right to expect; greatness should be judged after you've actually seen it.

why would g80 being a great product prevent ati from launching an even better one? seems they aimed higher.

Geo

21-Mar-2007, 03:16

why would g80 being a great product prevent ati from launching an even better one? seems they aimed higher.

Depends on your frame of reference, I suppose. A product that is as good as a great product is a great product, isn't it? I suspect that's not what other people mean, however. Someone stated that a product that was 20% faster than a great product would be disappointing, for instance. What would you call a great product that is as much greater than another great product was greater than whatever you were comparing it to? Umm, "really great"?

Yes there are limitations, and different priorities. We've seen, for instance, that HDMI sound seems to be in R600 silicon. I haven't heard it's in G80s. There's been indications that UVD may be in there too, so that'd be another chunk of silicon of unknown size (tho also unknown comparison to PV). There is no evidence they went the NVIO route and sucked that chunk of silicon out of the main die. Doing 512-bit must have cost them some mc control logic that didn't go elsewhere.

Razor1

21-Mar-2007, 03:22

heh that first paragraph gave me a headache :wink:

Geo

21-Mar-2007, 03:29

heh that first paragraph gave me a headache :wink:

Those last two sentences! :lol:

caffeinated

21-Mar-2007, 03:34

heh that first paragraph gave me a headache :wink:

Me too. I'm beginning to wonder how much wood could a woodchuck chuck if a wood chuck could chuck wood, and that just isn't right when I'm on a forum about 3D hardware :grin: .

Geo

21-Mar-2007, 04:48

Me too. I'm beginning to wonder how much wood could a woodchuck chuck if a wood chuck could chuck wood, and that just isn't right when I'm on a forum about 3D hardware :grin: .

This is a difficult question. I'd like to know more about the number of woodchucking units the particular woodchuck you have in mind has, whether they are tied on a 1-1 basis to his wood culling units or are disassociated to hide latencies, and if his woodchucking and culling scheduling units are robust, or 1st generation, or what.

icecold1983

21-Mar-2007, 04:51

Geo

21-Mar-2007, 04:56

geo would u agree that nvidia usually doesnt push the envelope and instead focuses on great margins? its usually ati who pushes the envelope of performance/clocks. nv30 excluded.

if a product launches half a year or more after another product at a higher price(its very unlikely that r600 will be cheaper than g80 at launch), that product has to be substantially better not to be a disappointment. otherwise u waited 6 months to pay more for something u had over half a year ago.

I think you're making assumptions on prices that may or may not be accurate. There are indications otherwise. Look at Richard's statement and recognize what broader means. Look at what Kombatant was quoted as saying at Rage3D.

I think Nvidia was fairly scarred from NV30, and claimed to learn some lessons. One of the major ones was "go wider rather than faster". I think it's not clear just yet how it works out vs R600, given clock domains and such. How do you square "not pushing the envelope" with 1.35GHz shaders? That's roughly, what? 60% or so faster than any piece of any previous GPU, from any manufacturer?

icecold1983

21-Mar-2007, 05:04

well im making assumptions on the price yes, but i will be shocked if when r600 launches, the street price isnt higher than g80s street price.

when g80 runs the close to the same temp as r580, and barely uses more power its hard to believe that they couldnt have done more with the card if they wanted to.

edit - just checked pricewatch, 8800gtx as low as 543, if r600 can be had that cheaply at launch, awesome. but im not expecting it. it also wouldnt be a stretch to imagine that g80s price will go down when r600 ships.

INKster

21-Mar-2007, 05:06

I think you're making assumptions on prices that may or may not be accurate. There are indications otherwise. Look at Richard's statement and recognize what broader means. Look at what Kombatant was quoted as saying at Rage3D.

I think Nvidia was fairly scarred from NV30, and claimed to learn some lessons. One of the major ones was "go wider rather than faster". I think it's not clear just yet how it works out vs R600, given clock domains and such. How do you square "not pushing the envelope" with 1.35GHz shaders? That's roughly, what? 60% or so faster than any piece of any previous GPU, from any manufacturer?

Yes, but since they went all scalar, it's likely that clock speed was planned because scalar made it easier to achieve.
So, we have to frame that 1.35GHz figure in the architecture of a specific part of the GPU, not necessarily in the need for speed vs the competition of the moment.
I mean, it was about efficiency vs brute force, not exactly what the comparatively huge 1350MHz number might have made us suppose at first glance.

well im making assumptions on the price yes, but i will be shocked if when r600 launches, the street price isnt higher than g80s street price.

when g80 runs the close to the same temp as r580, and barely uses more power its hard to believe that they couldnt have done more with the card if they wanted to.

edit - just checked pricewatch, 8800gtx as low as 543, if r600 can be had that cheaply at launch, awesome. but im not expecting it. it also wouldnt be a stretch to imagine that g80s price will go down when r600 ships.

Well, before 7800 GTX 512 came along, even the regular 7800 GTX'es were very difficult to overclock from the standard 430MHz up to ~500MHz with stock cooling.
Then, suddenly, the 7800 GTX 512 showed up with a 550MHz core, still built on the old 110nm half node, and sporting a redesigned cooler that was actually in use on the earlier QuadroFX 4500.

Geo

21-Mar-2007, 05:06

icecold1983

21-Mar-2007, 05:15

Geo

21-Mar-2007, 05:24

edit - just checked pricewatch, 8800gtx as low as 543, if r600 can be had that cheaply at launch, awesome. but im not expecting it. it also wouldnt be a stretch to imagine that g80s price will go down when r600 ships.

I said earlier I don't have any pricing info, but I've seen at least one published report of R600 launching at 500 Euros. That's probably the XT tho.

Pricing is almost never driven by cost, actually. Only profits are. :lol:

caffeinated

21-Mar-2007, 05:30

This is a difficult question. I'd like to know more about the number of woodchucking units the particular woodchuck you have in mind has, whether they are tied on a 1-1 basis to his wood culling units or are disassociated to hide latencies, and if his woodchucking and culling scheduling units are robust, or 1st generation, or what.

You can bet that they are chucking amazing...amounts of wood, that is. :grin:

Back to the topic:

I said earlier I don't have any pricing info, but I've seen at least one published report of R600 launching at 500 Euros. That's probably the XT tho.

Pricing is almost never driven by cost, actually. Only profits are. :lol:

And how profitable will R600 be after being late to the party, and possibly requiring finagling to improve performance and cooling?

Pete

21-Mar-2007, 05:32

A thought occurs: If R600 will launch at such limited quantities as some sites suggest, why bother paring it down to 512MB GDDR3? Why not just launch as high as possible, accepting limited initial yields? Wouldn't this also parallel the rumors that GDDR4 is in short supply and so kill two birds (widespread availability) with one stone (exclusivity at the high end)? B/c if it's widespread they want, they'll take care of that with the RV6x0 parts.

I'm just not sure how limited availability b/c of low yields (on whatever process :razz:) meshes with a delay to release what should be a more popular model.

icecold1983

21-Mar-2007, 05:32

I said earlier I don't have any pricing info, but I've seen at least one published report of R600 launching at 500 Euros. That's probably the XT tho.

Pricing is almost never driven by cost, actually. Only profits are. :lol:

what would that equate to in us dollars?

Tim Murray

21-Mar-2007, 05:41

jamesnmandy

21-Mar-2007, 05:48

My head hurts...so WTF is going on with R600?

It will be funny to see what other explanations we get in the next month or so.

It will be funny if a R600 part ever sees the light of day at this rate......i think they basically walked away from the R600 launch, left Nvidia to dominate the high end market, and hope to build up enough momentum with a "better late than never" R650/R700 launch to get back into the race.....

i visit this thread daily just as amusement to see how many people spend so much time speculating about a product that seems to be doomed while there are products out there already that will do what they want, i was personally excited about R600 like many months ago.....how long has it been?.....now i am bored with the mention of a product which has never been launched, they need to shit or get off the proverbial pot

Geo

21-Mar-2007, 06:01

And how profitable will R600 be after being late to the party, and possibly requiring finagling to improve performance and cooling?

Well, they have some stuff going for them tho. They appear to have 80nm vs 90nm. They have not needing a separate "NVIO" chip. Do you know what 1.35GHz in parts vs 800MHz for the whole (for example) does to yields? I surely don't. They might have some advantage from the asymmetry of the framebuffer size in the "one down" sku where NV cleaned up with 6800 GT (768MB vs 512MB).

what would that equate to in us dollars? Well, that can be variable. Often tho Euros = dollars for pricing. Not always, but often.

Ailuros

21-Mar-2007, 06:28

Well, they have some stuff going for them tho. They appear to have 80nm vs 90nm. They have not needing a separate "NVIO" chip. Do you know what 1.35GHz in parts vs 800MHz for the whole (for example) does to yields? I surely don't. They might have some advantage from the asymmetry of the framebuffer size in the "one down" sku where NV cleaned up with 6800 GT (768MB vs 512MB).

Question is now if the above makes enough difference to compensate for the sales/deals loses since Q4 2006 and the high costs of several respins to get whatever bugs out since last year.

I don't think anyone can answer your yield question that easily, but if I play around with some idiotic layman's math:

48 * 10 FLOPs * 1.1 GHz (clock domain) = 528 GFLOPs/s

Of course is such a scenario completely senseless since it would suggest only 3 clusters and probably a 384bit bus might have made more sense here too and a fundamentally different chip after all from ground up.

These things are extremely hard to compare, even more so for a layman like me.

Russell

21-Mar-2007, 06:40

Question is now if the above makes enough difference to compensate for the sales/deals loses since Q4 2006 and the high costs of several respins to get whatever bugs out since last year.

The several respins isn't a big deal. All chips have them, especially big ones. The A13 rev is only the third revision, not the thirteenth (from what I've read in this very thread). There's no need to recover from that; nothing is 100% the first time. The only reason it seems like a big deal is because we've been hearing so much about it.

Arty

21-Mar-2007, 08:24

Not sure if this (http://www.behardware.com/news/8691/cebit-release-date-of-the-r600.html) was posted yet:

CeBIT: release date of the R600?
One of the brand selling ATI's products gave us more information regarding the release of R600, or X2900, if we use the commercial name.

According to the documents that we saw, the evaluation samples of the Catseye 102-B600 delivered to ATI's clients and corresponding to the X2900 XT with 512 MB of GDDR3 memory, are scheduled to be released in March 22 (even if ATI has already sent demonstration computers to each manufacturer to show them to their own clients).

As usual now for high end models, ATI will deliver cards produced by subcontractors and the first deliveries of the final product are scheduled for the 21rst of April. The availability for end users will be at the same time as the product release, early May, the time for manufacturers to stick their logos, put the cards in boxes, and ship them.

The same document speaks of the Dragonshead 2 102-B007, or X2900 XTX, featuring 1 GB of DDR4 memory. Release dates are similar to the samples, 22nd of March, but the final products are expected April 1rst and mass availability mid-May. However, as another manufacturer told us that this card might not even be released, we don’t know really who is right and who isn't!

Silent_Buddha

21-Mar-2007, 08:50

Here's an odd thought considering how apparently paranoid AMD/ATI is about possible leaks on R600...

Considering that AMD are given each AIB a different board with what looks to be slightly different colored coolers with identifiers plastered over iin an attempt to dissuade leaks by making cards easily traceable back to the AIB...

Would it be the next logical step to release different and often contradictory information to different AIBs while sitting around and waiting to see what information gets leaked onto the net?

It's a bit too far out in conspiracy land for me to think it's happening. However, my imagination is having a field day thinking about it. :)

I mean if AMD told half the AIBs that it'll be on 80 nm and half of them that it would 65nm. And then told some of them that there were delays due to bugs, some that there were delays due to power leakage, some that there were delays due to waiting for a game bundle, and told some that everything was right on schedule.

All with the sole purpose of seeing which of their customers is leaking information to the press. As well as finding out which websites are getting inside information from which customers.

Yeah, it's totally Machiavellian, but that's what make my imagination go, hmmmm... :razz:

Regards,
SB

PS - yes, my mind wanders in odd directions with a lack of any concrete information.

xatnys

21-Mar-2007, 08:51

The same document speaks of the Dragonshead 2 102-B007, or X2900 XTX, featuring 1 GB of DDR4 memory.

:shock: At only the cost of an arm and leg! :lol:

Sobek

21-Mar-2007, 08:56

GDDR4? :razz:

neliz

21-Mar-2007, 10:16

PS - yes, my mind wanders in odd directions with a lack of any concrete information.

Sounds exactly like R420 launch eh?

It's easy to trace back the partner which leaked the info if you ship boards at 499.75, 500,75, 520.25, 523.25 etc.

trinibwoy

21-Mar-2007, 11:33

well im not expecting nvidia to do much with the ultra if its sitll 90nm, cuz it would cut into profit margins. id guess that a likely scenario is that the 8800 ultra is just the 8800 gtx at 635/2000.

Doubtful, the OC cards already hit those clocks. If the Ultra is going to be anything like the 7800GTX-512 (hopefully with better availability), core clocks would have to be a lot closer to 700. Or else they could just let XFX and BFG take on R600 :)

In the case of G70 it was a big step up to 90nm. The transition is a little easier now so I wouldn't be surprised in the least if Nvidia was already working on an 80nm shrink of G80 before the 90nm version even launched. They need to shrink that chip if they have any intentions of doing a GX2 anyway.

_xxx_

21-Mar-2007, 11:50

I wouldn't be surprised in the least if Nvidia was already working on an 80nm shrink of G80 before the 90nm version even launched. They need to shrink that chip if they have any intentions of doing a GX2 anyway.

I would be VERY surprised if G90 and G100 (or whatever the naming should be) aren't all but finished and ready for the first spin by now.

vertex_shader

21-Mar-2007, 12:49