The NEXT LAST R600 Rumours & Speculation Thread [Archive] - Page 3

icecold1983

14-Mar-2007, 01:17

The bandwidth/transistor/speed costs are too high compared to how it's a little better than smarter methods like MSAA.

imo theres a very noticeable iq difference between the 2, at the speed hardware is out pacing s/w its only a matter of time till we have so much extra speed that enabling 4xssaa across the board will be the norm, also unless im mistaken, ssaa works with everything, so every game would support aa.

trinibwoy

14-Mar-2007, 01:31

You're not the first person to make that argument ice but no matter how fast hardware gets MSAA+Transparency AA+higher resolution will always be faster than SSAA. I don't think we'll ever be at a point where there is enough spare horsepower to render at 4x the resolution and get playable performance in modern titles. MSAA is already at 8x and with hacks like CSAA it will only keep closing the IQ gap.

icecold1983

14-Mar-2007, 01:52

Bouncing Zabaglione Bros.

14-Mar-2007, 02:10

i know msaa will always be faster, but when we are at the point of getting several hundred fps in games at high res, it wont matter that ssaa is slower

heres how i see it, crysis is the most graphically advanced game we are likely to see for the next 2 to 3 years, hardware will likely be 3 times more powerful than an 8800 gtx at that time. id imagine a card 3x as powerful as an 8800 would be able to run crysis at 1920 x 1200 at well over 100 fps. im pretty sure at that point ssaa would be almost universally useable at a lot of resolutions.

The memory requirements will be a major issue. When your eating up 4x, 8x or even 16x what you otherwise would just for SS, that's a lot of space gobbled up that could be used for other things.

As a general rule, game developers can always gobble up more speed with more features, so I don't think you're going to see Crysis at 100 fps with all they eye candy turned on. Even if you did, you'd lose so much of it for SS, that MSAA will always be more attractive for the vast majority of people. The slight difference in quality is just not worth the massive drop in speed, which is why non-SSAA methods were developed in the first place. It's all about working smarter, not harder, even with ubercards like G80 and R600.

trinibwoy

14-Mar-2007, 02:30

Doesn't MSAA require the same framebuffer usage as SSAA?

Razor1

14-Mar-2007, 02:39

MSAA savings increase as you increase the number of textures in use its a 1 to 3 ratio of sample savings for the same level of AA for MSAA to SSAA.

trinibwoy

14-Mar-2007, 02:41

Huh?

I'm not talking about performance or textures. I believe the MSAA framebuffer is the same size as an SSAA framebuffer.

Razor1

14-Mar-2007, 02:44

SSAA doesn't sample just one color and sample per subpixel, it might do more depending on the multitexturing. So SSAA might need more framebuffer

trinibwoy

14-Mar-2007, 03:32

Heh, I'm still lost here.

I think you're confusing storage with the fact that the color and Z info for all samples in MSAA is the same. I'm not talking about how the samples are generated - just how they are stored. Whether it's MSAA (1 color/Z * 4) or SSAA (4 different color/Z) they both take up the same space. I'm not counting any framebuffer compression schemes though.

And multi-texturing has no relevance to AA AFAIK - if a fragment requires multiple textures all texture information would have already been blended into the sample before it's issued from the shader pipeline.

psurge

14-Mar-2007, 03:52

From what I understand, the worst case frame-buffer size for a given sample count is the same when comparing MSAA to SSAA, but the information content in an MSAA buffer is usually lower (read: in typical game scenes). This means that on average an MSAA buffer will compress better than an SSAA buffer, leading to a larger memory footprint (and bandwidth requirements) for SSAA.

That said, I remember several knowledgeable people saying that drivers allocate for the worst case (imagine the performance cliff you would fall off of if you happened to underallocate your framebuffer) in which case MSAA usually has a bandwidth advantage over SSAA, but never a space advantage. I'm pretty sure trinibwoys point is valid.

icecold1983

14-Mar-2007, 03:56

quick question, does the number of edges in a given scene affect the performance toll msaa takes?

3dcgi

14-Mar-2007, 05:00

Doesn't MSAA require the same framebuffer usage as SSAA?
Typical Full scene MSAA as implemented by Ati and Nvidia require the same amount of memory. Edge AA versions of MSAA like Matrox's FAA do not. As psurge said MSAA will require less bandwidth because the color buffer generally can be compressed better than SSAA.

quick question, does the number of edges in a given scene affect the performance toll msaa takes?
Yes.

Silent_Buddha

14-Mar-2007, 09:00

Wishful thinking, but I would love to see a return of some for of Rotated Grid SSAA as used on the 3dfx V5 series.

I still use the 4x version of that as the gold standard to which all other methods of AA have been lacking.

Considering the top R600 will have 1 gig of memory and gobs and gobs of bandwidth to spare. It would be a shame if some form of SSAA wasn't at least offered as an option.

Granted future games might well use most of that memory for something, however the vast majority of current games don't use 256 megs much less 512 megs.

So something like a maxed eye candy next gen Unreal Engine game may use a gig with highest options, but for myself, I'd bump down texture quality a bit if it allowed me to use 4x RGSSAA

The memory configuration of the R600 just screams SSAA to me.

I mean honestly, I've read a lot of good theories in the various R600 threads here, but maybe the simple explanation is the best. And what better way is there to enhance current Dx9 games than to offer a low perfomance hit for some version of SSAA?

Regards,
SB

AlexV

14-Mar-2007, 13:41

bigtabs

14-Mar-2007, 14:33

For some newer titles, sure.. but people don't stop playing games they've had for a year or two (or more).

I want to be using decent levels of SS on those games. IQ is increased and graphical horsepower isn't going to waste.

SirPauly

14-Mar-2007, 15:02

All, imho.

I've come to appreciate the mixed modes of CrossFire due to how well the super-sampled aspect offered quality. It reminds me very much of the quality of the V5's quality but lacking the hit, hehe!:)

All I am saying as edge quality seems to be there from quality aspects, well, there is other areas that need help as well.

Specular offers a lot of aliasing at times; as does Shaders at times. It's nice to have something with a virtual free hit at times to help curb this type aliasing.

It's nice to see the companies try to help curb and improve Alphas; to improve immersion.

Super-Sampled isn't forsakened totally it's more of a luxury feature right now.

AlexV

14-Mar-2007, 16:32

icecold1983

14-Mar-2007, 17:40

I think everybody realizes that as the ammount of per-pixel work grows(and it`ll certainly keep growing, with increased shader complexity, multiple fragments per surface etc.), SSAA becomes less and less of an option. No?

but assuming hardware continues to advance faster than s/w its only a matter of time till we are so far ahead of even the most demanding title. also users dual video cards could definitely enable 4x ssaa or higher in many many games even today

Geo

14-Mar-2007, 17:44

Again, as an only solution SS is gone for good. As an extra, either through a mixed mode, particular case(alphas) or a multi-GPU mode it may continue to exist. Shader aliasing should be dealt with by developers.IMHO.

Well, David Kirk said as much around the time of R520's release. That felt then, and still feels today, like a cop-out to a degree. Developers do not use sticks and mud to make games. They use tools. Tools that target a specific API and associated hardware and are provided by Microsoft, Nvidia, ATI. . . When you have a major IQ issue, like shader aliasing, it is those tools that need to provide developers with a friendly and effective way to deal with it. Do they today? Because if they don't, then I think we're justified in saying that the API and its support system is broken rather than dumping a bucket of s**t on the head of developers.

SirPauly

14-Mar-2007, 18:08

Shader aliasing should be dealt with by developers.IMHO.

Did hear this same mind-set when the subject was alphas years ago.

AlexV

14-Mar-2007, 20:21

And do you feel putzing around with alpha-tested textures and putzing around with per-material fragments is the same?Would you find it sexy that your overly long surface material that is a heavy hitter in terms of performance in its own right, would have to be rendered arbitrarly 4x times because there`s some shader aliasing somewhere and it`s the sexay to have a CP switch to deal with everything?I do not agree on this, sorry.

Geo, I agree. But I think devs are better equiped to handle the situation. They`re likely to employ shader-AA where it`s needed, and only there, instead of using a full-screen "sledgehammer".

What`s with the hardware going a helluva lot faster than SW and surpassing it to such a huge degree that every crude technique becomes feasible?Is that so?Why?Because the G80 rocks Oblivion?Shouldn`t we move past this mindset:every time a new gen is released, it rocks games available then and targeted at stuff that was available before. Will the G80 or R600 rock Crysis, for example, in the same manner?Doubtful. So we`re stuck with the old conundrum:use SSAA for older titles(which is cool), use something smarter for new ones where you don`t have performance to burn. Would you buy a card that has SSAA as the single method of approaching aliasing?

SirPauly

14-Mar-2007, 20:56

No, just limited past views about Alphas being tak'n care of by a sledgehammer approach instead of innovative adaptive views -- Kinda reminds me of what some are offering about Shaders now. Time-will-tell.

icecold1983

14-Mar-2007, 20:56

in 5 years i would be surprised if hardware wasnt leaps and bounds ahead of ANY videogame s/w.

Silent_Buddha

14-Mar-2007, 21:07

IMO, I'd love to go into the control application of whatever graphics card and when I go to change driver controlled AA see the option for Performance (MSAA, TSAA, etc) or Quality (some for of SSAA, hopefully RGSSAA).

Assuming they also didn't skimp on AF, I would buy that card in a heartbeat even if it was slightly slower than the competition. Always assuming of course that while Quality might not be usable in games with bleeding edge graphics that it was useable in past or future titles without bleeding edge graphics.

So, while one part of me is hoping that R600 brings something new and unthought of to the table with the massive bandwith and memory sizes. Another part of me is hoping it's just something as simple as useable SSAA.

And I only have Nvidia to thank for making me wait for R600. With all the horror stories I hear from people I play with online about the 8800 GTX + Vista 64 drivers. I'd own an 8800 right now if not that for. Such a shame as it's the first video card Nvidia has gotten completely right since the GF4.

Granted, I don't play much OpenGL games right now (due to no new ones in quite a while), but if ATI can make their driver work in Vista 64 more stably than it runs in Windows XP. Why can't Nvidia do the same? Grrrr...

Regards,
SB

trinibwoy

14-Mar-2007, 22:40

in 5 years i would be surprised if hardware wasnt leaps and bounds ahead of ANY videogame s/w.

You do realize that there is a vast wealth of offline software renderer algorithms out there that hardware just can't handle in real-time? Don't confuse what's possible on GPU hardware with what's possible on general purpose CPU's. DX10 is just now opening up the possibility for some of these offline algorithms to be hardware accelerated.

Hardware advances will be playing catchup with offline methods for a very long time to come so you won't be seeing hardware overtaking software anytime soon.

icecold1983

14-Mar-2007, 23:08

videogame s/w dude!! not offline renderers.

trinibwoy

14-Mar-2007, 23:17

You do know that videogame software and offline renderers are trying to accomplish the same task right? So when hardware gets fast enough to run algorithms previously unique to the offline space what do you think is going to happen?

AlexV

14-Mar-2007, 23:32

NO NO NO, YOU NOT UNDERSTAND!Well,umm, neither do I:|

icecold1983

14-Mar-2007, 23:42

ummm no one is going to code a video game that advanced for the pc because most people have sucky pcs so its rather a moot point.

when crysis gets released, im willing to bet performance on g80 or r600 is much higher than any other past quantum leap video game was on hardware of its time. for example, if u look at ut2k3, when it was released, the fastest card was the 9800 pro. even at 1024 x 768 with 4/8 it wasnt an entirely smooth experience on many maps. its very likely that g80 will be able to play crysis at a higher res with aa/af.

Shtal

15-Mar-2007, 03:31

The XNA team shared some of their goals for the next iteration of the API, DirectX 10.1, as well as a few of the longer-term aims for the progression of gaming on the PC..

Maybe that is why R600 is delay, so it may be well included on catalyst new driver_cd, the DX10.1 - for R600 release.

Shtal

15-Mar-2007, 06:14

Little bet ago I made a poll vote about what is your expectations about R600.
http://snappoll.com/poll/177165.php

So far the result kinda balance in different approaches.
Total votes I see of as this minute 85 + 1.

And here what people think :)

What do you expect out of R600.
"overall" - *said* about as good as G80 38% 33

*other said* another R300 36% 31

*other* not much to expect 16% 14

*other* above my expectations 9% 8
-------------------------------------------------------------------------------------
Some of you think that not much to expect out of R600, could be because of the delay issues?
hmm, very interesting....

Ailuros

15-Mar-2007, 06:24

when crysis gets released, im willing to bet performance on g80 or r600 is much higher than any other past quantum leap video game was on hardware of its time. for example, if u look at ut2k3, when it was released, the fastest card was the 9800 pro. even at 1024 x 768 with 4/8 it wasnt an entirely smooth experience on many maps. its very likely that g80 will be able to play crysis at a higher res with aa/af.

The fastest GPU when UT2k3 was released was the NV25, on which with a strong CPU the best you could get was 1024*768 with 2xAA/2AF (and that was right on the edge of playability too). Then the various filtering optimisations started to kick in and you could use a tad more AF even on NV25 (uhmm no wonder if you just filter on the first texturing stage LOL).

Then the 9700PRO made its appearance and not only was 4xAA possible even in higher resolutions than just 1024, but considering that similar to the above filtering optimisations came along by default, AF wasn't an issue either.

http://www.anandtech.com/video/showdoc.aspx?i=1647&p=8

http://www.anandtech.com/video/showdoc.aspx?i=1647&p=13

(Those were pre-release performance comparisons but I'm fairly certain that the game shipped quite a bit before the 9700PRO).

http://www.anandtech.com/video/showdoc.aspx?i=1683&p=18

Sobek

15-Mar-2007, 06:30

Actually UT2k3 shipped 2 months after the 9700 Pro launch.

icecold1983

15-Mar-2007, 07:02

yes im aware the benches make it look smooth, but i had the game from day 1 and when i upgraded to my 9800 pro i could not play smoothly on all maps at 1024 x 768 with 4x/8x. many maps had fps drops and mouse lag unless i turned off aa or lowered the res/texture detail.

AlexV

15-Mar-2007, 09:17

Why would things be different now?You do realise that there`s a helluva lot of things that weren`t feasible before due to API and hardware limitations, that now become useable, yes?And that that impacts performance?So, why would Crysis, for example, a game thought from the ground up with DX9 hardware in mind, that is doing a lot of per-pixel work, moreso than any game before it I dare say, be uber playable at high-res with high AA high AF?

icecold1983

15-Mar-2007, 09:58

never claimed it would be playable at super high res with aa/af, just that it will likely play faster than previous landmark games have on current hardware due to the rate at which hardware is advancing

AlexV

15-Mar-2007, 11:07

We`ll have to agree to disagree on this then. I simply don`t see HW progressing through an n! like function, whilst SW going through a log(n) one.

nagus

15-Mar-2007, 13:26

http://www.pcgameshardware.de/?article_id=571567

AWESOME!

zgemboandislic

15-Mar-2007, 13:57

http://www.pcgameshardware.de/?article_id=571567

AWESOME!

Finally, R600-related post...

Hmm...the leather jacket doesn't look plastic anymore - at least something positive! :twisted:

IbaneZ

15-Mar-2007, 14:31

http://www.tweaktown.com/news/7193/amd_roadmap_update_on_r600_details_confirmed/index.html

March 30?

R600 IS COMING

Som am I if that date's correct. :smile:

Arnold Beckenbauer

15-Mar-2007, 14:32

The Hunt For The Red Octo.. R600. (http://www.pcgameshardware.de/?article_id=571523) PCGH guys are hunting R600. :lol:
ATI's RV630 and RV610 (15 pics) (http://www.pcgameshardware.de/?article_id=571589)

Anteru

15-Mar-2007, 14:35

http://www.pcgameshardware.de/?article_id=571567

AWESOME!

Doesn't look that fantastic IMAO ... actually, after Toy Shop, I'd have expected ATi to bring a really yaw-dropping demo this time, kinda "toy-shop on steroids".

At least those guys say that the PCB is not abnormally long and looks similar to the 1950 XTX, so it seems they had non-OEM versions running.

trinibwoy

15-Mar-2007, 14:55

Looks like they were asked to pull the pics and complied.

Tim Murray

15-Mar-2007, 15:00

March 30 launch? Uh, I get the feeling that somebody didn't get the memo about the whole canceled Editor's Day thing.

Geo

15-Mar-2007, 15:03

http://www.tweaktown.com/news/7193/amd_roadmap_update_on_r600_details_confirmed/index.html

March 30?

R600 IS COMING

Som am I if that date's correct. :smile:

So far as I know, that's been overtaken by events and is from before the decision to push back by several weeks to make it a broader launch.

nicolasb

15-Mar-2007, 15:08

http://www.tweaktown.com/news/7193/amd_roadmap_update_on_r600_details_confirmed/index.html

March 30?

R600 IS COMING

Som am I if that date's correct. :smile:That's a rehash of a lot of stuff that was posted about a month ago - including what was, a month ago, the expected release date, and the news that a number of boards would be exhibited at CeBit. That date has since been pushed back to "Q2" - most likely mid-May.

There's one glaring inaccuracy in there, which is that the 9.5" XTX card (they say) uses 240W and has two 6-pin PCIe power connectors. That's physically impossible: you can't draw more than 225W peak from the slot and two 6-pin connectors.

The "13 revisions" thing is also rubbish. A13 means the third revision - one of the digits refers to silicon, the other to metal.

So, all in all, I would be pretty sceptical, if I were you.

IbaneZ

15-Mar-2007, 15:21

R600 is a 65nm Product and will be annaunced at a TechDay in less then 16 days.
http://forums.vr-zone.com/showthread.php?t=136197

http://www.k-hardware.de/news.php?s=K318002b636e5a8c7e84869802fb04d41&news_id=6395

A lot of news and rumors from Germany atm. :smile:

Arun

15-Mar-2007, 15:27

Yup, a lot of wrong rumours from Germany :p

trinibwoy

15-Mar-2007, 15:30

65nm would make me happy :D I want to believe !!!

Geo

15-Mar-2007, 15:57

I think I'd nominate 65nm as the "R520 has 32 pipes" for R600. And, whaddyaknow, same source flogging both. :wink:

w0mbat

15-Mar-2007, 16:03

Its longer than RV630 & RV610.
http://img92.imageshack.us/img92/1694/imgp5553ga2.jpg

Kaotik

15-Mar-2007, 16:06

Only one powerplug on that

w0mbat

15-Mar-2007, 16:08

Only one powerplug on that

R600 XL or R600 65nm? It should be a RV630, but the PCB is way too long.

Kaotik

15-Mar-2007, 16:10

Could it be that RV670 is actually closer than expected?

15-Mar-2007, 16:12

Its longer than RV630 & RV610.

I thought this was RV630Pro. Isn't this from the article with only RV610 and RV630 pics?

INKster

15-Mar-2007, 16:13

R600 XL or R600 65nm? It should be a RV630, but the PCB is way too long.

I'm betting on a R600 XL.

Kaotik

15-Mar-2007, 16:14

http://www.pcgameshardware.de/?article_id=571589

More pics of the card

edit: including RV610 shots

Geo

15-Mar-2007, 16:16

Mid-range with double-wide cooling? :shock:

Kaotik

15-Mar-2007, 16:19

Mid-range with double-wide cooling? :shock:

Midrange or not, it's not RV630 nor RV610

Geo

15-Mar-2007, 16:32

Midrange or not, it's not RV630 nor RV610

That is one confusing page, but yes I can see now there are three cards in that array. One single-wide marked with a white label "RV610", another single-slot marked "RV630" with a white label, and then whatever that double-wide is. ATI has not typically liked XL to be a double-wide. . .

jamis

15-Mar-2007, 16:32

Midrange or not, it's not RV630 nor RV610
Yes it is, they are saying it's RV630. The one without the cooler is RV630 too, but different PCB layout.
So the one with the cooler (and powerplug) is propably the highest model (RV630XT?) while the rest of the RV630's don't need extra power.

no-X

15-Mar-2007, 16:34

Mid-range with double-wide cooling? :shock:
Midrange or not, it's not RV630 nor RV610

Cooler is double-slot, but heatsink is quite small (less than 1/2 of the wind-tunnel space). And only one power connector. It could be Orloff or Kohinoor PCB according this (http://img6.picsplace.to/img6/26/4128_large_AMD_RV610_RV630.png). And the small one w/o power connector could be Sefadu(?)

Jawed

15-Mar-2007, 16:36

That big board looks to be about 5mm shorter than X1950XTX (whereas we're expecting R600XTX to be the same length). The area around the power is different from the VR-Zone "corner of R600XTX retail" we saw a while back, too. Obviously the single power connection is a big difference.

Jawed

Kaotik

15-Mar-2007, 16:38

15-Mar-2007, 16:47

It's RV630. The article also specifically states it....

jamis

15-Mar-2007, 16:47

So either it's the "biggest" RV630, mysterious RV670 or R600 XL/GT-type card
Can't you read CB? :lol: They are clearly saying that both are RV630 with different PCB layouts.

Jawed

15-Mar-2007, 16:53

So either it's the "biggest" RV630, mysterious RV670 or R600 XL/GT-type card
RV670 just hasn't been rumoured as due and it seems unlikely to me that R600XL would use a different PCB to R600XTX - that's what happened with X1800XL. GT/Pro would prolly come as board revisions later, if there was something worth trimming.

But it seems to me that R6xx is too hot for any of the higher performance cores to use anything other than a two-slot cooler or something like the specialist Zalman etc. that we saw on X1950Pro.

Another thing to note with what is apparently the "RV630XT" is that there's no heatpipes visible. So it's a relatively low-cost cooler. There's actually relatively little metal inside the shroud.

Jawed

Kaotik

15-Mar-2007, 17:01

Can't you read CB? :lol: They are clearly saying that both are RV630 with different PCB layouts.

No, I can't here read german any better than at FM :lol:

Arty

15-Mar-2007, 17:05

If thats RV630, then a dual-slot X2400 card? Sorry CJ, something doesnt add up there. :smile:

mao5

15-Mar-2007, 17:26

Today, we are meeting the latest rumor from CeBIT in Hanover, Germany. According to AMD and ATI graphics business firms AIB senior officials, R600 will be ATI AMD Techday 16 days after the formal sessions, But it is formally released yet clear. CeBIT07 venue other source, R600 will be formally announced in six weeks after that date in early May. In the production and manufacturing process, the R600 will be used outside the current 80nm production process. But according to the venue rumors, AMD hopes R600 chip production using 65nm technology, The reason is that under R600 chip yields 80nm process is not high, a more serious problem and leakage. R600 will continue to be performed by TSMC TSMC OEM. In addition, R600 Reorganized voltage source confirmed the rumor, Now the fact is that R600 R650, and the coming days may test results. R650 allegedly demanding more efficiency.

mao5

15-Mar-2007, 17:31

http://www.pcgameshardware.de/?article_id=571567

AWESOME!

hey, those pics were taken down now. Where are they now? in some other websites?

Jawed

15-Mar-2007, 17:33

If thats RV630, then a dual-slot X2400 card? Sorry CJ, something doesnt add up there. :smile:
Thinking more about the "lightweight" heatsink within that shroud, it seems like what we saw with X800XT then X850XT - a noisier single slot cooler or a quieter dual-slot cooler.

Is it a reasonable guess that RV630XT is about as hot as X850XT?

Jawed

Headstone

15-Mar-2007, 17:46

no-X

15-Mar-2007, 17:46

hey, those pics were taken down now. Where are they now? in some other websites?
...

http://4um.ocguru.cz/showpost.php?p=698142&postcount=197

Headstone

15-Mar-2007, 17:51

Those look like photos as opposed to screen grabs so it is going to be really hard to judge how good they are until we see the final product.

Geeforcer

15-Mar-2007, 18:00

http://img6.picsplace.to/img.php?file=img6/27/Imgp5304.jpg

Woa, is that a CGI or just a real photo? Either way, can you say Adrianne = pwnd?

INKster

15-Mar-2007, 18:05

http://img6.picsplace.to/img.php?file=img6/27/Imgp5304.jpg

Woa, is that a CGI or just a real photo? Either way, can you say Adrianne = pwnd?

Frankly, i think Adrianne looks more realistic.
This has that "cartoonish" look, with lifeless eyes...

MistaPi

15-Mar-2007, 18:14

INKster

15-Mar-2007, 18:18

Could the bigger PCB mean a RV630 with 256bit memory bus?
According to DailyTech no RV630 card will have 256bit mb, but also according to DailyTech there wont be a dual-slot RV630.

http://www.dailytech.com/article.aspx?newsid=6451

Maybe they were talking about reference cooling from ATI.
This double-slot cooler may be a specific vendor solution.

ChrisK

15-Mar-2007, 18:49

Sorry, if I missed it, but it doesn't seem to have been mentioned before:

This pcgameshardware.de page (http://www.pcgameshardware.de/?article_id=571589) says they're targetting a mid-April launch date for RV630. So did they get something wrong, or what's with the "family launch" AMD reps announced?

Kaotik

15-Mar-2007, 19:17

Maybe they were talking about reference cooling from ATI.
This double-slot cooler may be a specific vendor solution.

Vendor specific coolers usually have the vendor name big on 'em, not just ATI

Arty

15-Mar-2007, 19:28

Frankly, i think Adrianne looks more realistic.
This has that "cartoonish" look, with lifeless eyes...
The Adiranne pic looks better, the demo not so.

Shtal

15-Mar-2007, 19:34

Yup, a lot of wrong rumours from Germany :p

If so; then, 15 more days left for the truth! to be reveal :)

Also their might be something unexpected too - about R600, just like nobody suspected unify-shaders for G80....
[Edit: Nvidia probably waiting for R600 too, to see how it will turn out)

Arnold Beckenbauer

15-Mar-2007, 21:47

http://www.pcgameshardware.de/?article_id=571567

AWESOME!

Why didn't you translate? :sad:

The demo takes 680 MB of VRAM, alone 500 MB for textures, it uses soft particles and self-shadowing, 375 shaders are used, the standard PS is 430 instructions big.

First impressions: the R600 looks almost like X1950XTX and uses almost the same cooler.

IbaneZ

15-Mar-2007, 22:59

CeBIT: AMD speaks about R600

A little Q and A with a dude called Vijay Sharma.
http://www.xtremesystems.org/forums/showpost.php?p=2071080&postcount=134

Kaotik

15-Mar-2007, 23:04

Just curious, does the "RhOs" stand for something, don't remember seeing it before, yet it's present on both RV610 and RV630 boards

Jawed

15-Mar-2007, 23:07

Grab these presentation slides while you can:

http://www.rage3d.com/board/showpost.php?p=1334851342&postcount=3637

Jawed

trinibwoy

15-Mar-2007, 23:09

Lol, everything's in there except what we want to know.

nutball

15-Mar-2007, 23:14

Just curious, does the "RhOs" stand for something, don't remember seeing it before, yet it's present on both RV610 and RV630 boards

RoHS is Restriction of Hazardous Substances (http://en.wikipedia.org/wiki/RoHS). EU thing. Legislation. Environment. Law stuff. It means don't let your graphics card kill babies, etc. I guess it means that those graphics cards are vegetarian, or something.

Bouncing Zabaglione Bros.

15-Mar-2007, 23:20

RoHS is Restriction of Hazardous Substances (http://en.wikipedia.org/wiki/RoHS). EU thing. Legislation. Environment. Law stuff. It means don't let your graphics card kill babies, etc. I guess it means that those graphics cards are vegetarian, or something.

It's to do with recycling and disposal. EU won't accept any electronics that can't be dealt with safely without poisoning the environment after it's lifespan is over.

Russell

15-Mar-2007, 23:25

From that interview, regarding R600's performance vs the G80:
"You will be surprised…"

Here's hoping he means positively surprised, lol.

Arnold Beckenbauer

15-Mar-2007, 23:27

CeBIT: AMD speaks about R600

A little Q and A with a dude called Vijay Sharma.
http://www.xtremesystems.org/forums/showpost.php?p=2071080&postcount=134
And for the public?
The public will see with blow on of R600 everywhere in Computex which will take place in June since it will be launched during May
I can wait. :smile:
http://www.k-hardware.de/news.php?s=c&news_id=6399
K-Hardware guys say, that AMD confirmed R600 will come in 65 nm.

Bouncing Zabaglione Bros.

15-Mar-2007, 23:34

From previous threads, we've wondered what AMD can do with R600 that will top the surprise we felt when G80 arrived with unified shaders and scaler units. If R600 is 65nm, I will be very, very surprised. I don't think it's very likely, but you never know until it actually arrives...

Tim Murray

15-Mar-2007, 23:37

Notebooks, aka RV610/630, are 65nm. See last slide.

w0mbat

15-Mar-2007, 23:42

But they also said that R600 65nm is confirmed.

Jawed

15-Mar-2007, 23:43

Notebooks, aka RV610/630, are 65nm. See last slide.
Are ATI's older laptop GPU dies identical to the desktop ones?

Jawed

Jawed

15-Mar-2007, 23:45

From previous threads, we've wondered what AMD can do with R600 that will top the surprise we felt when G80 arrived with unified shaders and scaler units. If R600 is 65nm, I will be very, very surprised.
This whole thing continues to get more and more bizarre. It's like Sunset Beach for graphics geeks.

Jawed

jimmyjames123

15-Mar-2007, 23:51

The "surprise" is this: ATI is coming out with R600 on 80nm process, while also coming out with higher clocked R650 on 65nm process at the same time in the middle of the year, in addition to coming out with their midrange and low-end R6xx cards. The problem for them is that there is no doubt nV will have a significantly faster card than 8800GTX out by that time, and also will have much better and faster drivers than they currently have. It is certainly going to be interesting.

Kaotik

16-Mar-2007, 00:03

Are ATI's older laptop GPU dies identical to the desktop ones?

Jawed

Well if the powersaving features are disabled/enabled via drivers, they might be actually 100% identical, if it needs extra circuitry, they're not.

Kaotik

16-Mar-2007, 00:03

But they also said that R600 65nm is confirmed.

Only one sait claims they said, the slides ain't saying anything of such

jimmyjames123

16-Mar-2007, 00:04

Jawed

16-Mar-2007, 01:07

That is not the type of information that ATI would put in slides like these :)

What has people like Dave Baumann and others saying in the past about ATI delayed launches: R520 is late, but R580 is on time. Same goes with R600.
I don't think anyone was expecting "R650" (65nm) before the autumn though. So the idea of it being brought forward by such a significant period is a huge deal.

Jawed

Bouncing Zabaglione Bros.

16-Mar-2007, 01:29

I don't think anyone was expecting "R650" (65nm) before the autumn though. So the idea of it being brought forward by such a significant period is a huge deal.

Might be a way for them to redeem themselves with regards to the late execution of R600, especially if R680 was ahead of schedule and originally due for a quarter later than R600 is going to turn up anyway. R680 as a refresh against the G80 refresh makes sense.

I'm still sceptical that 65 nm is ready for something as complex as R600, and it would be such a bold move, it seems somewhat unbelievable. Would be great though, even if just for the effect on the price of R600.

wishiknew

16-Mar-2007, 01:29

From those slides, is the HDMI/HD audio feature the R600 equivalent of fetch4/df24 for r520?

Arty

16-Mar-2007, 01:42

From those slides, is the HDMI/HD audio feature the R600 equivalent of fetch4/df24 for r520?
Only if the competition has HDMI/HD audio. ;)

16-Mar-2007, 01:47

I'm still sceptical that 65 nm is ready for something as complex as R600, and it would be such a bold move, it seems somewhat unbelievable. Would be great though, even if just for the effect on the price of R600.

Someone on a Dutch forum pointed me to this link: http://www.k-hardware.de/news.php?s=Kcced1a97b825370a52a842fe607b2f0a&news_id=6399

In short it says that AMD confirmed to them that R600 is a 65nm product. And it also has all the slides at the end that were already leaked onto the web.

Silent_Buddha

16-Mar-2007, 02:00

One interesting thing from those slides.

Video decode is completely decoupled from the 3D rendering engine, as opposed to R5xx series where Avivo video decode was tied to and reliant upon the 3D rendering engine. Thus a R580 has faster HW decode of video than say a R530.

This poses a couple of very nice benefits.

1. Universally lower CPU utilization when decoding HD Video across all R6xx lines.

2. With the graphics layer rewrite for Windows Vista. It should be entire possible to run a 3D application (game or otherwise) at full speed while also watching HD Video with full hardware acceleration.

Granted number 2 may not be all that important to all that many people. But considering Vista's desktop is rendered using the 3D acceleration it should prove for an even more responsive desktop environment when playing HD Video.

And of particular interest to me. Since I happen to raid a lot. It'll be a bit of a bonus to be able to watch full screen HD Video on my second monitor while I'm raiding in whichever MMORPG I'm currently playing.

Not to mention I'd much prefer to watch TV or video streams on my second monitor while gaming rather than having my TV + DVR + Surround Sound Receiver + 5.1 Channel Amp, etc all running just so I can watch TV while gaming.

While not entirely game related, as a graphics and video junkie this has me extremely excited if well executed.

Regards,
SB

Kaotik

16-Mar-2007, 03:02

Something caught my eye in the slides - of course it might be just case of using random core in the pics, but in the slide where they compare Xenos and R600, the chip isn't turned 45 degrees on the package, like it's been on the earlier pics.

Shtal

16-Mar-2007, 03:05

The "surprise" is this: ATI is coming out with R600 on 80nm process, while also coming out with higher clocked R650 on 65nm process at the same time in the middle of the year,

R600 will be called X2900XTX

So that mean that R650 will be X2950XTX. Something is not right with calculations.... So small number differentness between R600 and R650.
[Edit: example - between R300 and R350 was Radeon 9700Pro and Radeon 9800Pro] "Unless if it is going to be a miracle that R600 be 65nm tech"

16-Mar-2007, 03:39

And of particular interest to me. Since I happen to raid a lot. It'll be a bit of a bonus to be able to watch full screen HD Video on my second monitor while I'm raiding in whichever MMORPG I'm currently playing.

Last time I tried watching something on my second monitor during a raid, I let the maintank die and we wiped. Try to explain to 39 other people that you were distracted by an episode of Charmed. :oops:

Kaotik

16-Mar-2007, 04:01

Last time I tried watching something on my second monitor during a raid, I let the maintank die and we wiped. Try to explain to 39 other people that you were distracted by an episode of Charmed. :oops:

Don't worry, now you will only have to explain that to 24 people since the new raids are max of 25ppl :lol:

But ye, I know what you mean, luckily I'm specced all shadow and no-one notices if one DPS is slacking a bit

Anon Lamer

16-Mar-2007, 06:07

Mid-range with double-wide cooling? :shock:

I've noticed that too... but at 71 w TDP (8600 GTS 256(?)) or 121/128 (256/512 Rv530) theres no alternative. 8600GTS is about equal to a x1950 pro and Rv530 to a X1900XTX. The practical limit for single slot cooling seems to be around 50W TDP. Although there are x1950pros with single slot cooling, they are not known as being noiseless or cold running. Dont forget that these "mid range chips" have as many transistors and as high a clock as last generations high performance range.

INKster

16-Mar-2007, 06:25

I've noticed that too... but at 71 w TDP (8600 GTS 256(?)) or 121/128 (256/512 Rv530) theres no alternative. 8600GTS is about equal to a x1950 pro and Rv530 to a X1900XTX. The practical limit for single slot cooling seems to be around 50W TDP. Although there are x1950pros with single slot cooling, they are not known as being noiseless or cold running. Dont forget that these "mid range chips" have as many transistors and as high a clock as last generations high performance range.

But they are also made on a smaller process.
Therefore, they run on lower voltage ranges than the old top of the line models.

Dave Baumann

16-Mar-2007, 07:13

Try to explain to 39 other people that you were distracted by an episode of Charmed. :oops:

Pheobe, Paige, Piper... What explaination is actually needed?

(Strange world we live in! :p )

Anteru

16-Mar-2007, 09:46

Grab these presentation slides while you can:
http://www.rage3d.com/board/showpost.php?p=1334851342&postcount=3637
Jawed

Nice! What makes me wonder, why no "breathtaking DX10" performance PR in that? They talk a lot about the other features, but I remember the last leaked presentation (mid-2006?) said something about incredible performance increases why this presentation does not say anything about "faster than the competition" ... also nothing about increased IQ compared to the competition by new AA modi. Strange.

Give me some leaked benchmark, please :)

chavvdarrr

16-Mar-2007, 11:56

R600 will be called X2900XTX

So that mean that R650 will be X2950XTX. Something is not right with calculations.... So small number differentness between R600 and R650.
[Edit: example - between R300 and R350 was Radeon 9700Pro and Radeon 9800Pro] "Unless if it is going to be a miracle that R600 be 65nm tech"What if they use optical shrink 80->65, so they only got more and cheaper chips, but not better speed ?
With 512-b bus, it really seems that price/performance may be a problem, not the raw performance.

remember what AMD did in the past with 0.18 - > 0.13 -> 0.09 - first release chips with same speed, but binned for quantity/power , and only later increase speed when whole process tech matures (ie thoroughbred A vs B)

Arty

16-Mar-2007, 13:23

VR-Zone (http://forums.vr-zone.com/showthread.php?t=136434)

http://img92.imageshack.us/img92/3614/r600vrzoneam4.jpg

AMD is keeping a close watch on their partners over at CeBIT making sure nothing is leaked to the media. We saw some R600 cards but they are all marked with the respective AIB company name all over the card. As such, no one is allowed to take any photos. However, nothing escape from us as usual so here's a side view for your viewing pleasure. More later if there is enough interests around here.
That smells bull as this photo looks like a stock photo rather than snapped up a from a partner at CeBIT.

NocturnDragon

16-Mar-2007, 13:25

What if they use optical shrink 80->65, so they only got more and cheaper chips, but not better speed ?

A optical shrink was possible from 90 to 80nm but I 65nm it's a totally different node. You cannot optically shrink it.

Kaotik

16-Mar-2007, 13:35

VR-Zone (http://forums.vr-zone.com/showthread.php?t=136434)

That smells bull as this photo looks like a stock photo rather than snapped up a from a partner at CeBIT.

I don't see what makes that look more like a "stock photo" than from some partners card @ CeBIT, most cards use, after all, the stock cooling with their own logos printed on, and the white BG is clearly just photochopped in that

Arty

16-Mar-2007, 13:39

I don't see what makes that look more like a "stock photo" than from some partners card @ CeBIT, most cards use, after all, the stock cooling with their own logos printed on, and the white BG is clearly just photochopped in that
Its looks like the same pic but different angle that VR teased a couple of weeks back.

Kaotik

16-Mar-2007, 13:45

Its looks like the same pic but different angle that VR teased a couple of weeks back.

And if the teaser pic was from a retail board, why should another retail board look any different?

Jawed

16-Mar-2007, 13:45

That new VRZone pic shows a cooler without a fan.

Jawed

chavvdarrr

16-Mar-2007, 13:48

A optical shrink was possible from 90 to 80nm but I 65nm it's a totally different node. You cannot optically shrink it.maybe not optical shrink, but without spending time ( and they don't have time!) and resources for speed tuning atm, aiming just for lower price-per-die

Arty

16-Mar-2007, 14:11

That new VRZone pic shows a cooler without a fan.

Jawed
The previous pic

http://resources.vr-zone.com/newspics/Feb07/12/R600XTX-Retail.jpg

jimmyjames123

16-Mar-2007, 15:39

Another thing that would "surprise" some people is that the R600 is a scalar architecture. IIRC, this is the way that Rys and some others are leaning towards.

neliz

16-Mar-2007, 16:49

Another thing that would "surprise" some people is that the R600 is a scalar architecture. IIRC, this is the way that Rys and some others are leaning towards.
I don't think a scalar architecture or marginally increased video playback falls in the category "surprise" .. it needs to be something that makes everybody WANT one..

no-X

16-Mar-2007, 17:03

I don't think a scalar architecture or marginally increased video playback falls in the category "surprise" .. it needs to be something that makes everybody WANT one..
e.g.? Something like "buy R600 and get a free cookie" or something more like full MSAA 8x with performance drop of todays 4x? :wink:

Kaotik

16-Mar-2007, 17:06

I don't think a scalar architecture or marginally increased video playback falls in the category "surprise" .. it needs to be something that makes everybody WANT one..

I wouldn't really call the UVD "marginally increased video playback" tbh, since it's not affected by any 3D stuff you're doing, unlike current solutions.

What I see, is that AMD/ATI has a clear winner in their hands without any doubts - but it's only for one sector - HTPC's - the RV6xx's integrated audio solution is plain brilliant, no extra cabling, one card handles it all from the PC to your TV/Amplifier/whatever via one HDMI cable.

So there's something that makes everybody interested in HTPCs to want one - however I think at this point, it's even smaller market than the highend is.

{Sniping}Waste

16-Mar-2007, 17:14

Found some info at OCW with some bench numbers ranges for the X2900XT that could be real or not on 3Dmark 05 17XXX and 3Dmark 06 11XXX.
http://forums.ocworkbench.com/bbs/showthread.php?t=60799

16-Mar-2007, 17:35

The title of that article says X2900GT instead of XT. And the G and X aren't even close to eachother on my keyboard. Wonder which one it is...

{Sniping}Waste

16-Mar-2007, 17:39

The title of that article says X2900GT instead of XT. And the G and X aren't even close to eachother on my keyboard. Wonder which one it is...
Is there even a GT ver of the R600. All I have heard of is the XTX, XT, XL, PRO. I think the GT is a mis type.

IbaneZ

16-Mar-2007, 17:43

Perhaps it's R600 Garland TX?

I'm bored..

INKster

16-Mar-2007, 17:44

VR-Zone (http://forums.vr-zone.com/showthread.php?t=136434)

That smells bull as this photo looks like a stock photo rather than snapped up a from a partner at CeBIT.

No, that photoshop work on the background is too imperfect to be a stock photo.
I think it is in fact a CeBIT shot.

Anteru

16-Mar-2007, 18:13

Found some info at OCW with some bench numbers ranges for the X2900XT that could be real or not on 3Dmark 05 17XXX and 3Dmark 06 11XXX.
http://forums.ocworkbench.com/bbs/showthread.php?t=60799

That's quite close to what the 8800 GTX does ... may be real, and that would explain why they are quite silent about the performance of the R600. Though I hope for the R600 to be ~20% faster ...

{Sniping}Waste

16-Mar-2007, 18:17

That's quite close to what the 8800 GTX does ... may be real, and that would explain why they are quite silent about the performance of the R600. Though I hope for the R600 to be ~20% faster ...

If its the X2900xt then its about 15% to 25% slower then the X2900XTX so if the numbers are right it will be around the same speed of a 8800GTX. Remember the X2900XT is only 512meg of GDDR3 and has a slower clock speed then the X2900XTX. The X2900XTX is 1 gig of GDDR4.

w0mbat

16-Mar-2007, 18:22

Shtal

16-Mar-2007, 18:29

What if they use optical shrink 80->65, so they only got more and cheaper chips, but not better speed ?
With 512-b bus, it really seems that price/performance may be a problem, not the raw performance.

remember what AMD did in the past with 0.18 - > 0.13 -> 0.09 - first release chips with same speed, but binned for quantity/power , and only later increase speed when whole process tech matures (ie thoroughbred A vs B)

Their might be another theory that by going from 80nm ---->to 65nm. ATI/AMD will have space to add extra feature into the chip/core.

Like example:
A. Add More Shaders
B. Increase texture units
C. Increase ROP's
D. Faster Core frequency

INKster

16-Mar-2007, 18:30

If the X2900 GT is as fast as 8800GTX ATi has nothing to worry about
:grin:

But will it be as fast as 8800 Ultra ? ;)

Nakai

16-Mar-2007, 18:33

Like example:
A. Add More Shaders
B. Increase texture units
C. Increase ROP's
D. Faster Core frequency

A. No
B. No
C. No
D. Yes, orlower power consumption at the aimed clocks.

If the X2900GT has enough power to beat the G80 how strong should the X2900XTX be?
Maybe the R600 can score around 24k at 3DMark05?^^:wink:

mfg Nakai

Shtal

16-Mar-2007, 19:28

A. No
B. No
C. No
D. Yes, orlower power consumption at the aimed clocks.

If you take an example like intel CPU.

90nm Pentium 4 dual core had 1MBx2 cache memory
65nm Pentium 4 dual core had 2MBx2 cache memory
When Intel shrink from 90nm ----> to 65nm they had space to add extra L2 cache.
Now Intel 65nm (Core2) Quad-Core total has 8MB L2 cache. And Intel's plan that when they go to 45nm tech, their new Quad-Core will have Total 12MB L2 cache.

[Edit: I'm not saying I'm correct, it is just a theory. (chavvdarrr said) that it might be just a shrink from 80nm to 65nm. That is it: "R600 to R650 not much change"]
"And I understand the point about different node from 80nm to 65nm"

Arty

16-Mar-2007, 20:02

I think I'd nominate 65nm as the "R520 has 32 pipes" for R600. And, whaddyaknow, same source flogging both. :wink:

PCWatch (http://pc.watch.impress.co.jp/docs/2007/0317/cebit06.htm)
Furthermore Bergman touches concerning note PC, "the R600 generation is produced with the process rule of 65nm. Because of this, watt efficiency (efficiency around electric power consumption) it is the expectation which is improved largely ", that you expressed.
I think Bergman's words are the reason for the confusion. :lol:

Other things of note:
CeBIT launch cancelled because of "board redesign"
R600 250W +
RV630 75-100W +
RV610 25-35W +
Launch probably in April (15 days to tech day?)

Geo

16-Mar-2007, 20:27

Something weird is going on with that whole 65nm R600 thing.

Theo at Inq says today:

We have talked with several high-ranking AMD executives and confirmed this rumour. So, from March 16th at 19:59 CET - the R600 which you will buy in stores in May will be 65nm, and will consume almost one third less power than the 80 nanometre one

I don't know that I believe he would be so reckless as to say "several" high-ranking AMD executives confirmed that if not even one had. But the consistent message we are hearing is it just isn't true.

nonamer

16-Mar-2007, 20:30

That sounds absurd. It would mean R600 was scrapped and replaced with a new chip under the same name. I also wouldn't expect that chip in mass quantities either until fall.

neliz

16-Mar-2007, 20:34

Geo, could it be that your sources are having working production samples, just like the R520 developer cards turned out to be XL's?

Shtal

16-Mar-2007, 20:43

The 80nm R600 die - as it was leaked - will now come to life only in a very limited amount of chips, since AMD decided to solve their problems by pulling all resources to go 65nm across the board, including this 720 million heavy monster. This drastically reduces power consumption and enables AMD to clock R600 to 1 GHz or even more. Mass production R600 WILL BE 65nm.

This means that even though AMD is more than six months behind in terms of G80 vs. R600 battle, AMD actually now has six month levy over NV55, more known to people as ex-G81, currently known about G90 (or NV55 in halls of Satan Clara).

I don't know if I could believe it or not. But it sounds cool! :) R600 65nm....

the R600 which you will buy in stores in May will be 65nm, and will consume almost one third less power than the 80 nanometre one.

[Edit: I wonder if G81/NV55 be 65nm also and available exact same time frame]

trinibwoy

16-Mar-2007, 20:46

An R600 at 65nm would be an infinitely bigger coup than any surprises G80 dropped on us. That would be a massive process advantage. But was TSMC really in a position months ago to produce a chip the size of R600 on 65nm?

Arty

16-Mar-2007, 20:47

Is 65HS even available?

3dilettante

16-Mar-2007, 20:58

Rumors had new silicon revisions at 80nm since January. Making a jump to 65 nm would have to be planned pretty far in advance. Why would they continue to make respins on an 80nm variant?

nonamer

16-Mar-2007, 21:08

Shtal

16-Mar-2007, 21:11

Rumors had new silicon revisions at 80nm since January. Making a jump to 65 nm would have to be planned pretty far in advance. Why would they continue to make respins on an 80nm variant?

It could be possibility that on March 30, 2007. ATI/AMD will release 80nm R600 in small availability, then in May 2007 ATI/AMD will start doing big production 65nm R600.

Slappi

16-Mar-2007, 21:20

Jawed

16-Mar-2007, 21:20

Rumors had new silicon revisions at 80nm since January.
Counter-rumoured by "A13 was the last one seen," which hit in October/November I think...

Making a jump to 65 nm would have to be planned pretty far in advance. Why would they continue to make respins on an 80nm variant?
The jump to 65nm would have been in progress anyway, if it was to release in the autumn. I can't say I believe it has really been pulled forwards by 4 months or so.

It's worth noting that clock speed on RV530 is about 94% of R520 - i.e. in the same time frame and the same process, there was practically no difference in core clock. If 65nm was working well for ATI, maybe it wasn't so hard to go with "R650"?

The 80nm transition seemed to give ATI a lot of trouble last year, so perhaps they just have better luck with full nodes rather than half-nodes?...

Anyway, I'm still dubious. If anyone sees a ~330m2 R600 though, phone in pronto, alright?!!

Jawed

3dilettante

16-Mar-2007, 21:25

One would think the cessation of 80nm testing and production in January would have been noted.

Does anyone have stats on the number of wafer starts for that node from the last time R600 revisions were confirmed and now? TSMC may not make such information public, but I know some information exists for other companies with fabs.

The test runs may not too large, but a discontinuation of such efforts by a large customer might be noticeable.

The discontinuation of a production contract of that kind would also impact the financial outlook for TSMC, but I don't know if they need to post that kind of information.

jamis

16-Mar-2007, 21:38

Just ask yourselves this: Why is it that we havn't seen a single photo of retail R600 yet? They've been so bloody secretive havn't they? :grin:

jimmyjames123

16-Mar-2007, 21:53

The title of that article says X2900GT instead of XT. And the G and X aren't even close to eachother on my keyboard. Wonder which one it is...

The title said "GT", but the actual post said "XT". Definitely a typo there.

caffeinated

16-Mar-2007, 22:16

So this "information" about ATI dropping R600 and going to .65 nm is still questionable? To be honest, I just really don't want to hear "When R600..no wait, I mean when R650 is released" for another 5 months. Please spare the public and put out a product for heaven's sake.

Shtal

16-Mar-2007, 22:51

I don't know where you guys are getting your info from but the truth, after talking to Hector himself, is that 80nm was too slow and 65nm would take to long so they compromised. On April 15 AMD will release the X2925 aka R625 on a 72.5nm process with 768mb ram. The chip will run at 906.25 mhz and the card will require a 3x2 and a 3.5x2 connection. It will max at 192W.

Thank me later.

I know that the Inq is not reliably source; some of the times they were wrong, but some of the times they were right.

here goes nothing! I simply hear to much rumors now that I no longer believe until truth reveal from AMD themselves.
Fake or not fake ?? http://www.theinquirer.net/default.aspx?article=38292

Geo

16-Mar-2007, 22:56

Geo, could it be that your sources are having working production samples, just like the R520 developer cards turned out to be XL's?

Neliz, anyone who's been around this industry for any significant length of time knows there is a great deal of sleight-of-hand that goes on. We could go down a long list on both sides of feints, dekes, planted rumours, crippled silicon or boards sent out that way on purpose, leaks of exploratory silicon, presentations telling not-quite-truths handed out to partners with the express knowledge they'd eventually leak to the public, etc etc etc. I've said it before, I'll say it again --we never know for 100% certain until we have a production sample in hand. And we don't with R600 yet.

Having said that, if the R600 released to the public in May is 65nm instead of 80nm, then this one pushes the "extreme pipelines" of R420 into the shade for disinformation campaigns from ATI. As of right now, personally, I'm still sitting on 80nm.

Russell

16-Mar-2007, 23:04

Rumors had new silicon revisions at 80nm since January. Making a jump to 65 nm would have to be planned pretty far in advance. Why would they continue to make respins on an 80nm variant?

This quote actually said a lot to me:

"This drastically reduces power consumption and lets AMD clock the R600 to 1GHz or even more."

Remember the rumors of a new respin (that we assumed would be released at a later date) that allowed 1GHz+ R600 chips? Though this is far from certain, at least two pieces of this puzzle fit together. As for the previous rumours, perhaps they were either intentionally misleading or simply misinformed, much like 99% of the G80 ones were.

I will say this as well. Theo is one of the more sensible, less sensationalist people they have at the Inq. And he's saying this like it's 100% certain, which is far from the usual Inq style (wherein they allow an escape route in case they're wrong). In short, I really really want to believe this is the surprise we were promised by AMD.

Bouncing Zabaglione Bros.

16-Mar-2007, 23:27

I will say this as well. Theo is one of the more sensible, less sensationalist people they have at the Inq. And he's saying this like it's 100% certain, which is far from the usual Inq style (wherein they allow an escape route in case they're wrong). In short, I really really want to believe this is the surprise we were promised by AMD.

It does kind of fit with "you will be pleasantly surprised" and "there is no R600XTX" - that place has been taken by the R680 as XTX. It also might explain the editor day cancellation and the release date pushing back. Someone high in AMD saw that 65nm was coming along nicely, and took a massive execution leap to get ahead of the competition, and join up all the R6xx family.

Being able to get to 1 ghz or better seems likely to give pretty dominant performance and might explain the reason for the 512 bit bus. While introduced on R600, it's real advantage and target was the R680 running at higher speeds.

If R600 was a power hog and running so late it was almost into the back of the refresh, at the same time as 65nm was working really, really well with better, more cost effective yields, maybe the numbers were preferable too.

I dunno. I want to believe as it suddenly brings back that "WOW" factor with added "ZOMG!!" It's one of the few things that would make up for the R600 delay mess, but I'm wary of building up my hopes on something so... exactly in line with what all the fan boys would like! It's too good to be true I tell ya!

Slappi

16-Mar-2007, 23:32

JUST IN!!!

AMD has skipped 65nm and also 32nm 16nm 8nm 4nm 2nm 1nm and gone straight to Ludacris Silicone.

Russell

16-Mar-2007, 23:37

It does kind of fit with "you will be pleasantly surprised" and "there is no R600XTX" - that place has been taken by the R680 as XTX. It also might explain the editor day cancellation and the release date pushing back. Someone high in AMD saw that 65nm was coming along nicely, and took a massive execution leap to get ahead of the competiton, and join up all the R6xx family.

If R600 was a power hog and running so late it was almost into the back of the refresh, at the same time as 65nm was working really, really well with better, more cost effective yeilds, maybe the numbers were preferable too.

I dunno. I want to believe as it suddenly brings back that "WOW" factor with added "ZOMG!!" It's one of the few things that would make up for the R600 delay mess, but I'm wary of building up my hopes on something so... exactly in line with what all the fan boys would like! It's too good to be true I tell ya!

The G80 was also too good to be true. But it is true (though I confess I don't actually own one...the people who do could all by lying and they may not actually exist...).

One thing people keep commenting on is how they couldn't have suddenly switched to 65nm. They're right. They couldn't have. It's not uncommon however for chip firms to work on two processes simultanously for testing and evaluation purposes (much like software firms do...ala Apple's x86 OS they secretly were working on all those years). I would call it an expensive but necessary measure to take. After all, they couldn't have decided some 3-4 years ago when they started designing R600 that 80nm or 65nm would be the manufacturing process de jour. They would typically have developed two and later abandoned one when it became clear that process X would be the way to go. Perhaps in this case they simply left their options open, keeping both on the development table. Then, as you said, they made the final decision to jump ship from 80nm and cancelled the editors day.

Or, perhaps, this is just another false Inq rumour. If that's the case I'll be dissapointed.

icecold1983

16-Mar-2007, 23:46

JUST IN!!!

AMD has skipped 65nm and also 32nm 16nm 8nm 4nm 2nm 1nm and gone straight to Ludacris Silicone.

love that movie.

http://www.youtube.com/watch?v=zvd3kaupZ60

couldnt resist!

caffeinated

17-Mar-2007, 00:28

love that movie.

http://www.youtube.com/watch?v=zvd3kaupZ60

couldnt resist!

:grin: That was great...it made me smile. :lol2:

Geeforcer

17-Mar-2007, 01:30

Ok, let's say R600 is 65nm. When are we going to be able to buy one? If a "simple" respin takes a month.... Redoing a chip in 65nm (not the optical 90->80 shrink, mind you) and getting all the kinks out must take months and months. We are talking late May street time, no?

overclocked_enthusiasm

17-Mar-2007, 02:03

Is TSMC the only fab option for ATI at 65nm? What would it take for AMD's fabs to be setup to manufacture GPUs?

Also, May seems to be pretty darn optimistic for R650 being pulled in. I have a feeling May will turn into June in short order.

Razor1

17-Mar-2007, 02:38

AMD has SOI 65nm which might complicate the matter?

Silent_Buddha

17-Mar-2007, 03:01

While I'm not putting a lot of stock into the retail board is 65 nm, there are a few things that could be inferred if true.

1. All OEM boards are 80 nm R600. Thus the need for larger cooling than retail channel could accept due to heat.

2. All Retail board are 65nm R600/650 whatever. Considering how late

I mean, with virtually no substantiated rumors of any respins past A13 which was out Nov(?) of last year, and with virtually no info/leaks on anything remotely retail...

Also, this would jive with some rumors that R600 was hotter/no able to reach the performance they expected.

I would think they've been at least playing with 65 nm for a while, and maybe the redesign for 65nm was a lot easier than they expected?

I'd imagine a move to 65 nm would be the most closely kept secret ever, and likewise would hugely surprise the competition who was planning on competing with a 80 nm chip.

Personally, I think this is probably the least likely situation. However, with rumors increasing about the possibility of a 65nm retail R600...who knows.

Regards,
SB

Bouncing Zabaglione Bros.

17-Mar-2007, 03:20

I have to say, this 65nm rumour has for the last couple of days almost exactly followed the same pattern as the 512 bit bus rumour... and look what happened there.

Tim Murray

17-Mar-2007, 03:59

I have to say, this 65nm rumour has for the last couple of days almost exactly followed the same pattern as the 512 bit bus rumour... and look what happened there.
Well..

1. A13 apparently came back in January.
2. What are the odds that they would get 65nm right the first time and not need another spin on a chip of R600's size?
3. Look up how much a 65nm tapeout costs. Assume an 80HS one is slightly less. They're not going to just eat the 80HS costs like that.

Sorry guys, I just don't see any indication whatsoever that it's true.

Shtal

17-Mar-2007, 04:31

Just a question?

Does anybody here suspect/think what would happing on March 30, 2007. Or rumored from Germany was right all the lone. About the delay till May 2007.

Is rumorer from Germany still right or wrong?? no March Lunch.... :( or both right and wrong.

Cuthalu

17-Mar-2007, 06:16

icecold1983

17-Mar-2007, 06:39

man with all these rumors its getting hard not to expect a big jump over g80.

Unknown Soldier

17-Mar-2007, 06:47

To be announced in May?

Erm.. ok ... much later than I expected... still if it's 65nm .. .then good fer them.

Just let teh price be reasonable for all the waiting that we've done.

US

Russell

17-Mar-2007, 07:12

To be announced in May?

Erm.. ok ... much later than I expected...

US

The talk around teh intarweb, including this forum, has been a May release for some time now :)

Geeforcer

17-Mar-2007, 07:39

Well, the May launch and 65nm sort of add up.

Techno+

17-Mar-2007, 08:24

http://www.xbitlabs.com/news/video/display/20070316211223.html

geforce 8800 ultra, 8600, 8500, 8400, 8300 to have a unified launch.

neliz

17-Mar-2007, 10:25

for what it's worth, halfway trough february digitimes had an article about 65nm production for only a few clients, Qualcom, Xilinx and Broadcom were the most pronounced ones.
They expected production to ramp up only during the third quarter of 2007.
So in any form, unlike R520 which was pushed on even though their goals were not reached AMD did the smart thing and pull R650 forward which just seems to benefit the end-users.

hoom

17-Mar-2007, 11:01

May release someday :roll:

Arnold Beckenbauer

17-Mar-2007, 11:24

http://developer.valvesoftware.com/wiki/Valve_time
-----------------------------------------------------------------------
http://forums.ocworkbench.com/bbs/showthread.php?p=408085

Small question: The X2900XT is clocked 10-15 % lower than XTX?

vertex_shader

17-Mar-2007, 11:37

man with all these rumors its getting hard not to expect a big jump over g80.

The expectations very high, AMD looks confident about R600, but i don't know from here they are just looks confident because of the business partners (AMD can't say if the R600 not so great), or really confident about the R600, time will tell only, still some weeks left.

Unknown Soldier

17-Mar-2007, 11:46

May release someday :roll:

hahha .. great pun :D

I would change it slightly though. ;)

Geo

17-Mar-2007, 11:57

I have to say, this 65nm rumour has for the last couple of days almost exactly followed the same pattern as the 512 bit bus rumour... and look what happened there.

True. But I still don't think so this time. There's been some "I spy with my little eye" at cebit now and the word filtering back is that it's a honker.

vertex_shader

17-Mar-2007, 12:20

R600: Exclusive co-operation between AMD and valve?
Link (http://www.google.com/translate?u=http%3A%2F%2Fwww.pcgameshardware.de%2F %3Farticle_id%3D571781&langpair=de%7Cen&hl=en&ie=UTF8)

jamis

17-Mar-2007, 12:21

Just a reminder: it's not only the Inq which has 65nm rumors, but these as well:
http://www.techpowerup.com/?27451
http://www.vr-zone.com/?i=4796
http://www.k-hardware.de/news.php?s=K1fbfa7da649e0ce1f232e8a064ef0dc0&news_id=6395
The 65nm story originated from K-hardware which both vr-zone and techpowerup are referring as their source, The Inq may be following K-hardware too but not telling (and lying about that "we have talked with several high-ranking AMD executives" bit)

vertex_shader

17-Mar-2007, 14:05

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

Nothing special in this side of the card, looks like as a x1950xtx.

I not understand why AMD not show the card in Cebit without working to the public? only some weeks left to the launch.

kyniskos

17-Mar-2007, 15:01

Maybe they finally took some lessons in marketing. Let the audience invent their own fantasies and make sure you are the talk of the town. By barely showing of any product, they are getting us into a frenzy.:idea:

Bouncing Zabaglione Bros.

17-Mar-2007, 15:20

True. But I still don't think so this time. There's been some "I spy with my little eye" at cebit now and the word filtering back is that it's a honker.

It wouldn't be the first time the "real" product has been held back while the lower specced versions, or even engineering samples have been shown as the intended product. More for the disinformation campaign.

Don't squash our 65nm speculation just yet!

Geo

17-Mar-2007, 16:27

Don't squash our 65nm speculation just yet!

Far be it from me to crush your dreams! Having said that, my apologies in advance if I slip and call you "Brimstone". :smile:

DemoCoder

17-Mar-2007, 16:48

You would have thought ATI would have learned their lesson about launching nextgen cards in partnership with Valve. :)

nicolasb

17-Mar-2007, 18:10

Not sure if this has been posted yet:

http://www.cdrinfo.com/Sections/News/Details.aspx?NewsId=20005

One interesting quote:

Bergman also said that AMD is preparing an 65nm version of the R600, that will feature a lower power consumption and will be used in the future DirectX 10-enables notebooks, possibly available later this year. That probably explains where all the "R600 is 65nm" rumours are coming from: there will be a 65nm version, but later in the year and intended for notebooks.

There's also a really scary quote:

The R600 could be considered as the second generation of the ATI Xenos GPU, which powers Microsoft's Xbox 360 game console. It features 48 Unified Shaders and fully supports DirectX 10 and Shader Model 4.0. If R600 really only has 48 shaders then ATI is in trouble.

INKster

17-Mar-2007, 18:18

Not sure if this has been posted yet:

http://www.cdrinfo.com/Sections/News/Details.aspx?NewsId=20005

One interesting quote:

That probably explains where all the "R600 is 65nm" rumours are coming from: there will be a 65nm version, but later in the year and intended for notebooks.

There's also a really scary quote:

If R600 really only has 48 shaders then ATI is in trouble.

There are a number of inaccuracies in that article.
RV610 a mainstream part and RV630 a low end one ?

I wouldn't give it too much credit just yet.

no-X

17-Mar-2007, 18:29

If R600 really only has 48 shaders then ATI is in trouble.
And why is R600 3-times bigger than Xenos? Maybe ATi suffed initially small R600 with polystyrene to made it look more majestic? http://forum.3dfx.cz/images/smiles/icon_shy.gif

INKster

17-Mar-2007, 18:32

And why is R600 3-times bigger than Xenos? Maybe ATi suffed initially small R600 with polystyrene to made it look more majestic? http://forum.3dfx.cz/images/smiles/icon_shy.gif

Maybe because Xenos is part of a console, so there's no need for OpenGL, DirectX (at least not like the desktop version, tied to Windows) ?
Also, Xenos will never have to deal with resolutions like the ones R600 has to endure in the future (2560x1536, insane AA and AF levels, etc).

Kaotik

17-Mar-2007, 18:33

Not sure if this has been posted yet:

http://www.cdrinfo.com/Sections/News/Details.aspx?NewsId=20005

One interesting quote:

That probably explains where all the "R600 is 65nm" rumours are coming from: there will be a 65nm version, but later in the year and intended for notebooks.

There's also a really scary quote:

If R600 really only has 48 shaders then ATI is in trouble.

Wasn't it confirmed that R600's have 320 multiply-accumulate units already? That doesn't fit in the "48" no matter how you try

w0mbat

17-Mar-2007, 19:59

The R600 could be considered as the second generation of the ATI Xenos GPU, which powers Microsoft's Xbox 360 game console. It features 48 Unified Shaders and fully supports DirectX 10 and Shader Model 4.0.

Yes, the Xenos has 48US. But they dont say anything about R600.

BTW: R600 Techdemo: http://www.youtube.com/watch?v=EJ6aMxPh6k0

overclocked_enthusiasm

17-Mar-2007, 20:37

Bouncing Zabaglione Bros.

17-Mar-2007, 21:02

http://www.theinquirer.net/default.aspx?article=38300

"WITH SO MANY HACKS constantly talking to AMD's partners about the R600 and tapping in the dark about the launch, everything is at fever pitch in the press office.
But the details are as follows. AMD is preparing the biggest launch in history of graphics chips.

Since, we understand, the company pulled an unexpected move with a complete switch to a 65 nm process at TSMC, it seems that a new manufacturing alliance is emerging, since AMD is currently in various partnership deals with IBM and so Sony, Toshiba, Chartered, UMC and of course, TSMC.

AMD will probably ship well over 100 million 65 nanometre chips in 2007, and decided to start their own graphics adventure with a bang. We learned that the firm is inviting a ton of partners for a huge 3D-wide PR briefing - close to 200 journos will arrive at the press event and the plan is for each and every member of the press to get a taste of 65 nano chippery. µ "

Not much new there but it is fresh in terms of when it was posted.

That might explain why Editor's Day was cancelled - it wasn't big enough for the massive launch of AMD into the graphics arena. Looks like they are going for the "hearts and minds" and trying to get themselves a rep for graphics that the likes of Intel doesn't have. They are relaunching themselves as a company, as well as R600 as a range of graphics products.

If it's true, it finally looks like ATI is learning how to do PR - even if it took AMD buying them to do it for them.

R300King!

17-Mar-2007, 21:06

That Youtube Ruby demo looks nice. I didn't get any sound though. Was there suppose to be sound or was that my computer? Too bad it's not in higher rez.
Can someone put it up on Youtube in higher rez if you have it, please? I'm not gready. :D

So 65nm has been Inqufirmed? (Inq+confirmed)

Definition:
In·qu·firm (In-quah-ferm)
1. Absolutely, positively confirmed, maybe.
2. Confirmation taken with a pillar of salt.
3. Pure speculation and guessing by the Inquirer.

;)

zgemboandislic

17-Mar-2007, 21:13

Not sure if this has been posted yet:

http://www.cdrinfo.com/Sections/News/Details.aspx?NewsId=20005

One interesting quote:

That probably explains where all the "R600 is 65nm" rumours are coming from: there will be a 65nm version, but later in the year and intended for notebooks.

There's also a really scary quote:

If R600 really only has 48 shaders then ATI is in trouble.

How could it possibly have only 48 shaders when R580 has the same number? Pure bullshit

IbaneZ

17-Mar-2007, 21:16

BTW: R600 Techdemo: http://www.youtube.com/watch?v=EJ6aMxPh6k0

Cool, thanks!

Ruby still got it. I can't wait to see that demo running on my brand new R600. :smile:

Twinkie

17-Mar-2007, 21:35

With this much hype built everyday.. i wonder if AMD can deliver.

Bouncing Zabaglione Bros.

17-Mar-2007, 21:54

With this much hype built everyday.. i wonder if AMD can deliver.

Considering how bored everyone was getting with "R600 fatigue" from months of nothing but speculation (and even that was tailing off), I'm surprised that things have again managed to get whipped back up into a frenzy of interest. Maybe a cunning plan, eh?

pjbliverpool

17-Mar-2007, 22:03

Yes, the Xenos has 48US. But they dont say anything about R600.

BTW: R600 Techdemo: http://www.youtube.com/watch?v=EJ6aMxPh6k0

Holy CGI Batman!

That looks AMAZING!

Makes me wonder what the hell nvidia's demo team spend their days doing! I want a demo like that for my G80! :sad:

The very final scene was the best for me, the smile looked incredible!

Twinkie

17-Mar-2007, 23:26

Holy CGI Batman!

That looks AMAZING!

Makes me wonder what the hell nvidia's demo team spend their days doing! I want a demo like that for my G80! :sad:

The very final scene was the best for me, the smile looked incredible!

Adrianne looks more realistic especially the eye.Ruby still has that cartoony look.

Although the ruby demo looks better overall.

Russell

17-Mar-2007, 23:28

What really strikes me about this 65nm rumor is how suddenly everybody on pretty much every site is talking about it as though it's 100% certain. There's a few naysayers in forums (such as here) but that's about it. Even with the 512-bit thing there were sites that didn't buy it at first.

Geo

17-Mar-2007, 23:37

Silent_Buddha

17-Mar-2007, 23:47

If everyone is starting to believe that R600/650 will be 65 nm and launched with the rest of the family, and it isn't true, wouldn't an ATI/AMD rep clarify things? Afterall can you imagine the disappointment of their supporters if everyone is claiming 65 nm and miracles...and it turns out to be just plain old 80 nm and mortal?

I've still got a healthy dose of skepticism that it's 65 nm, but I'm open the possibility that it might happen.

Regards,
SB

INKster

17-Mar-2007, 23:52

I've still got a healthy dose of skepticism that it's 65 nm, but I'm open the possibility that it might happen.

Regards,
SB

It will happen.
The question is "when". ;)

Power_man

18-Mar-2007, 00:01

It will happen.
The question is "when". ;)

The answer is "soon" :-P .

INKster

18-Mar-2007, 00:11

The answer is "soon" :-P .

How "soon" is your "soon" ?

caffeinated

18-Mar-2007, 01:19

Hmmm...AMD is already shipping 65 nm chips. I know because I have one in my machine...but it's a Brisbane. Could they have mixed up the switch of all processors to 65nm with the R600? Reading that second Inq article, the wording could actually work either way. Sure, journalists will get a taste of 65nm chippery...Brisbane, etc. It is worded funny.

Twinkie

18-Mar-2007, 02:27

This place should be renamed to beyond*English* :lol:

aeryon

18-Mar-2007, 02:28

Direct from AMD :
R600 is 80nm
Rv610/630 are 65nm
period

Bouncing Zabaglione Bros.

18-Mar-2007, 02:36

Direct from AMD :
R600 is 80nm
Rv610/630 are 65nm
period

The rumour isn't that R600 is on 65nm, it's that the refresh (R650/R680) is on 65nm and is launching at the same time, with R600 as a short production run for OEMs, and mainstream retail cards being the 65nm refresh.

AMD could tell you that R600 is on 80nm perfectly correctly. That still doesn't tell us if they are launching a 65nm refresh with a different product number at the same time or not.

trinibwoy

18-Mar-2007, 02:43

Yeah it looks like everybody took the 65nm R600 notebook tidbit way too far. Anybody who is disappointed because R600 isn't 65nm has only themselves to blame - ATi certainly hasnt made that promise. Although the inquirer is talking as if they've gotten the inside scoop - but given their track record they might consider misinterpreting public statements as "scoops". Who knows?

The Ruby demo is very cool though. Nvidia's demos are suckage in comparison in terms of entertainment value. I didn't get any sound either R300King so don't sweat it :)

Jawed

18-Mar-2007, 02:59

trinibwoy

18-Mar-2007, 03:02

Yeah I noticed that too lol. It was like she was wearing a hair-lmet :lol:

icecold1983

18-Mar-2007, 03:17

ill be disappointed if r600 isnt at least 30% faster across the board.

Bouncing Zabaglione Bros.

18-Mar-2007, 03:21

What's up with her hair? When she's in the cockpit her hair moves, yet when she's boarding at 100 miles per hour her hair's rock steady. Has she got super-hero hair gel or something that turns on and off: off when she's flirting and on when she's boarding? Or maybe her hair's magnetic and when she puts her hood up it locks it all in place?

We need to know.

Jawed

If you're asking based on that awful quality Youtube video, when Ruby's snowboarding she actually has the hood up on her coat. You can briefly see it on one of the close up shots as she's coming down the mountain. The front of the hair sticking out of the hood is actually moving in the wind. At least that's what I think I can see amongst all the macroblocks.

aeryon

18-Mar-2007, 03:22

The rumour isn't that R600 is on 65nm, it's that the refresh (R650/R680) is on 65nm and is launching at the same time, with R600 as a short production run for OEMs, and mainstream retail cards being the 65nm refresh.

AMD could tell you that R600 is on 80nm perfectly correctly. That still doesn't tell us if they are launching a 65nm refresh with a different product number at the same time or not.

I will be very very surprised if high end R6xx part on 65nm will be available before end of 2007 since now TSCM 65nm hi speed process has poor yield and nobody uses it in large scale production...

caffeinated

18-Mar-2007, 03:31

Yeah I noticed that too lol. It was like she was wearing a hair-lmet :lol:

LOL for coining a new word..a hairlmet. I think I'll be using that in RL as soon as possible :lol: I know a woman with a hairlmet!

trinibwoy

18-Mar-2007, 03:46

ill be disappointed if r600 isnt at least 30% faster across the board.

Well that's a reasonable expectation given that they know exactly what they're aiming for and have known for quite a long time.

LOL for coining a new word..a hairlmet. I think I'll be using that in RL as soon as possible :lol: I know a woman with a hairlmet!

Heh, as long as you say it behind her back :)

Shtal

18-Mar-2007, 04:31

Since, we understand, the company pulled an unexpected move with a complete switch to a 65 nm process at TSMC, it seems that a new manufacturing alliance is emerging, since AMD is currently in various partnership deals with IBM and so Sony, Toshiba, Chartered, UMC and of course, TSMC.
http://www.theinquirer.net/default.aspx?article=38300
AMD will probably ship well over 100 million 65 nanometre chips in 2007, and decided to start their own graphics adventure with a bang.

Can somebody clear things up for me?
If ATI for example were never merge with AMD or in other words bought up by AMD.

ATI then would never had opportunity to make sudden change and move directly to 65nm tech without AMD.

INKster

18-Mar-2007, 04:32

http://www.theinquirer.net/default.aspx?article=38300

Can somebody clear things up for me?
If ATI for example were never merge with AMD or in other words bought up by AMD.

ATI then would never had opportunity to make sudden change and move directly to 65nm tech without AMD.

AMD's 65nm SOI process has nothing to do with TSMC's 65nm tech.
So, ATI R600 65nm derivatives would still be a reality, takeover or not.

Shtal

18-Mar-2007, 04:41

AMD's 65nm SOI process has nothing to do with TSMC's 65nm tech.
So, ATI R600 65nm derivatives would still be a reality, takeover or not.

The reason I ask is because the Inq quoted "since AMD is currently in various partnership deals with IBM and so Sony, Toshiba, Chartered, UMC and of course, TSMC."

I wasn't sure that this information was important to know that AMD has partnership deals. Or why theinquirer has to phrase that?

Kaotik

18-Mar-2007, 04:43

The reason I ask is because the Inq quoted "since AMD is currently in various partnership deals with IBM and so Sony, Toshiba, Chartered, UMC and of course, TSMC."

I wasn't sure that this information was important to know that AMD has partnership deals. Or why theinquirer has to phrase that?

It's not relevant, AMD stated that ATI chips will continue to be built by TSMC and UMC, just like before, when the merger happened.

Shtal

18-Mar-2007, 04:45

It's not relevant, AMD stated that ATI chips will continue to be built by TSMC and UMC, just like before, when the merger happened.

Thanks :)

[Edit: But still, AMD might have some kind effect on ATI to speed up the process.]

Arty

18-Mar-2007, 05:20

Fuad is now saying that there will also be a Lasso version of R600 (external).

Kaotik

18-Mar-2007, 05:22

Fuad is now saying that there will also be a Lasso version of R600 (external).

The way I've understood it, Lasso is independent of the card - as in, you should be able to stick pretty much any, if not any, PCIe card to it

Arty

18-Mar-2007, 05:46

The way I've understood it, Lasso is independent of the card - as in, you should be able to stick pretty much any, if not any, PCIe card to it
Yes, but the leaked picture showed a low-profile X850. The Asus XG Station also has the ability to house relatively low-profile cards only. Maybe I shouldnt read too much into it.

Unknown Soldier

18-Mar-2007, 08:20

OCZ wants in on R600

The spokesperson was quick to state that it has no exclusivity deals in place and that the company also has a very strong working relationship with ATI (and now AMD), mainly thanks to the CrossFire certification programme on some of its memory modules.

OCZ hopes to make full use of this relationship in order to start shipping products based on the eagerly anticipated and somewhat delayed R600 GPU.

News Source: Bit-Tech (http://www.bit-tech.net/news/2007/03/16/ocz_wants_in_on_r600/)

zgemboandislic

18-Mar-2007, 08:31

It's not relevant, AMD stated that ATI chips will continue to be built by TSMC and UMC, just like before, when the merger happened.

And nVidia stated that unified shaders are not the way to go yet...

HAL

18-Mar-2007, 08:46

http://translate.google.com/translate?u=http%3A%2F%2Fwww.sdbear.com%2F2%2Flib% 2F200702%2F15%2F20070215056.htm&langpair=zh%7Cen&hl=it&ie=UTF-8&oe=UTF-8&prev=%2Flanguage_tools
:wink:

Anarchist4000

18-Mar-2007, 09:02

They didn't necessarily state that all ATI/AMD chips will be built by a specific company. TSMC could be handling the midrange parts while IBM(for example) creates the high end part. If the midrange parts are getting physically larger more fab space would be required.

I'd still consider it possible that UVD, HDMI/audio, or possibly the ROPs could be pulled off of the core and included as a separate chip on the cards. Those in turn could be manufactured on 65nm by a company other than TSMC. High clockspeeds aren't really required for that functionality so including them onto the main core wouldn't make a whole lot of sense, similar to NVIO.

Arty

18-Mar-2007, 09:03

Thats more than a month old, since then we've had X2900 yada yada yada .. :lol:

Geo

18-Mar-2007, 14:11

ill be disappointed if r600 isnt at least 30% faster across the board.

So you'll be disappointed if 8800 Ultra isn't 30% faster across the board too?

Unknown Soldier

18-Mar-2007, 15:46

So you'll be disappointed if 8800 Ultra isn't 30% faster across the board too?

That's 8900 Ultra for you mister!

/me runs :D

trinibwoy

18-Mar-2007, 15:59

So you'll be disappointed if 8800 Ultra isn't 30% faster across the board too?

I don't get the correlation.....?

Unknown Soldier

18-Mar-2007, 16:10

Ppl expect the 8800 Ultra (http://www.xbitlabs.com/news/video/display/20070316211223.html) to be at least as fast as the R600 with it's faster memory and faster core.

US

Geo

18-Mar-2007, 18:16

I don't get the correlation.....?

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? (Unless, of course, we're just discussing ATI fanbois who think their favorite IHVs parts should be 30% faster across the board because. . . they want them to be, and that's sufficient logic; or NV fanbois who want to create an expectations level for ATI/AMD that their own favorite company can't match either so that they can claim mock disappointment in a product they had no intention of buying in the first place). I've often heard people say its a timing issue, and that parts released later must be faster than parts released sooner. Well, if that's the logic being applied here than the expectations for R600 and this supposed Ultra should be about the same, right?

icecold1983

18-Mar-2007, 18:19

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? (Unless, of course, we're just discussing ATI fanbois who think their favorite IHVs parts should be 30% faster because. . . they want them to be, and that's sufficient logic; or NV fanbois who want to create an expectations level for ATI that their own favorite company can't match either). I've often heard people say its a timing issue, and that parts released later must be faster than parts released sooner. Well, if that's the logic being applied here than the expectations for R600 and this supposed Ultra should be about the same, right?

smaller process, insane bw, more memory, more transistors. ati isnt my favorite ihv and im not a fanboy. get real

willardjuice

18-Mar-2007, 18:26

I am actually really excited for the audio portion. Ironically, the R600 will be the first sound card that is able to send multi-channel (up to 7.1) uncompressed linear PCM out of a computer digitally. This is a pretty significant step in the audio frontier. Who even cares about its graphical performance? :razz:

Kaotik

18-Mar-2007, 18:26

nicolasb

18-Mar-2007, 18:29

RV6xx's have the audio controller, not R600Source?

Geo

18-Mar-2007, 18:30

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.

The reality of the matter is there are no reasonable expectations for "across the board". R300 vs NV30 couldn't pass the standard you just blithely rolled out.

Edit: Here you go: http://www.sharkyextreme.com/hardware/videocards/article.php/3211_1449721__3 The "disappointing" R300 failing to beat GF4 (released 8 months earlier) by 30% "across the board". In fact, look at those 800x600 numbers vs R8500! Disaster, doom, etc.