Next NV Infomania

[Dave Baumann]

Sorry, yeah, that should be G71.

[Geo]

So, we have a name, some promo advert material, and reasonable looking specs. We have a pic of the card (probably). We have some performance info of the "give with one hand and take away with the other" variety that marketing people just love. Y'know, TWICE AS FAST (in some specific shaders).

So this really does look like a card announcement hittings its checkpoints on the way to Der Tag. But where is the heavy breathing at www.nvidia.com if they are less than a week out??

Of course, we have none of this for R520. . .

[Chalnoth]

So this really does look like a card announcement hittings its checkpoints on the way to Der Tag. But where is the heavy breathing at www.nvidia.com if they are less than a week out??
I don't remember "heavy breathing" on the nVidia site before the launch of the NV40. Could be wrong, though.

Anyway, we typically get info on when NDA's expire a week or two before launch. So nVidia's next high-end product is probably still a bit further out. This may be a hint that nVidia's trying harder this time to ship on their announcement date.

[digitalwanderer]

Last I heard on the G70 performance was 50% more than a 6800 GT. :?

[DegustatoR]

So this really does look like a card announcement hittings its checkpoints on the way to Der Tag. But where is the heavy breathing at www.nvidia.com if they are less than a week out??
They're more than a week from out :) But this time we can expect a resonably decent retail presence right from the start. I think that first G70s will hit retail in July or August at worst.

[DegustatoR]

Last I heard on the G70 performance was 50% more than a 6800 GT. :?
Depends on what performance you're comparing...

[Evildeus]

The launch is on the 22/6 isn't it?

[Geo]

The launch is on the 22/6 isn't it?

Seen that date suggested recently; haven't accepted it yet --unless someone here knows better.

[tEd]

Only small improvements for AA,AF? :(

[trinibwoy]

Only small improvements for AA,AF? :(

Yeah I'm not too thrilled by the released specs. What exactly have these "hundreds of people" been doing "for several years" to come up with this chip. If G70 is just a slightly tweaked NV40 shouldn't that statement have been applied to the NV40 instead?

[Dave Baumann]

Given that we are still slap bang in the middle of the DX9 / SM3.0 cycle, what do we expect? IMO, it'll still be interesting.

[JoshMST]

Only small improvements for AA,AF? :(

Yeah I'm not too thrilled by the released specs. What exactly have these "hundreds of people" been doing "for several years" to come up with this chip. If G70 is just a slightly tweaked NV40 shouldn't that statement have been applied to the NV40 instead?

Well, if you are unhappy with this product release, why not apply for a job at NVIDIA and whip them into shape? I am pretty sure that designing a 300 million transistor part would be a walk in the park for you! :wink:

As for me, I like new stuff. I don't care if it is truly revolutionary or evolutionary, I like the change nonetheless. I think that this will be a fun summer, how about you? Well... unless you live in the southern hemispheres... then it will be an interesting winter.

[trinibwoy]

Well, if you are unhappy with this product release, why not apply for a job at NVIDIA and whip them into shape? I am pretty sure that designing a 300 million transistor part would be a walk in the park for you! :wink:


You missed the point. I never said I wasn't impressed with NV40. But the comments I quoted above were said in reference to the RSX architecture. If, as speculated, RSX is very similar to G70 but G70 is a slightly tweaked NV40 then the "hundreds of people" working during "several years" to develop the part doesn't make sense to me.

[Ailuros]

Well, if you are unhappy with this product release, why not apply for a job at NVIDIA and whip them into shape? I am pretty sure that designing a 300 million transistor part would be a walk in the park for you! :wink:



That's according to my understanding RSX claimed transistor count. Not only is RSX@90nm/550MHz, but AFAIK it's not the same exact GPU as the desktop variant either. Damn close probably, but I have severe doubts that G70@110nm will count as many transistors. Could be wrong though.

[psurge]

Well it remains to be seen whether G70 is a slightly tweaked NV40 or not. I suppose the other possibility is that the RSX isn't in the same generation as G70. Out of curiosity - does anyone know when G70 taped out?

Cheers,
Serge

[JoshMST]

I think a lot more work was done on RSX than you give them credit for. While it will probably have all of the features of G70, it will undoubtedly have more. The area where the most work was probably done was the interface with the rest of the system. The way it supposedly can intelligently allocate and request resources is something that looks to be much more complex than designing a PCI-E interface. I of course could be wrong, but the way the RSX works together with the Cell probably isn't an easy task.

[Geo]

I think a lot more work was done on RSX than you give them credit for.

If that's true, Josh, then it is also a lot more work than NV themselves --including Jen-- is giving themselves credit for. They have been very vocal to the financial analysts that the Sony money is practically "free" from a resource committment pov. Jen actually told them that revenue is "all margin".

[nAo]

small improvements for AA -> still no MSAA on 64bit rendertargets :(
R500 with AA support on 32/64bit fp render target and almost free 4x (at least with 32 bit buffers) will probably have an edge over G70/RSX image quality.

[Geo]

small improvements for AA -> still no MSAA on 64bit rendertargets :(

Well, I'm not so sure in a marketing-aimed document that would qualify as "large", even tho the impact in games is significant.

[nAo]

dunno, my bet is "small AA improvements" -> gamma corrected AA.

[DegustatoR]

dunno, my bet is "small AA improvements" -> gamma corrected AA.
...which you can enable in current Forceware in OpenGL on any NV4x right now :)

[nAo]

dunno, my bet is "small AA improvements" -> gamma corrected AA.
...which you can enable in current Forceware in OpenGL on any NV4x right now :)
Maybe this time it will be exposed for DX applications too ;)

[ChrisRay]

dunno, my bet is "small AA improvements" -> gamma corrected AA.
...which you can enable in current Forceware in OpenGL on any NV4x right now :)
Maybe this time it will be exposed for DX applications too ;)

It can be exposed in Direct3d or OpenGL. Its just buggy and if you switch from gamma corrected back to normal AA. Your gamma is buggy until you reset your PC. Its easily exposed in Rivatuner or Nhancer.

[Reverend]

If, as speculated, RSX is very similar to G70
Can you point me to such speculations?

[trinibwoy]

If, as speculated, RSX is very similar to G70
Can you point me to such speculations?

Sure, about a bazillion threads in the console forum during E3 mania. Not only that but Nvidia reps pretty much said that RSX is based on their next gen GPU and there was not much effort put into the Sony deal on their part.

[Geo]

If, as speculated, RSX is very similar to G70
Can you point me to such speculations?

PS3 is going to be all gross margins for us.

That's what Jen told the financial analysts two weeks ago.

Then there is this interview that hammers the point home about as hard as you can without fiery letters in the sky:

http://www.xbitlabs.com/articles/editorial/display/ces2005.html

Are they bullsh*tting? In this industry one can never entirely discount the possibility I suppose. :)

[Ailuros]

If, as speculated, RSX is very similar to G70
Can you point me to such speculations?

Sure, about a bazillion threads in the console forum during E3 mania. Not only that but Nvidia reps pretty much said that RSXis based on their next gen GPU and there was not much effort put into the Sony deal on their part.

Watch the highlight and what it should indicate. Albeit the differences between R500 and R520 are obviously huge compared to RSX/G70, I could eventually also say that R520 is based on R500 and I could just mean that the belong both to the SM3.0 family of GPUs.

There's quite a difference between the above and "very similar".

[Ailuros]

If, as speculated, RSX is very similar to G70
Can you point me to such speculations?

PS3 is going to be all gross margins for us.

That's what Jen told the financial analysts two weeks ago.

Then there is this interview that hammers the point home about as hard as you can without fiery letters in the sky:

http://www.xbitlabs.com/articles/editorial/display/ces2005.html

Are they bullsh*tting? In this industry one can never entirely discount the possibility I suppose. :)

Sorry folks but this particular interview sounds as it should:

And we’ve been working with them to produce a customized version that is customized specifically to connect that to the cell processor, so that they could work together.

We’ve been working with them on the whole development platform for the software developers, which is one of the advantages to them with going outside, not doing it inside just in terms of the availability of tools, the availability of systems and things. What we developed was to come up with content and quickly and efficiently. In fact we engaged with them quite some time ago, so we‘ve been working with them for quite some time on the development platform overall for the new technology.

If you think that desktop G70 will be customized for Cell or in extension a UMA based architecture or that they pulled this one off in half a year, then you haven't been reading between the lines enough.

It's not that RSX is a quantum leap compared to G70, but "very similar" and minimal time is definitely out of place.

[Geo]

Okay, and how do you get "all gross margins" unless your expenses were very minimal?

[Mulciber]

Okay, and how do you get "all gross margins" unless your expenses were very minimal?

if it was based on a nearly complete g80 there would be no contradiction

[Xmas]

Oh, and one of our Belgium members over at Elite Bastards [...]

Since I'm dutch[...]
:lol: :lol:


dunno, my bet is "small AA improvements" -> gamma corrected AA.
...which you can enable in current Forceware in OpenGL on any NV4x right now :)
But it's slower.
I also hope for gamma-adjusted downsampling in hardware, it shouldn't be overly expensive (drop quincunx instead!). But, if they want to do it The Right Way (tm), they need gamma-adjusted alpha blending as well.

[Chalnoth]

Don't forget texture filtering.

[PeterAce]

If, as speculated, RSX is very similar to G70
Can you point me to such speculations?

http://www.beyond3d.com/forum/viewtopic.php?p=521398#521398

[Geo]

Okay, and how do you get "all gross margins" unless your expenses were very minimal?

if it was based on a nearly complete g80 there would be no contradiction

If that were true, then ATI could say the same thing about R500. But they aren't saying anything like that. The two companies sound very different in their own descriptions of the amount of effort they put into these two products.

[digitalwanderer]

Oh, and one of our Belgium members over at Elite Bastards [...]

Since I'm dutch[...]
:lol: :lol:

There's a difference between being Dutch and being Belgium? :|


























(Ok, I goofed. :oops: )

[Chalnoth]

There's a difference between being Dutch and being Belgium? :|
Yes. One is a state of being a nation.

[rendezvous]

If you are Bulgium you are a country located in western Europe :)

If you are dutch you are a from the Netherlands, (another country in western Europe).

[digitalwanderer]

If you are Bulgium you are a country located in western Europe :)

If you are dutch you are a from the Netherlands, (another country in western Europe).
Well they're all in Western Europe, they're practically the same place... http://66.224.5.66/board/images/smilies/bleh2.gif











;)

[nutball]

Well they're all in Western Europe, they're practically the same place... http://66.224.5.66/board/images/smilies/bleh2.gif

lol, you crazy canuck :P

[digitalwanderer]

lol, you crazy canuck :P
Ah, we're all on the same continent...I get it. http://69.93.88.162/forum/images/smiles/rofl.gifhttp://69.93.88.162/forum/images/smiles/rofl.gifhttp://69.93.88.162/forum/images/smiles/rofl.gif

Back on topic before we honk off the powers that be and get this excellent thread closed, does anyone know anything else about nVidia's new AA? I'm REALLY interested to hear more about that.

I guess raw speed/power just doesn't impress me as much anymore since there really isn't a whole lot taking advantage of it yet, I want to know how I can use that speed/power to make me games look better.

[JoshMST]

I think Jen's comment about it being all gross margins is because they are not producing the chip, so the money they receive from Sony will be "free" that way. Now, while the actual deal is not for public consumption, we can expect that Sony paid NVIDIA up front most of the development costs for designing the RSX (just as Microsoft did with the original X-Box GPU/MCP). So, with Sony paying for development, and then paying NVIDIA a royalty for each chip produced, then it really is "free money". Usually NVIDIA has to pay the development costs for its GPUs, then produce the part, and only when the board partners start buying the chips do they make a profit. This deal is all profit for NVIDIA.

NVIDIA also has found a killer app for Cg, and all PS3 developers will have to use Cg. So, when these developers port over the PS3 games to the PC, then they will probably use Cg. The tools for the PS2 were terrible, and many are expecting the tools for the PS3 to be nearly on par with what Microsoft offered with the X-Box (really outstanding tools). I bet that Sony also paid for NVIDIA to develop many of these tools, and because Cg was already around, that is again another "easy" profit for NVIDIA.

[Geo]

Ah, so you're a fan of the theory that this deal was consummated long before it was announced? Well, this threadlet will be a good one to come back to for scorekeeping somewhere down the road. I'll promise to not point and laugh if the rest of you will. :lol:

[Dave Baumann]

Josh - talk to someone at NVIDIA about the impact of the work they did on the XBox.

[JoshMST]

Haha, yeah, I guess I do subscribe to that theory. If you look back at how rocky the relationship between NV and MS was, I think that Sony approached NV very shortly after the release of the X-Box, and whispered a lot of "sweet nothings" into their ear. Otherwise, why would NV take such a hard stance against MS with all of the dealings of the X-Box? Also, due to things like backwards compatibility, MS probably wanted to keep NVIDIA as their primary source of parts. While the X-Box didn't topple the PS2, it was still very successful (especially for being the first foray into consoles for Microsoft).

So my feeling is the NV really weighed their choices between Sony and Microsoft, and they probably looked at the 3:1 advantage that Sony had over MS in sales with the PS2. Just looking at the way things turned out, I am pretty sure that NV didn't really try to get the contract for the X-Box 360. The timeline from the original X-Box to the overall design of the current X-Box 360 really points to Sony approaching NV very early on about creating the graphics portion of the PS3. And if this is the case, I wonder how nervous Sony got when NVIDIA released the NV30 while ATI came out with the R300? I bet there were some executives at Sony really sweating it out if they made the right deal!

[rendezvous]

Now, while the actual deal is not for public consumption, we can expect that Sony paid NVIDIA up front most of the development costs for designing the RSX (just as Microsoft did with the original X-Box GPU/MCP).

Huang has on several (atleast two) occations stated that the deal with SONY on the GPU design and tools are on a "per completement basis" IIRC.

[thatdude90210]

So my feeling is the NV really weighed their choices between Sony and Microsoft, and they probably looked at the 3:1 advantage that Sony had over MS in sales with the PS2. Just looking at the way things turned out, I am pretty sure that NV didn't really try to get the contract for the X-Box 360.
Yeah, Nvidia had to choose between MS or Sony. Afterall, the Rsx used up all of their engineers, all 50 of them.

[Xmas]

Don't forget texture filtering.
sRGB filtering is already supported. But actually, you don't need it for gamma-adjusted AA to work correctly (still, it's a great thing). One issue is getting linearized color values into the shader (filtering), the other is storing non-linear values in the framebuffer (framebuffer writes, blending, downsampling).

[Ailuros]

So my feeling is the NV really weighed their choices between Sony and Microsoft, and they probably looked at the 3:1 advantage that Sony had over MS in sales with the PS2. Just looking at the way things turned out, I am pretty sure that NV didn't really try to get the contract for the X-Box 360.
Yeah, Nvidia had to choose between MS or Sony. Afterall, the Rsx used up all of their engineers, all 50 of them.

That misconception was broken down when it first appeared; even if it would be the case (which I doubt) who exactly says that quantity overrules quality for one and secondly in the past as today at some small IHVs 50 "engineers" (well it also depends what exactly you mean with engineers) is the usual size of a design team.

[Geo]

So my feeling is the NV really weighed their choices between Sony and Microsoft, and they probably looked at the 3:1 advantage that Sony had over MS in sales with the PS2. Just looking at the way things turned out, I am pretty sure that NV didn't really try to get the contract for the X-Box 360.
Yeah, Nvidia had to choose between MS or Sony. Afterall, the Rsx used up all of their engineers, all 50 of them.

That misconception was broken down when it first appeared; even if it would be the case (which I doubt) who exactly says that quantity overrules quality for one and secondly in the past as today at some small IHVs 50 "engineers" (well it also depends what exactly you mean with engineers) is the usual size of a design team.

Certainly these things are a matter of degree, and (personally at least) my reference points are NV's effort levels for Xbox1 and ATI's for Xbox2.

Also, the original context for this threadlet are:


Yeah I'm not too thrilled by the released specs. What exactly have these "hundreds of people" been doing "for several years" to come up with this chip. If G70 is just a slightly tweaked NV40 shouldn't that statement have been applied to the NV40 instead?

and

You missed the point. I never said I wasn't impressed with NV40. But the comments I quoted above were said in reference to the RSX architecture. If, as speculated, RSX is very similar to G70 but G70 is a slightly tweaked NV40 then the "hundreds of people" working during "several years" to develop the part doesn't make sense to me.

So 50 people for some months (maybe a year? Unless you buy Josh's theory that this was in place for a long time before we heard of it) can certainly accomplish some worthwhile things on a value-add front, but --one hopes!-- not nearly what "hundreds" can for several years.

[PeterAce]

I was uner the impression that Nvidia already had 200 engineers working on it's 'next generation technology' for several years before the Sony deal. At which point point Nvidia assigned 50 of their engineers for the 'customizing for communication with CELL'.

[Geo]

We also don't know how many of those 50 are software engineers rather than hardware engineers, given the toolset was by all accounts a major part of the deal.

Which I thought worked out well for NV and finally justified their investment in CG. It will also be interesting to see if they leverage any of that additional CG work back into the pc version (i.e. if they do some additional front-end work to go with the custom-api back-end work).

[Geo]

I was uner the impression that Nvidia already had 200 engineers working on it's 'next generation technology' for several years before the Sony deal. At which point point Nvidia assigned 50 of their engineers for the 'customizing for communication with CELL'.

Well, that's what we're talking about --the rumored specs for G70 (and we're all assuming that's what "next generation technology" is) above/beyond what NV40 is/can do don't seem to do justice to 200 engineers for several years. . .

Edit: Actually, what makes sense to me (assuming those numbers have some basis in reality, which I'm willing to do :) ), and given Wavey's comment upstream about "DX9 9 era", is that is more of a SM3.0-era accounting, and rolls in the NV40 investment as well, rather than a delta between NV40 and G70/RSX. Y'know, the pinnacle of the era for them, which in total cost, etc.

[Ailuros]

Edit: Actually, what makes sense to me (assuming those numbers have some basis in reality, which I'm willing to do :) ), and given Wavey's comment upstream about "DX9 9 era", is that is more of a SM3.0-era accounting, and rolls in the NV40 investment as well, rather than a delta between NV40 and G70/RSX. Y'know, the pinnacle of the era for them, which in total cost, etc.

Nothing against that theory; IMO IHVs might have changed their roadmaps somewhere around 2002 or so when it became clear that Longhorn/WGF2.0 will face another delay. If we'll see WGF2.0 GPUs in 2006, then probably very close to it's end, if not early 2007 after all.

ATI started through the DX9.0 era with SM2.0, moved to SM2.0_extended and is only now about to roll out it's SM3.0 design and of course had to cancel R400 due to the above change probably.

On the other hand NVIDIA started immediately with SM2.0_extended and moved then to SM3.0 with which they're going to stay until WGF2.0 and I'd speculate that whatever they had initially planned for WGF2.0 was further pushed into the future.

I'm not saying that NVIDIA consumed tons of resources for RSX, yet I doubt that a console design (even if it contains a desktop GPU derivative) can be implemented with a twist of your left hand either. If I'm not mistaken there are buswidth differences between RSX and G70.

If you're wiling to see the whole affair from a more spherical POV I very much doubt that NVIDIA needed a bunch of software engineers for RSX, as ATI could have pulled R420/480 off with an equal amount and similar engineers.

One more thing: Josh's article contains also some speculations for the timespan before the NV40/R420 release. While I can believe that clockspeed changes are possible virtually "last minute", the paragraph about supposed 8*2/16*1 for NVIDIA seems out of place; I believe there is a story behind it, but it doesn't mean that NV40 and R420 hadn't been planned 4 quad designs when their specifications were set in stone and it doesn't occur 6 months or a 1 year prior to release either.

[Mulciber]

Im pretty sure he was explaining that the 8*2 rumor was a diversionary tactic by nVidia

[CJ]

G70 up close & personal: http://www.hardwarezone.com/articles/view.php?id=1603&cid=3&pg=1

[trinibwoy]

Can't wait!! :roll:

According to our sources, the card is hot, so hot that we did not managed to even perform a live run. The card is obviously damaged due to heat and upon closer inspection, we saw brown burnt marks around the card, near the PCI Express connector and even at the fan plug. The 10-layer PCB which holds the GPU also suffered heat damage and we saw blistering of the PCB at areas near the GPU.

[neliz]

Nice info...
But where am I going to leave 150Watts of pure heat?

[_xxx_]

Nice info...
But where am I going to leave 150Watts of pure heat?

Feed them into your girlfriend ;)

[nutball]

Nice info...
But where am I going to leave 150Watts of pure heat?

I think I'll be leaving it in the wall socket! :)

[DSC]

The GPU is dated 0505 aka week5 2005. Since February there's G70 silicon, obviously it's not gonna be as good as later batches of silicon as TSMC refines the process for the new chip.

[gsgrunt]

The GPU is dated 0505 aka week5 2005. Since February there's G70 silicon, obviously it's not gonna be as good as later batches of silicon as TSMC refines the process for the new chip.

Does this give any clues as to potential availability? Does it rule out late summer or indicate that late summer could be a strong possibility?

[DegustatoR]

GPU is packaged in Korea -- that is IBM packaging if i'm not mistaken. So this is probably the very first revision of G70 (made at IBM's 130nm?).

[rendezvous]

I think it's difficult to come up with information of when the cards will be availiable from a chip, packaged in february, that according to the source wasn't working too well.

[Chalnoth]

Nice info...
But where am I going to leave 150Watts of pure heat?
150 Watt actual power draw doesn't make any sense. It's doubtful you'd ever get a graphics card to actually work that close to the spec.

[_xxx_]

Nice info...
But where am I going to leave 150Watts of pure heat?
150 Watt actual power draw doesn't make any sense. It's doubtful you'd ever get a graphics card to actually work that close to the spec.

Usually they'll take the peak power draw for such recommendations.

[trinibwoy]

Early Anand coverage - http://www.anandtech.com/tradeshows/showdoc.aspx?i=2431&p=2. Single-slot on G70 so far.

[Chalnoth]

Usually they'll take the peak power draw for such recommendations.
Even peak it's rather doubtful, since you'd need a very exacting distribution of power draw between the different voltage inputs.

[_xxx_]

Oh well, it's the same old story - just exaggerating and making sure people with low-quality PSU's can use the card. So if they say 150W, I'd rather think of 70-80 for real. Whatever, we know what it's about :)

[Chalnoth]

Well, right. A 150 watt number may be a very good one to give to system builders/motherboard manufacturers. But it won't directly translate to heat, since the board would never be capable of actually drawing that amount.

If, and I stress if, this is a number that nVidia is throwing around, then real-world peak power draw might realistically be around 90-110 watts or so.

[karlotta]

to go OT abit, a dual pcix board with two cards running 75w each your 150 is there. then the cpu and lil ram/sound... 200w is a posibliity?

[Evildeus]

So the G70 is a 300+m chip?

[wireframe]

If that is the final reference and proposed retail design, I must say I am somewhat disappointed. It also irks me that people think a single slot solution is a good thing. I don't care about the temperature and heat of the GPUs as long as it is directly exhausted and this requires an extra slot. It seems so logical that a high performance device like a 3D accelerator would warrant such a solution that I am amazed at how Nvidia has gone from a good approach in the 5800 Ultra to this (of course I am assuming that this thing is hot, which it may not be).

Oh well, I'll have to hope for a dual slot cooling, with strong emphasis on an exhaust system, from some card vendor or something more to my liking from the competition.

[DegustatoR]

The only thing that matters is how noisy it is. I wouldn't call FX5800 cooler a good solution ;) If this cooler is small _and_ quiet then it is great. If it is noisy then i'd prefer a bigger, but quiet design.

[wireframe]

The only thing that matters is how noisy it is. I wouldn't call FX5800 cooler a good solution ;) If this cooler is small _and_ quiet then it is great. If it is noisy then i'd prefer a bigger, but quiet design.

I was referring to the direct exhaust, but I should probably have made a note of not desiring the whole Leafblower Effect(tm). It's my understanding that ATI's x850 series aren't terribly noisy and they have direct exhaust, which I am mostly concerned with here (I don't think anybody wants noise, but we have to accept some as a trade-off).

[borntosoul]

i have to second that, i really think nvidias cooling designs arent as good as ati, just look at the cooling of the 850xt----it looks like it was designed by a formula one team. nvidias gpu really need some more work in my opinion.

[trinibwoy]

Has Computex started yet?

[HVZ]

Has Computex started yet?

this is boring, wheres the g70 performance leaks....

[Pete]

But the 5800U's cooler didn't exhaust hot air, it was just a taller version of a regular cooler. It exhausted hot air in the case.

I think Abit's OTES cooler, debuted with the GF4, was the first one for gaming cards that exhausted hot air outside of the case. I believe the Arctic Cooling VGA Silencer was the second, and ATI's X850 cooler was the third.

Actually, I don't think nV has ever made a cooler that dumped hot air outside the case. That's a bit puzzling, considering their leading-edge efforts in most other respects. Why bother making a double-height cooler without taking the tiny (and achingly obvious) extra step of exhausting hot air directly out of the case?

[borntosoul]

on cooling- i think nv are affraid because they copped so much back lash with the 5800.

[wireframe]

But the 5800U's cooler didn't exhaust hot air, it was just a taller version of a regular cooler. It exhausted hot air in the case.

This is what I am talking about: http://techreport.com/reviews/2003q2/geforcefx-5800ultra/index.x?pg=2

[stevem]

But the 5800U's cooler didn't exhaust hot air, it was just a taller version of a regular cooler. It exhausted hot air in the case.
FXFlow had external intake & exhaust venting on the 5800. Abit's OTES had internal intake/external exhaust..

[Unknown Soldier]

We can’t say much about the G70 as we are bound by NDA, but all of the cards at the show are single-slot solutions which is refreshing.

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2431&p=2.

Single slot G70's .. wow. I wonder if that's the reason why hardwarezone (http://www.hardwarezone.com/articles/view.php?id=1603&cid=3&pg=1) said the card they got was shot?

Still single slot. I wonder if ATI might quake a bit if the R520 is dual slot.

US

[CJ]

We can’t say much about the G70 as we are bound by NDA, but all of the cards at the show are single-slot solutions which is refreshing.

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2431&p=2.

Single slot G70's .. wow. I wonder if that's the reason why hardwarezone (http://www.hardwarezone.com/articles/view.php?id=1603&cid=3&pg=1) said the card they got was shot?

Still single slot. I wonder if ATI might quake a bit if the R520 is dual slot.

US

According to the Inq ATi also has a single slot cooling for the R520: http://www.theinq.net/?article=23563 But to me that looks like a dual slot.

And I'd rather have a dual slot that blows hot air out of my case than one (presumably) hot single slot that keeps all the hot air inside my case.

[Pete]

Whoops, I blanked on that one. I thought I Googled to double-check, but obviously not. I was probably thinking of the 5900U (http://techreport.com/etc/2003q2/new-gfx/index.x?pg=1) or 5950U (http://www.beyond3d.com/previews/nvidia/nv38/index.php?p=2) coolers. Sorry.

I guess nV innovates in basically all areas, then, though obviously with varying degrees of success. :) I wonder if different plastic and a larger-diameter blower would've quieted the FX Flow down, as maintaining a separate airflow sounds like a good idea.

[kemosabe]

We can’t say much about the G70 as we are bound by NDA, but all of the cards at the show are single-slot solutions which is refreshing.

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2431&p=2.

Single slot G70's .. wow. I wonder if that's the reason why hardwarezone (http://www.hardwarezone.com/articles/view.php?id=1603&cid=3&pg=1) said the card they got was shot?

Still single slot. I wonder if ATI might quake a bit if the R520 is dual slot.

US

Do you guys still expect R520 to see the light of day? Seems more and more doubtful to me - I say they're pushing for an earlier R580 release in late summer/early fall. ATI could be bluffing, but they're not good poker players and there seems to be a lack of confidence in R520's ability to outperform G70.

[wireframe]

Do you guys still expect R520 to see the light of day? Seems more and more doubtful to me - I say they're pushing for an earlier R580 release in late summer/early fall. ATI could be bluffing, but they're not good poker players and there seems to be a lack of confidence in R520's ability to outperform G70.

I am very puzzled why R520 would be cancelled given a lack of performance in relation to G70. The G70 seems like a very predictable update to NV40. If G70 was a 32 pipeline card with decent frequency I see where ATI may have underestimated Nvidia, but from the looks of things I think the G70 specs are 'conservative' and I see no reason why ATI should not be able to match them unless they have not used their time wisely. However, looking at the launch schedule for CrossFire, it seems that this coincides with the timeframe we would expect new hardware and it doesn't seem likely that ATI would be launching both X850-CrossFire and 'X950-CrossFire' at the same time, so something certainly is amiss.

[Apple740]

it doesn't seem likely that ATI would be launching both X850-CrossFire and 'X950-CrossFire' at the same time

But why not? As if there will be a large market for 2 x X8xx while a new generation is around the corner? I strongly doubt that much people will be running to the shop for a 2 x X8xx purchase these days. Crossfire seems nice, but for this generation cards they are 6 months too late.

[digitalwanderer]

I strongly doubt that much people will be running to the shop for a 2 x X8xx purchase these days.
I strongly doubt that many ever will be, for either companies dual card solution! :?

Seriously, amid all this argument methinks we're all losing focus a bit on the very real fact that ANY dual GPU rig is gonna be expensive....we're talking "Review Edition" equipment here.

I almost seriously think that SLI/Crossfire was never really intended for broad adoption, just for early bragging rights/pissing contest reasons. :?

[reever]

ATI could be bluffing, but they're not good poker players and there seems to be a lack of confidence in R520's ability to outperform G70.

Do you always believe everything you read?

[kemosabe]

Digi, forget SLI/Crossfire and get working on the scoop about what's ailing R520. Baumann's hints have come to a dead stop and MuFu has been missing in action, so it's all up to you now. :wink: :lol:

[kemosabe]

ATI could be bluffing, but they're not good poker players and there seems to be a lack of confidence in R520's ability to outperform G70.

Do you always believe everything you read?

Uhh, where do you see any mention about something I read? Ever try speculating based on your own interpretation of what's happening and what's not?

[Altcon]

I almost seriously think that SLI/Crossfire was never really intended for broad adoption, just for early bragging rights/pissing contest reasons. :?
Of course they're for bragging rights! (it's a piss on OUR confusing wind contest).
Who in their right mind would develop games that require this kind of solution?
Hell, I can't even imagine that 5 years from now it will be a popular/mainstream solution.
Most people are buying SLI boards due to the Nforce 4's many features, If ATI's crossfire boards are likewise equiped it'll be an instant hit.
Oh, and some 3 or 4 trillionares will actually be using dual X8xx's or dual 6800's of some type.
Do you know of anyone who has an SLI setup running dual 6800 GT's or upwards (I'm counting out reviewers who didn'y pay for their rigs)?
I sure don't...

[ANova]

I think you guys are reading too much into the R520's supposed delay.

[Geo]

As if there will be a large market for 2 x X8xx while a new generation is around the corner? I strongly doubt that much people will be running to the shop for a 2 x X8xx purchase these days. Crossfire seems nice, but for this generation cards they are 6 months too late.

Yes, but the rub is this can hardly be news to ATI either, and yet they went forward with it. Hence the sinking feeling some of us are having as to whether there may be "another shoe" yet to drop.

[digitalwanderer]

I think you guys are reading too much into the R520's supposed delay.
http://mastabeta.com/forum/images/smilies/yep.gif

I don't think there is anything sinister or bad behind "the delay", I think it's a strategic business move.

Why should they release it even if it's ready if they're still selling X800 well and they have a lot of them?

I don't like that reasoning either being as I just want the shiny/newest now, but I really think that's all there is behind it.

Blame the suits.

[Apple740]

I strongly doubt that much people will be running to the shop for a 2 x X8xx purchase these days.
I strongly doubt that many ever will be, for either companies dual card solution! :?


I wish i had received a dollar for each sold Sli ready motherboard. :lol:

[Apple740]

Why should they release it even if it's ready if they're still selling X800 well and they have a lot of them?

They need a card that can do FP blending. I know a lot of people
who bought a X800, even recently, and found out that they can't play Farcry and Splinter Cell with HDR because their card doesn't support this. A very dissapointing experience if you just purchased a 500$ card.
Valve's new demo level Lost Coast requires sm3.0/fp blending for max IQ. Heck, Valve holds the release up because their bed-partner hasn't a card which can play it! :roll: The Steam forums are full of Nvidia fanboys laughing their pants off. (the forum is VERY big, many people who doesn't favor Nv nor Ati, and are in the need for a new card see this. "hmmm, i need a Nv card" is most likely what they'll think)

I know, a lot of people/gamers don't even know what HDR is, but each day more and more will. Even at Driver(Ati)heaven there is a thread called "if you want HDR don't buy Ati".

[Moloch]

Why should they release it even if it's ready if they're still selling X800 well and they have a lot of them?

They need a card that can do FP blending. I know a lot of people
who bought a X800, even recently, and found out that they can't play Farcry and Splinter Cell with HDR because their card doesn't support this. A very dissapointing experience if you just purchased a 500$ card.
Valve's new demo level Lost Coast requires sm3.0/fp blending for max IQ. Heck, Valve holds the release up because their bed-partner hasn't a card which can play it! :roll:

I know, a lot of people/gamers don't even know what HDR is, but each day more and more will. Even at Driver(Ati)heaven there is a thread called "if you want HDR don't buy Ati".
Do they realize that hdr for the nv40 is slow?
When you can run at 1600x1200 it will be a worthy options, but untill then it isn't.

[CJ]

it doesn't seem likely that ATI would be launching both X850-CrossFire and 'X950-CrossFire' at the same time

But why not? As if there will be a large market for 2 x X8xx while a new generation is around the corner? I strongly doubt that much people will be running to the shop for a 2 x X8xx purchase these days. Crossfire seems nice, but for this generation cards they are 6 months too late.

They are not late for everyone. There are plenty of people who have a PCIe X800XT, X800XL or X800XT PE who can still easily play the majority of games now without needing SM3.0. Those people have the chance to double performance or get better IQ in a few months and prices are bound to drop after August for those CrossFire cards.

[Chalnoth]

Do they realize that hdr for the nv40 is slow?
When you can run at 1600x1200 it will be a worthy options, but untill then it isn't.
Have you actually played a game in HDR? It really does add an entirely new dimension to the lighting. Even though it has its faults, once I saw HDR lighting in Far Cry, I couldn't stand to play without it. It ran just fine in 1024x768 on my GeForce 6800 (I haven't played it recently, so with my faster CPU I may no longer be able to accept the low framerate, I don't know), so I expect you could do 1280x960 or higher with a GT or Ultra just fine.

In the end, I expect that most people that have actually seen HDR in action won't agree with you that it's not necessary until it's available at the highest resolutions.

[wireframe]

They are not late for everyone. There are plenty of people who have a PCIe X800XT, X800XL or X800XT PE who can still easily play the majority of games now without needing SM3.0. Those people have the chance to double performance or get better IQ in a few months and prices are bound to drop after August for those CrossFire cards.

This makes so very little sense. If people are finding faults in Nvidia's SLI offerings, then ATI's must be met by extreme criticism. You said it yourself, the owners of X800/X50 can easily play most games, but without SM 3.0. So why would they need more power? With the Geforce 6800 it makes much more sense: you get over-powering performance now, but these SLI setups should be able to survive well into the future because they support SM 3.0 and blending. The ATI situation is looking something like if Nvidia would have responded to the Readeon 9700 with Geforce 4 SLI.

Look at it this way. Why would anyone with a X850 want to to seek out a limited SKU X850-Crossfire for $550 *and* a soon-to-be-introduced (read: low initial volumes and growing pains) Crossfire compatible mainboard when they can find an SLI capable mainboard almost anywhere and G70 is looming at the same $550 price point and will offer the option to pair a second one in SLI? It really makes no sense to me, unless the customer has serious objections to using Nvidia hardware.

This should all be taken in referced to the X800/X50 cards and not as a critique of Crossfire. Crossfire is fine, but the implementation in terms of current options is just not logical. Two SM 3.0, blending capable R520s would make sense. You'd have to be a total ATI zealot, extremly curious, or stupid to get two x850 cards when most of the time one is 'enough' and 5,000 of them won't be enough when you want to see that flashy new game using SM 3.0 and HDR effects.

[dan2097]

2 X800XLs is less unreasonable though...

[CJ]

This makes so very little sense. If people are finding faults in Nvidia's SLI offerings, then ATI's must be met by extreme criticism. You said it yourself, the owners of X800/X50 can easily play most games, but without SM 3.0. So why would they need more power?

Not every game that will be released within the next year will use SM3.0. There are plenty of games who will still use 'plain old SM2.0' but still ask a lot more of a graphicscard. CrossFire will come in handy there. And it's not like higher performance is the only thing going for CrossFire. There are some IQ freaks out there who would love to have 14xAA and 32xAF.

[wireframe]

This makes so very little sense. If people are finding faults in Nvidia's SLI offerings, then ATI's must be met by extreme criticism. You said it yourself, the owners of X800/X50 can easily play most games, but without SM 3.0. So why would they need more power?

Not every game that will be released within the next year will use SM3.0. There are plenty of games who will still use 'plain old SM2.0' but still ask a lot more of a graphicscard. CrossFire will come in handy there. And it's not like higher performance is the only thing going for CrossFire. There are some IQ freaks out there who would love to have 14xAA and 32xAF.

But some people are already moaning because games like Splinter Cell: Chaos Theory requires SM 3.0 + blending/HDR. What makes you think this situation will go away? It will just get 'worse' if you ask me and it should. We want better graphics, right? And this requires certain functionality from the hardware. Your argument could have been used when Radeon 9700 was new: But SM 1.x will still be supported... So what? Do you honestly think that people would have been happy playing games like Far Cry on Geforce 4 SLI using SM 1.x when others are playing it in DX 9/SM 2 mode? No way! And guess what, it has even progressed from there so that you can enable a limited version of HDR in Far Cry. What on Earth makes you think this will come to a halt and everyone will suddenly think "you know, I don't want better graphical effects. I just want higher frame rates with the same graphics I have now," when they could have both?

Well, we'll probably never see eye-to-eye on this one because I thought the X800 was somewhat castrated at launch because it did not have the same feature set as the 6800. Then again, I probably wanted Geforce 6800 Ultra x 1.2 and 512MB memory for everything to be alright so maybe I am a bit overly critical.

Just so that you understand, Splinter Cell: Chaos Theory runs fine on a single 6800 Ultra in 1600*1200 using all SM 3.0 effects to the max and HDR. Far Cry with HDR runs fine at 1280*1024. Neither are using FSAA, of course. Sure, there are some areas in both games that drop the frame rate to where you think you would like a bit more power, but it's not to the point where it is unplayable and I think both games have areas that are just troublespots no matter what (especially Far Cry). Furthermore, and probably more importantly, these games look really good with HDR enabled. It's not just a gimmick (although the Far Cry implementation may be a bit in that direction). It really adds something to the gaming experience. Wait 2 years and you will probably refuse to play without it.

[Jawed]

[H] can't get playable framerates in SC:CT at the settings you suggest:

http://www.hardocp.com/article.html?art=NzYwLDc=

With the BFGTech GeForce 6800 Ultra OC, what we found was that, with all these settings at their highest values, we could only play at 1024x768 resolution with 8XAF. When HDR is enabled, Anti-Aliasing is disabled. Therefore, you cannot enable AA with HDR. As such, if you wanted to play with all the graphics features turned on with the BFGTech 6800 Ultra OC, it looks like you will have to play at a relatively low 1024x768 resolution to get acceptable performance.
I don't know what SLI would achieve in this scenario, I don't remember seeing an SLI benchmark for this game. There must be one out there.

But anyway, I agree, X8xx is a dead end features wise, and you have to be a certain kind of enthusiast to want to double it up in the current climate. Some of those enthusiasts are the type that happily change their system every six months, so let's leave em to it, shall we :wink:

Unless G70 can't do SLI, I really think ATI should have announced R520 and Crossfire simultaneously.

Why hasn't NVidia officially announced G70 at Computex?!!

Jawed

[wireframe]

[H] can't get playable framerates in SC:CT at the settings you suggest:

http://www.hardocp.com/article.html?art=NzYwLDc=

With the BFGTech GeForce 6800 Ultra OC, what we found was that, with all these settings at their highest values, we could only play at 1024x768 resolution with 8XAF. When HDR is enabled, Anti-Aliasing is disabled. Therefore, you cannot enable AA with HDR. As such, if you wanted to play with all the graphics features turned on with the BFGTech 6800 Ultra OC, it looks like you will have to play at a relatively low 1024x768 resolution to get acceptable performance.

Well, you are just gonna have to take my word for it. I only use 4x anisotropic filtering (can't remember if this is for some reason other than when I was benchmarking), but it runs just fine at 1600*1200 with all the bells and whistles. I have only played up to and including half the bank level so maybe there is some level beyond that will require dropping the resolution to have it play smoothly. However, keep in mind that this is not a super-fast paced shooter and 30 FPS can look smooth.

Perhaps HardOCP was too busy getting bechmarking numbers that look 'normal' in the 60+ range or I am simply much much better than them :P

[wireframe]

Unless G70 can't do SLI, I really think ATI should have announced R520 and Crossfire simultaneously.

We had some good indication that G70 "7800 GTX" is SLI capable already, I thought, or is that all bogus? I also don't understand why ATI would blow their wad with Crossfire on R4xx when this will just reduce the performance gap between these cards and the next. I thought they would want to give more incentives to buy the R520 to get that uber-uber gaming experience "that no other card on the market can provide."


Why hasn't NVidia officially announced G70 at Computex?!!

Maybe all these 'leaks' are very premature and there is no reason to expect G70 or R520 soon. Maybe Nvidia is playing it conservative, seeing what ATI is doing. I suppose they could go ahead and put the G70 out there and claim all honors (annihilation strategy), but this would also eat into their own NV4x sales. Then again, they may want/need to take some time to refine the G70, using any time afforded them by ATI to up their clocks (I must admit I hesistate in speculating that ATI is completely falling apart. It just doesn't feel right.) It's probably safe to speculate that "this is how milking happens."

[Jawed]

Well we might take the paucity of Xenos (90nm) chips for XB360 demonstration at E3 as an indicator that ATI isn't having such a great time with 90nm.

The rumours about ASE packaging:

http://www.xbitlabs.com/news/video/display/20050530100343.html

would seem to imply that ATI is further with R520 than NVidia is with G70.

Jawed

[Domell]

http://www.bit-tech.net/news/2005/05/31/xfx_g70_running/

Look at This!!!!! G70 is running ....!!!

[Geo]

Maybe all these 'leaks' are very premature and there is no reason to expect G70 or R520 soon. Maybe Nvidia is playing it conservative, seeing what ATI is doing. I suppose they could go ahead and put the G70 out there and claim all honors (annihilation strategy), but this would also eat into their own NV4x sales. Then again, they may want/need to take some time to refine the G70, using any time afforded them by ATI to up their clocks (I must admit I hesistate in speculating that ATI is completely falling apart. It just doesn't feel right.) It's probably safe to speculate that "this is how milking happens."

I go back and forth on this, but mostly what I get from ATI right now is a sense of being out-of-sync --like watching the mouths and hearing the dialogue of an old Godzilla movie. I think this comes from having so many balls in the air (remember, they were going to be stretched by R500 *before* they belatedly decided to jump on the SLI bandwagon). Then Rialto went long, and I just mostly get a sense of the proverbial snowball getting out of hand. Not that the stuff isn't good, just they are so stretched that the pieces of the strategy aren't fitting together as well in time as they'd like.

R520 will be the touchstone on this theory tho. If it is performance/feature competitive or better with G70, then we'll know it was just all timing issues driven by resource constraints playing havoc with the overall strategy fitting together rationally. If it is a disappointment, then it was a mad scramble to realign.

Some think it could all be Longhorn-realignment, but I'm not there yet --we've known late 2006 for Longhorn for some time, yet that ATI roadmap was pretty fresh and not showing any signs of that kind of conscious realignment.

[Apple740]

With the Geforce 6800 it makes much more sense: you get over-powering performance now, but these SLI setups should be able to survive well into the future because they support SM 3.0 and blending.

Except that (current) Sli doesn't help in blending performance over a single card. Interesting to see if this issue is solved with R520-Crossfire and/or 7800GTX- Sli.
And... will we have blending & FSAA together!!?? :D

[Apple740]

http://www.bit-tech.net/news/2005/05/31/xfx_g70_running/

Look at This!!!!! G70 is running ....!!!

As we mentioned in a previous article, NVIDIA will not launch G70 at the show, but we expect it to release a couple of weeks after everything has calmed down a little.

Calmed down? What was all the heat then? Did the cards caught fire? :lol:

[Chalnoth]

Except that (current) Sli doesn't help in blending performance over a single card. Interesting to see if this issue is solved with R520-Crossfire and/or 7800GTX- Sli.
And... will we have blending & FSAA together!!?? :D
Considering that render to texture is going to be more and more common, this is likely something that nVidia is going to be working on. I don't really see any reason why you can't get excellent SLI performance out of your typical HDR scenario with a bit of knowledge of how the rendered texture is used in the tonemapping pass. It'll require a fair amount of effort on nVidia's part, though, and application-specific settings.

[HaLDoL]

http://www.bit-tech.net/news/2005/05/31/xfx_g70_running/

Look at This!!!!! G70 is running ....!!!

As we mentioned in a previous article, NVIDIA will not launch G70 at the show, but we expect it to release a couple of weeks after everything has calmed down a little.

Calmed down? What was all the heat then? Did the cards caught fire? :lol:
How funny from an ATI ..... .

Anand says: NVIDIA plans on sticking it to ATI with G70 and offering widespread availability of their new GPU very soon. Manufacturers at the show have already indicated that the first shipments of G70 boards will be in their hands by the second week of June.
Boards are on display at the show, behind closed doors of course. Of course, in Taiwan nothing is ever secret - and thus we’ve had the ability to play around with a number of G70 cards at the show. We can’t say much about the G70 as we are bound by NDA, but all of the cards at the show are single-slot solutions which is refreshing.
Only a handful of NVIDIA’s closest partners have been given G70 designs to show off at Computex, the rest are told to wait until further notice.

So there are more cards than just 2, the cards are single slot and are currently under NDA. I still have my hopes on a computex launch.

[elroy]

Could be a launch-and-on-the-shelf-next-day sort of launch, which would be refreshing.

[Unknown Soldier]

Could be a launch-and-on-the-shelf-next-day sort of launch, which would be refreshing.

That's what was said ATI were gonna do .. except ... you still have to wait a month for Crossfire and even longer for crossfire graphics cards (X800XL).

I'm seriously starting to get worried about ATI.

US

[DegustatoR]

This shows the strength of XFX's close relationship with NVIDIA, with the graphics giant giving away its only working silicon to one of its board partners.
Well it's definately not the _only_ working silicon if it was given to XFX (who outsource board production to Sparkle if i'm not mistaken and then just slaps their stickers on them) :)

[Broken Hope]

A UK online retailer is claiming availability in 2-3 weeks. I wonder how close to release we really are?

[pharma]

Ooh .... Looks like a G70 Ultra may be lurking around! :D

I get the distinct impression from talking to certain industry insiders at Computex that they're holding back a no-holds-barred Ultra SKU, packed with 512MiB of framebuffer and clocks to shame the GTX, while they wait and see what ATI push out with their R520 graphics processor.


http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD0xMjI3

Pharma

[trinibwoy]

I'm loving this sh!t. :lol: :lol: Nvidia and ATi yanking each others' nuts while we gossip and guess and hope and pray. These companies are amazing man - they have people begging to give them their money :lol:

[compres]

Why are some people so obviously biased? I hope you are at least getting paid by either company, becouse otherwise just it seem like irrational behavior.

[trinibwoy]

One observation we’ve made whilst roaming the show floor is the lack of nVidia G70 samples. G70 is nVidia’s new graphics core, which will be labeled the ‘GeForce 7’ series. (ed. note – we did see G70 this morning with nVidia but now we cannot say much as we are under NDA but nVidia’s usual technology demonstration was pretty fine looking) Interestingly, whilst discussing this with a product manager at ASUS, we learnt that the reason for this has much to do with attempted strong-arm tactics by nVidia. Apparently, they refused to provide reference samples to vendors that were displaying ATI Crossfire solutions, and since Crossfire is arguably the biggest announcement of the show, it’s not surprising that many vendors chose to stick with ATI – hence the lack of G70.

WTF? Maybe Nvidia is EVIL!!! :twisted:

[digitalwanderer]

Why are some people so obviously biased? I hope you are at least getting paid by either company, becouse otherwise just it seem like irrational behavior.
Man has never been a rational creature.

[HVZ]

id beg to differ about crossfire being bigger then nvidias net gen chip. i doubt many people here actually care about crossfire.

[ERK]

One observation we’ve made whilst roaming the show floor is the lack of nVidia G70 samples. G70 is nVidia’s new graphics core, which will be labeled the ‘GeForce 7’ series. (ed. note – we did see G70 this morning with nVidia but now we cannot say much as we are under NDA but nVidia’s usual technology demonstration was pretty fine looking) Interestingly, whilst discussing this with a product manager at ASUS, we learnt that the reason for this has much to do with attempted strong-arm tactics by nVidia. Apparently, they refused to provide reference samples to vendors that were displaying ATI Crossfire solutions, and since Crossfire is arguably the biggest announcement of the show, it’s not surprising that many vendors chose to stick with ATI – hence the lack of G70.

I'm shocked.

[Chalnoth]

That doesn't make much sense to me. nVidia wouldn't be handing out the samples at the show itself.

[Geo]

One observation we’ve made whilst roaming the show floor is the lack of nVidia G70 samples. G70 is nVidia’s new graphics core, which will be labeled the ‘GeForce 7’ series. (ed. note – we did see G70 this morning with nVidia but now we cannot say much as we are under NDA but nVidia’s usual technology demonstration was pretty fine looking) Interestingly, whilst discussing this with a product manager at ASUS, we learnt that the reason for this has much to do with attempted strong-arm tactics by nVidia. Apparently, they refused to provide reference samples to vendors that were displaying ATI Crossfire solutions, and since Crossfire is arguably the biggest announcement of the show, it’s not surprising that many vendors chose to stick with ATI – hence the lack of G70.



Damn, that is very interesting. If one were to take that at face value (I'd call this "anecdotal") the conclusion one would have to reach is that new SLI/Crossfire has more cachet amongst the partners than a new generation release. Whew! Have we gotten to that place? That seems bubblish to me (which assumes the bubble will burst).

[DegustatoR]

Nothing new here. Exclusive NV board makers (Leadtek, Sparkle etc.) were getting the first shipments of NV4x last year too. So this is most probably going to be the same with G70 and all the following cores. It's business :)

[Chalnoth]

Nothing new here. Exclusive NV board makers (Leadtek, Sparkle etc.) were getting the first shipments of NV4x last year too. So this is most probably going to be the same with G70 and all the following cores. It's business :)
Now that makes a bit more sense.

[compres]

I dont think its good bussiness for them though, if I had a bussiness I would try to grab as much resellers as possible.

Maybe I am wrong, how much do you think this strategy helped nVidia on the nV40 release? And does ATI have the same strategie on exclusive resellers like saffire?

[Altcon]

I dont think its good bussiness for them though, if I had a bussiness I would try to grab as much resellers as possible.

Maybe I am wrong, how much do you think this strategy helped nVidia on the nV40 release? And does ATI have the same strategie on exclusive resellers like saffire?
I think it helps them KEEP some partners exclusive. It also boosted sales for those partners since NV40 was a hit. It's good business to keep loyal partners happy and not straying into enemy territory.
Here increased sales of companies like Albatron (who've never been heard of here prior to NV40) showed those fruits.
If you thnk about it, ATI does the exact same thing with Sapphire who've always got the first pick of new VPU cores, and hence always ship worldwide first.

[trinibwoy]

It's official - http://www.gamespot.com/news/2005/06/02/news_6126883.html

Nvidia has revealed that it will announce its next-generation GPU on June 21 at a launch event in San Francisco. An invitation sent to Nvidia partners, analysts, and members of the press promises invitees an "evening of full-throttle entertainment" filled with "revolutionary technologies, hot 3D content, and exciting gaming action as we unveil our next-generation GPU."

[Chalnoth]

Can't wait to see those new demos :) I seriously hope that I can run 'em on my 6600 GT.

[DegustatoR]

You'll most probably be able to but the speed would be around 5 fps i guess :)

[Chalnoth]

You'll most probably be able to but the speed would be around 5 fps i guess :)
I doubt it'd be that bad. If they run at all (i.e. without emulation), they likely won't be hugely slower than the NV40 demos, which run pretty well on my 6600 GT.

[Love_In_Rio]

http://photo.gznet.com/photos/1148314/1148314-E23BmoGnIJ.jpg

[max-pain]

http://img248.echo.cx/my.php?image=019di.jpg
http://img16.echo.cx/my.php?image=023yh.jpg
http://img16.echo.cx/my.php?image=032wi.jpg
http://img16.echo.cx/my.php?image=040ro.jpg
http://img16.echo.cx/my.php?image=051ph.jpg
http://img16.echo.cx/my.php?image=066fv.jpg
http://img16.echo.cx/my.php?image=074ji.jpg
http://img16.echo.cx/my.php?image=087iy.jpg
http://img16.echo.cx/my.php?image=093eb.jpg
http://img152.echo.cx/my.php?image=107mj.jpg
http://img152.echo.cx/my.php?image=116ao.jpg
http://img152.echo.cx/my.php?image=122ej.jpg

[Bouncing Zabaglione Bros.]

It's official - http://www.gamespot.com/news/2005/06/02/news_6126883.html

Nvidia has revealed that it will announce its next-generation GPU on June 21 at a launch event in San Francisco. An invitation sent to Nvidia partners, analysts, and members of the press promises invitees an "evening of full-throttle entertainment" filled with "revolutionary technologies, hot 3D content, and exciting gaming action as we unveil our next-generation GPU."

That's the marketing launch - any bets as to when they'll ship or go into general availability? Go on Nvidia, surprise us all and actually ship quantity soon afterwards.

[Unknown Soldier]

GC16 Memory? Isn't that the same speed as the ATI X800XT PE?

That would be good for 560Mhz, maybe 600Mhz?

It's also 0.06 bigger than the GT6800

Hmm.. how many transistors does the GT6800 have again?

US

[rendezvous]

Hmm.. how many transistors does the GT6800 have again?

US

222M

source: Beyond3D chip table (http://www.beyond3d.com/misc/chipcomp/)

[Unknown Soldier]

So how many transistors do you think they can stuff into 0.06mm (length and height)?

US

[no-X]

It's also 0.06 bigger than the GT6800

G70 is much bigger than NV40 - width is almost the same, but height isn't.

[PaulS]

width is almost the same, but height isn't.

That's my story and I'm sticking to it :twisted:

[Unknown Soldier]

Not according to these two pics?

GT6800
http://img152.echo.cx/my.php?image=116ao.jpg

GTX 7800
http://img152.echo.cx/my.php?image=107mj.jpg

Like I said 0.06mm bigger.

It's the GTX core ... and there's been no other mention (officially) of a GTX Ultra. Also I would expect the GTX to be a G70 core.

US

[_xxx_]

That must not really be the size of the core, it's underneath and could as well be bigger or smaller.

[DegustatoR]

So how many transistors do you think they can stuff into 0.06mm (length and height)?
NV40 is 130nm. G70 is 110nm.

[Unknown Soldier]

width is almost the same, but height isn't.

That's my story and I'm sticking to it :twisted:

Ye .. actually the height does look longer. interesting

[IbaneZ]

Why are there 4 cables for the fan? Blue, green, black and yellow. Never seen more than 3 before. Maybe it has a reverse mode in case the GPU gets too cold? :lol:

[fellix]

Haven't read whole thread, but is it possible that Nvidia just increase the number of fragment pipes (24 or 32) but leaving the number of ROP's untouched?!
I think that >16 color/Z writes per cycle, are already out-of-bound for the common 256-bit DRAM interface. :roll:

[no-X]

Not according to these two pics?

GT6800
http://img152.echo.cx/my.php?image=116ao.jpg

GTX 7800
http://img152.echo.cx/my.php?image=107mj.jpg

Like I said 0.06mm bigger.

I mean this:

http://img173.echo.cx/img173/6114/g70nv405ql.jpg

[DegustatoR]

Haven't read whole thread, but is it possible that Nvidia just increase the number of fragment pipes (24 or 32) but leaving the number of ROP's untouched?!
I think that >16 color/Z writes per cycle, are already out-of-bound for the common 256-bit DRAM interface. :roll:
ROPs were changed however their quantity is expected to be the same as in NV40 (16c/32z).

[nAo]

ROPs were changed however their quantity is expected to be the same as in NV40 (16c/32z).
They addressed issues with 64 bits buffers for sure and it would be nice if MSAA is avaliable on FP RT too.
Support for a 32bit FP format as R500 supports it would be even nicer ;)

[Unknown Soldier]

Ye no-X .. saw that .. but only after I posted after getting back with someone on the phone.

I guess I should refresh before posting.

Also as DegustatoR said .. it's on a 0.11nm base now too ... unlike the 0.13 for the GT6800.

Another 75 million transistors or more?

I suppose we can expect 300Million .. but how much more is the question i'm asking. Then again .. they have got SM3.0 already so other changes would be minimal, i'd expect.

US

[digitalwanderer]

That's the marketing launch - any bets as to when they'll ship or go into general availability? Go on Nvidia, surprise us all and actually ship quantity soon afterwards.
They will ship in July and beat ATi to market with this one by about a month, bank on it.

[DegustatoR]

They will ship in July and beat ATi to market with this one by about a month, bank on it.
More than a month i think...

[Jawed]

It seems to me that NVidia is going to do a phased cut-over with G70. All we've heard about so far is replacements for 6800GT/Ultra.

So when 7800GT/GTX hit the streets in July, presumably 6800GT/Ultra will immediately cease to be listed as current SKUs (though there'll be old stock to clear out?).

At the same time lower cards like 6600GT will carry on without any changes.

NVidia's entire line-up will then be 110nm.

How long will NVidia keep on making 6800 Ultra/6800GT (i.e. NV40/45). I'm wondering if NVidia will keep the 130nm lines in production or chop them off as fast as possible. The latter seems most likely to me.

Jawed

[DegustatoR]

It seems to me that NVidia is going to do a phased cut-over with G70. All we've heard about so far is replacements for 6800GT/Ultra.

So when 7800GT/GTX hit the streets in July, presumably 6800GT/Ultra will immediately cease to be listed as current SKUs (though there'll be old stock to clear out?).

At the same time lower cards like 6600GT will carry on without any changes.

NVidia's entire line-up will then be 110nm.

How long will NVidia keep on making 6800 Ultra/6800GT (i.e. NV40/45). I'm wondering if NVidia will keep the 130nm lines in production or chop them off as fast as possible. The latter seems most likely to me.

Jawed
Don't forget about NV48. It is possible that they'll use it to lower prices on the current 6800 line -- 6600 at $179, 6800 at $199, 6800GT at $299, 7800GT at $399 and 7800GTX at $499.

But that's just a wild guess :)

[CJ]

Don't forget about NV48. It is possible that they'll use it to lower prices on the current 6800 line -- 6600 at $179, 6800 at $199, 6800GT at $299, 7800GT at $399 and 7800GTX at $499.

But that's just a wild guess :)

Where does that leave the regular 7800? :wink:

[DegustatoR]

Where does that leave the regular 7800? :wink:
What regular 7800? :) Then again it might be like this: 6600 at $179, 6800 at $199, 6800GT at $299, 7800 at $399, 7800GT at $499 and 7800GTX at $549 ;)

[no-X]

regular 7800 is 7800 STD?

[incurable]

Personally, I wonder what the SRP for their Ultra variant is currently planned to be. $649? $749?

[Xmas]

That must not really be the size of the core, it's underneath and could as well be bigger or smaller.
Underneath what? This is not a heatspreader, it's the flip side of the die.

[DegustatoR]

Personally, I wonder what the SRP for their Ultra variant is currently planned to be. $649? $749?
Is there an Ultra variant planned? 8)

[CJ]

Where does that leave the regular 7800? :wink:
What regular 7800? :)

You'll see soon enough. :wink:

And yes no-X. "regular 7800" is the standard 7800 without any suffix. There will be a regular, a GT and a GTX. Ultra is no where to be found.

[kemosabe]

Can you be as discretely prolific on the R520 side of the equation? :wink:

[incurable]

Personally, I wonder what the SRP for their Ultra variant is currently planned to be. $649? $749?
Is there an Ultra variant planned? 8)
Don't know, but there better be. GTX sounds so ... hmmm ... slow. Who's gonna pay $100 extra over a simple GT for an X? :twisted:

[wireframe]

There are rumors Nvidia is holding back the Ultra until ATI (finally) responds. This makes a lot of sense, seeing as how it really should be ATI launching a new tech "and teaching the Nvidia boys and girls how DX9.0c(+) is done." It's been a bit strange with the previous releases including the Ultra/XT-PE from the start. These SKUs make a lot more sense as refreshes or responses to increase pressure on the competition instead of as an immediately available flavor.

At any rate, if I were Nvidia and I was launching a refresh on a part that is already ahead of the competition, I would not play all my cards simply because I would not have to. I could release the refresh, which exceeds my previous top-of-the-line (Ultra), brings some other refinement, and then wait for ATI to play their cards, using the time to refine my process, stock up on needed ingredients, and comfortably analyze the situation.

Although this Ultra part is already a speculation (is it really though?), I think it might be more fun to speculate on how ATI will play this through. Given that G70 launches before R520, what would you do with R520 once it launches? Would you also play conservative, only just matching or exceeding the G70 and try to force Nvidia to play the Ultra card? Or would you go balls-to-the-wall and sock it to 'em? Of course this depends on what ATI knows R520 capable of process-wise (frequency and yields), but even if ATI knows they could easily top the GT/GTX, they could choose to hold back (god knows they have been milking like an army of farm hands lately. heh)

At any rate, it will be good if Nvidia launches something soon. This cauldron needs some stirring.

Yes: this was reply 666 :twisted:

[DegustatoR]

I don't think we'll see 7800 Ultra but i do suppose that 90nm G70 refresh may come out sooner than later...

[Xmas]

And yes no-X. "regular 7800" is the standard 7800 without any suffix. There will be a regular, a GT and a GTX. Ultra is no where to be found.
Ah, I hate those names without a suffix. I mean, when there's a suffix you use it, but when there is none you still have to add something like "regular", "non-Pro", whatever, to make it clear you mean a specific model. I think that's a bad move marketing-wise, and I thought ATI had it right this time... until they presented the "plain" X800 :?

[EasyRaider]

Why use a suffix at all? They already have 4 digits in the model number. I would suggest something like GeForce 7810, 7820...

Then again, I kind of like the Ultra moniker. It really sounds like a no-holds-barred product.

[reever]

There are rumors Nvidia is holding back the Ultra until ATI (finally) responds. This makes a lot of sense, seeing as how it really should be ATI launching a new tech "and teaching the Nvidia boys and girls how DX9.0c(+) is done."

Um, didn't the *exact* same rumours come out for the NV35 and NV40?

[Dave Baumann]

There are rumors Nvidia is holding back the Ultra until ATI (finally) responds.

Just the images of a single slot cooler would tend to imply they are holding something back to me.

[weeds]

One observation we’ve made whilst roaming the show floor is the lack of nVidia G70 samples. G70 is nVidia’s new graphics core, which will be labeled the ‘GeForce 7’ series. (ed. note – we did see G70 this morning with nVidia but now we cannot say much as we are under NDA but nVidia’s usual technology demonstration was pretty fine looking) Interestingly, whilst discussing this with a product manager at ASUS, we learnt that the reason for this has much to do with attempted strong-arm tactics by nVidia. Apparently, they refused to provide reference samples to vendors that were displaying ATI Crossfire solutions, and since Crossfire is arguably the biggest announcement of the show, it’s not surprising that many vendors chose to stick with ATI – hence the lack of G70.

WTF? Maybe Nvidia is EVIL!!! :twisted:

Or maybe nvidia is smart and not evil,

http://www.beyond3d.com/forum/viewtopic.php?t=23484&postdays=0&postorder=asc&sta rt=220

[CMAN]

Just the images of a single slot cooler would tend to imply they are holding something back to me.

Clever....I never thought of it like that. So from now on, if it doesn't have massive cooling, then there is something faster out there. Gotta love competition.

[Chalnoth]

Well, yeah. We're way past the point where chips can realistically be limited in clockspeeds by their design/processes before they're limited by heat.

[ANova]

Or maybe nvidia is smart and not evil

No, nvidia is evil, but smart too. When have you seen a super villain that was mentally challenged?

[CJ]

Can you be as discretely prolific on the R520 side of the equation? :wink:

You mean like saying a certain vendor has R520 on their roadmap for August (which I already hinted at in another thread)? :wink: Could mean anything... launch end of July, availability in August. Or Launch in August with immediate availability. That same vendor has G7800 and GF7800GT scheduled for Q3, while GF7800GTX is scheduled for June 21/22

[ANova]

Can you be as discretely prolific on the R520 side of the equation? :wink:

You mean like saying a certain vendor has R520 on their roadmap for August (which I already hinted at in another thread)? :wink: Could mean anything... launch end of July, availability in August.

Specifically July 26th

[IbaneZ]

If 7800 GTX actually will be available in a month, do you think ATI will leak info about R520? I bet they want to stop people from buying 7800's.
"Psst! Look what you get if you just wait a month or two" :)

I'm gonna buy just one more video card this year, and i have to make the right choice. 8)

Last year was crazy, and i'm not gonna buy that many 2005.
X800 Pro->6800GT->X800 Pro VIVO->X800XT->X800XT PE->X850XT->X850XT-PE (my currect card).

ATI should send me a set of stake knives or something. :wink:

[Martin Eddy]

If 7800 GTX actually will be available in a month, do you think ATI will leak info about R520? I bet they want to stop people from buying 7800's.
"Psst! Look what you get if you just wait a month or two" :)

I'm gonna buy just one more video card this year, and i have to make the right choice. 8)

Last year was crazy, and i'm not gonna buy that many 2005.
X800 Pro->6800GT->X800 Pro VIVO->X800XT->X800XT PE->X850XT->X850XT-PE (my currect card).

ATI should send me a set of stake knives or something. :wink:

You're crazy. :lol: Who buy's seven video cards in one year?

[_xxx_]

That must not really be the size of the core, it's underneath and could as well be bigger or smaller.
Underneath what? This is not a heatspreader, it's the flip side of the die.

Not always, it's sometimes covered in plastic.

[trinibwoy]

If 7800 GTX actually will be available in a month, do you think ATI will leak info about R520? I bet they want to stop people from buying 7800's.
"Psst! Look what you get if you just wait a month or two" :)

I'm gonna buy just one more video card this year, and i have to make the right choice. 8)

Last year was crazy, and i'm not gonna buy that many 2005.
X800 Pro->6800GT->X800 Pro VIVO->X800XT->X800XT PE->X850XT->X850XT-PE (my currect card).

ATI should send me a set of stake knives or something. :wink:

You're crazy. :lol: Who buy's seven video cards in one year?

Can't all X800Pro VIVO's unlock to 16 pipes and make X800XT speeds? Seems to me like you could've skipped everything after the VIVO - did you see a big difference going from the X800XT to X850XT PE?

[Xmas]

Not always, it's sometimes covered in plastic.
I've yet to see a chip with that power dissipation covered with plastic. Besides, it wouldn't look like that.

[Chalnoth]

Yeah, it wouldn't seem all that intelligent to cover a hot chip with an insulator.

[IbaneZ]

Can't all X800Pro VIVO's unlock to 16 pipes and make X800XT speeds? Seems to me like you could've skipped everything after the VIVO - did you see a big difference going from the X800XT to X850XT PE?

Well, my Vivo barely made it past 500 MHz@16 pipes, so i figured i'd just go for a real XT instead.
As for the X850XT to XT-PE switch, i was really into benching 3dmark a few months back, so i wanted the best.
And no, i don't notice any difference in games at all. But the PE OC's like a beast. :)

[DarN]

Last year was crazy, and i'm not gonna buy that many 2005.
X800 Pro->6800GT->X800 Pro VIVO->X800XT->X800XT PE->X850XT->X850XT-PE (my currect card).

ATI should send me a set of stake knives or something. :wink:
Someone's making way too much money! :shock:

Yeah, it wouldn't seem all that intelligent to cover a hot chip with an insulator.
LOL. Maybe it's a reverse psychology ploy.

[kemosabe]

Can you be as discretely prolific on the R520 side of the equation? :wink:

You mean like saying a certain vendor has R520 on their roadmap for August (which I already hinted at in another thread)? :wink: Could mean anything... launch end of July, availability in August. Or Launch in August with immediate availability. That same vendor has G7800 and GF7800GT scheduled for Q3, while GF7800GTX is scheduled for June 21/22

Well if the GTX is shipping by the end of June, a one-month head start would be enough to put a large dent in X850XT/PE sales, so ATI certainly has something to worry about even if R520 is very competitive. Looking out a bit further, though, a more important objective would be the mid-range segment and especially the mobile market where the 0.09u advantage will likely be far more tangible if ATI can deliver for the fall refresh. Usually their mobile parts launch shortly after the desktop boards.

[digitalwanderer]

I don't think nVidia will get a months jump on availability, but I think they will get a few weeks.

[trinibwoy]

I don't think nVidia will get a months jump on availability, but I think they will get a few weeks.

digi, you really need to start posting some 'reasons' along with all these claims you're making lately. Or are you just blindly shooting from the hip?

[egore]

Is it possible that Nvidias mid range products could be on 90nm and there high end on 110nm. That would give Nvidia the same 90nm advantage as ATI in the mid range segment.

[Pete]

digi's reasons are undoubtedly NDAed. Shotting from the hip, though? He doesn't seem to need a reason to do that to nVidia, does he? ;)

[digitalwanderer]

digi's reasons are undoubtedly NDAed. Shotting from the hip, though? He doesn't seem to need a reason to do that to nVidia, does he? ;)
And if I were aiming for nVidia, I'd shoot a hell of a lot bigger caliber...don't ya think? ;)

[trinibwoy]

digi's reasons are undoubtedly NDAed. Shotting from the hip, though? He doesn't seem to need a reason to do that to nVidia, does he? ;)

Oh, well I didn't think he was privy to NDA info since most (all?) people under NDA don't spout off public opinions regarding performance based on that info without providing the info - I hope that makes sense. And wouldn't he also need insider access to G70 info as well. Maybe digi is the most connected man on the internet!! :shock:

And yeah, he has never needed a reason to pop Nvidia one :)

[IbaneZ]

Hype machine is waking up.

"Prepare to be enlightened" Coming in 11 days...etc etc. :)
Flash at Gainward.de: http://www.gainward.de/new/main.html

And Gibbo at Overclockers.uk speaks again. 22k in 3dmark 05? Even for an SLI setup that seems a bit high. 8)
http://www.nvnews.net/vbulletin/showthread.php?t=51755

[digitalwanderer]

Oh, well I didn't think he was privy to NDA info since most (all?) people under NDA don't spout off public opinions regarding performance based on that info without providing the info
I'm not under any NDAs, that doesn't mean I don't know some stuff covered by NDAs that I just keep to meself.

You're right about people not making stuff like that public knowledge, but it would probably shock you to know how freely people talk about it privately as long as they trust the discretion of who they're talking to.....and I'm pretty good about being discreet and teasing with it anyways. ;)

You wouldn't believe the PMs & e-mails I sometimes get out of the blue, and I LOOOOOOVE it! :D

I hope that makes sense. And wouldn't he also need insider access to G70 info as well.
Yup he would.

[digitalwanderer]

Hype machine is waking up.

"Prepare to be enlightened" Coming in 11 days...etc etc. :)
Flash at Gainward.de: http://www.gainward.de/new/main.html

And Gibbo at Overclockers.uk speaks again. 22k in 3dmark 05? Even for an SLI setup that seems a bit high. 8)
http://www.nvnews.net/vbulletin/showthread.php?t=51755
22k does seem high, if that's true colour me impressed. :shock:

11 days sounds right, but is that for the announcement or availability?

If they're available in 11 days nVidia will probably get a 3-4 week jump on ATi...although ATi will announce about a week after whatever Gainward enlightens us with.

[DegustatoR]

Considering tweaked FPs, highly shader-limited application (3DM05) and 3DM05 GF6800U SLI scores i'd say that this score is indeed possible (though it do look a bit higher than i was expecting).

As for availability -- we'll have to see for ourselfs but it looks like 7800s should be available right after the announcement... Though this info may be wrong :)

And ATI won't announce anything in about a week after 21st of June.

[egore]

22k in 3dmark05? Wouldn't you be system limited to score that high?

[BrynS]

...And ATI won't announce anything in about a week after 21st of June...
I'm thinking Orton may drop a hint of R520 performance in the conference call (June 23rd ?) relative to the then recently released G70 specs and peformance and probably also provide a bit more clarity on the concerns raised by Goldman Sachs et al.

Cheers,


BrynS

[Geo]

The really interesting part of the Inq piece was that the release of technical materials (and samples?) to the press was today.

So let the leaking begin!! :lol:

[trinibwoy]

Well geo, it's begun - http://www.theinquirer.net/?article=23874

Inq says it's going to be called the GFX. TBH the released specs sound a bit underwhelming in light of all the unmitigated hype and 32-pipe rumours surrounding R520 at the moment.

[rashly]

We don’t have any idea why you need 64-bit texture filtering nor blending but we guess developers do.
Gotta love the Inq. :lol:

[trinibwoy]

We don’t have any idea why you need 64-bit texture filtering nor blending but we guess developers do.
Gotta love the Inq. :lol:

I wonder what level of filtering and blending they think we all need :lol:

[allnighter]

We don’t have any idea why you need 64-bit texture filtering nor blending but we guess developers do.
Gotta love the Inq. :lol:
Ahh the good old Fudo. :lol: Not having a clue is is his naturall environment. He's happy.

[no-X]

Well geo, it's begun - http://www.theinquirer.net/?article=23874

Inq says it's going to be called the GFX. TBH the released specs sound a bit underwhelming in light of all the unmitigated hype and 32-pipe rumours surrounding R520 at the moment.

Reversed to GTX :D

http://www.theinquirer.net/?article=23889

[_xxx_]

Not always, it's sometimes covered in plastic.
I've yet to see a chip with that power dissipation covered with plastic. Besides, it wouldn't look like that.

Ok, I didn't mean "plastic" like iMac housing, but rather some special mixtures (with ceramics?). I have no idea about this chip, but I've seen some. This stuff doesn't look much different than the heat spreader on a CPU, whatever the mixture may be.

[Xmas]

Ok, I didn't mean "plastic" like iMac housing, but rather some special mixtures (with ceramics?). I have no idea about this chip, but I've seen some. This stuff doesn't look much different than the heat spreader on a CPU, whatever the mixture may be.
CPU heat spreaders are typically made of metal. Besides, you can see the epoxy underfill around the die here. I'm very certain that what you see here is the bare die.

[Dave Baumann]

xxx, are you thinking of older wirebond packages? These are shrouded in a ceramic casing.

[no-X]

GF 7800GTX 256MB = $639 (?)

http://www.theinquirer.net/?article=23915
http://ibuypower.com/ibp/store/configurator.aspx?mid=136

[ERK]

GF 7800GTX 256MB = $639 (?)

http://www.theinquirer.net/?article=23915
http://ibuypower.com/ibp/store/configurator.aspx?mid=136

I presume that is $639 MORE than the default card in the rig?

EDIT: which is a NV 6200.

[Xmas]

That's still a lot cheaper than the 6800Ultra 512MiB they're offering :lol:

[R300King!]

Seems the option for the 7800 is gone now. wtf? :roll:

[trinibwoy]

Seems the option for the 7800 is gone now. wtf? :roll:

Yeah it vamoosed. If the GTX is available for retail at ~ $700 ($150 > MSRP) within a week after release then it would still be a hell of a launch.

[Geo]

Seems the option for the 7800 is gone now. wtf? :roll:

Yeah, having your indiscretion show up on the front page of the Inq does have a tendency to create that kind of result. :P

[trinibwoy]

http://www.cooltechzone.com/index.php?option=content&task=view&id=1406

Perhaps the most interesting news is that G70 cards will be in stock and ready to ship from various online vendors starting at 6:01a.m. on the very same day (users in countries with earlier time zones will be able to get the cards accordingly).

[Geo]

http://www.cooltechzone.com/index.php?option=content&task=view&id=1406

Perhaps the most interesting news is that G70 cards will be in stock and ready to ship from various online vendors starting at 6:01a.m. on the very same day (users in countries with earlier time zones will be able to get the cards accordingly).

That would be very cool. Let's hope it is true, and that ATI is inspired to "go thou and do likewise". :D

[digitalwanderer]

An announcement/release would be cool! :D

[_xxx_]

xxx, are you thinking of older wirebond packages? These are shrouded in a ceramic casing.

Maybe, kinda like initial Pentiums had if I recall that correctly? I'm not really sure, quite some years since I last looked at that stuff.

[Pete]

GF 7800GTX 256MB = $639 (?)

http://www.theinquirer.net/?article=23915
http://ibuypower.com/ibp/store/configurator.aspx?mid=136According to the AT thread on this, it was $639 Canadian, which means ~$500 plain American dollars. Not too bad for a new card with higher clocks and 50% more pipes than the 6800U, right off the bat.

[trinibwoy]

Yep, and in the same thread they pointed out that iBuyer is US based and the prices are in $US. Not sure where that guy got Canadian from.

[CMAN]

Price inflation due to time and availability? (Such as last summer)

[Pete]

LOL, trin, that's ridiculous, then. Funny that even though availability is rumored from day one, vendors will still gouge. Hopefully they just won't gouge for as long.

[ondaedg]

If I recall correctly, didn't the same thing happen with the 6800s and x800s when the first batches started to hit the net?

[neliz]

Uhm, 3dmark05 score of the 7800GTX: 7700 points:

http://www.hardspell.com/news/showcont.asp?news_id=14305
(wihtout CPU tests)

[IbaneZ]

Well, if the 7800 GTX "only" scores 7700 in 3dmark-05 i dunno if i'm that interested. I actually hope it's a fake. :)

[neliz]

but it's only a GPU benchmark, not incorporating the normal CPU tests etc.

[IbaneZ]

but it's only a GPU benchmark, not incorporating the normal CPU tests etc.

So? The CPU test won't make the final score higher/lower. It's only the 3 game tests that counts. :)

[Evildeus]

So 40% faster than a 6800U? Not too bad.

[Unknown Soldier]

If it only hits 7700 points .. then i'm shocked. If the R520 hits the rumoured 9000+ scores .. then it's gonna walk all over the G70.

If dave thinks that a Dual slot G70 Ultra is still going to be announced sometime .. how much faster than the GTX can it get? Certainly only 200-300 points?

What about the G80? Any rumours on when it's gonna be announced? If Nvidia were able to release the NV40 and NV45 at just about the same time .. is it possible that Nvidia are holding onto the G80 to see what the R520 is like?

US

[Xmas]

If it's legit, that's lower than I expected. But at 600MHz mem clock, I guess the card is heavily bandwidth/output limited, especially with shadow buffer stuff. And maybe the current drivers aren't taking full advantage of the G70 changes.

btw, if it were completely bandwidth limited, increasing the mem clock to 700MHz would result in a score of ~9000 based on this.

[Jawed]

I presume you're basing that on the results of overclocked 6800Ultras?

I'll take your word for it, because I don't feel like getting into 3DMk-speculation bingo.

Jawed

[Unknown Soldier]

The 6800GT and Ultra .. what were the performance differences?(too lazy to look)

50Mhz?

Can we expect only a difference of 50Mhz for the 7800 Ultra over the GTX?

US

[phenix]

If you consider that X850 XT PE scores around 6000 without sweat this 7700 figure doesn't seem all too much. But then again it seems it really is memory bandwith limited.

[dizietsma]

Ati have enjoyed a 10% to 15% advantage over the 6800U in 3dmark05 so

6800U => 5200
X800XT PE => 5800
7800GTX => 7800

Doesn't seem too shabby with not being earth shattering.

[Unknown Soldier]

You forgot X850XT PE > 6400

[trinibwoy]

Ati have enjoyed a 10% to 15% advantage over the 6800U in 3dmark05 so

6800U => 5200
X800XT PE => 5800
7800GTX => 7800

Doesn't seem too shabby with not being earth shattering.

I think your numbers are a bit off. My GT at Ultra scores in the 5500 range on an XP. Most PE scores I've seen are in the 6000-6100 range.

[trinibwoy]

If it only hits 7700 points .. then i'm shocked. If the R520 hits the rumoured 9000+ scores .. then it's gonna walk all over the G70.

If it only hits 7700 not only would it be disappointing but it opens up the market for X850 Crossfire which now becomes a lot more appealing. What ever happened to the rumours of a single G70 being faster than Ultra SLI (which scores 9000+) ?

[Dave Baumann]

Remember the limitations of 3DMark05...

[trinibwoy]

Remember the limitations of 3DMark05...

:evil:

[Unknown Soldier]

Hmm ye . .3DMark05 doesn't support SLI atm?? or something similar like that if I remember correctly.

US

[Rys]

It's fairly vertex fetch limited. So wasted performance waiting for geometry to hit the vertex hardware.

EDIT: and in saying that, give the expected performance of R520 a rethink too, since it also suffers the same fate :!:

[PeterAce]

Does putting 3DMark05 in a higher res like 1600x1200 make the benchmark more pixel fillrate limated (rather than vertex fetch limated like at 1024x768)?

I can't test this as the demo version doesn't let you control anything.

[Pete]

Hexus' and D-L's benches seem to suggest so, as the gap between the two top cards declines slightly (http://www.ixbt.com/video2/images/his-5/m.png) to significantly (http://www.hexus.net/content/reviews/review.php?dXJsX3Jldmlld19JRD0xMjQ2JnVybF9wYWdlPTY =).

Mind the gap!

[nAo]

Umh..I'm hearing G70 might be able to handle AA on transparent primitives via multisampling OR supersampling.

[Demirug]

Umh..I'm hearing G70 might be able to handle AA on transparent primitives via multisampling OR supersampling.

http://www.beyond3d.com/forum/viewtopic.php?t=23866

[Xmas]

So you missed this thread (http://www.beyond3d.com/forum/viewtopic.php?t=23866)? :D

ah, too slow...

[nAo]

Sorry, I've not missed that thread but I forgot it could use supersampling too on transparent primtives :oops:

[russo121]

Something is wrong with this.... IMO it's too slow.... but it can be explained if:
- the clock was downgraded
- this card has a lot of new features (directed to wfg2 - if it makes any sense) and almost kept nv45 performance....

But if that's the real score, I'm sort of disappointed....

[bans3i]

Do you think it is possible that Nvidia also has a 90nm part of the G70, like RSX?

[trinibwoy]

Not necessarily "wrong". If we were looking at a fillrate limited scenario then the increase in clocks and pipes would amount to a score near 9000 on a 24-piped 430Mhz NV40. There are only two possible scenarios to explain this score AFAIC - either each G70 pipe is weaker than an NV40 pipe or 3dmark05 is not fillrate limited at 1024x768 on the G70. We know it's not CPU limited so that leaves bandwidth or vertex limitations. Based on Dave's hint and the ensuing comments I would bet that we will see much better scaling as resolution is increased and we become more fillrate limited.

My real concern is AA performance - it doesn't look like G70 will have much more bandwidth to work with.

If there are any holes in that logic please feel free to point them out!