NVIDIA GF100 & Friends speculation

[ChronoReverse]

It's also notable that the numbers for ATI are on the 10.2 drivers (and they're not hiding that fact).

While Nvidia drivers are sure to come out in the future to improve performance, when the reviewers use 10.3(a) against the release drivers, it's likely the gap will close a bit.

 

[annihilator]

The only significant changes in 10.3 are Hawx and Dirt 2 performance. Also I wonder with which Nvidia drivers these tests were done.

All in all, if we are talking about drivers, Fermi is more likely to gain future performance enhancements - 5870 is architecturally much more similar to 4870 than GTX 480 is to GT200's. Plus Cypress was released some eight months before.

 

[Squilliam]

Good one - but an X-Fi can do some more things, so it'd add value on its own yet again.

BTT: Most people i know are still using regularly connected 2.1 - 5.1 speakers, not routed with several HDMI-extenders through their receivers in their living rooms and back to the room where the PC sits.

It's a great-to-have feature though for living room/HT PCs, but in my personal opinion it is not very useful for high-end gaming cards. It is no more decisive for my buying decision as for example the video decoders capabilities (again: for high-end gaming cards that sit in high-end gaming rigs).

Personally I've noticed a lot of people sticking gaming PCs onto TVs (whether in the main room or specialised gaming rooms/offices) for the ability to use the display as a dual purpose television / monitor. Its quite simply the easiest way to get a >24" monitor for a reasonable price and being able to stream audio is a good bonus. A 1920/1080 resolution can be an advantage for gaming in terms of getting value for money from the GPU.

Personally I've used it in my HD 5870 VapourX and my HD 5750 HTPC, its a very useful feature especially when for whatever reason I want to relocate the beast into the living room.

 

[ZerazaX]

PR slides + small sample size so you can't really take an average of the %'s and say that's what the card performs.

 

[trinibwoy]

That's far more games than most (all?) reviews use.

 

[jimbo75]

What makes me wonder about those slides is the amount of .0's.

Look at how many fps totals end in .0 The fact that there are a few that don't make me think the "benchmarker" added them as an afterthought, ie he thought he better add some 0.7's and 0.9's to add credibility.

But yeah..66% or more of them end in .0 and that makes them all look suspicious. In that last slide 11/14 5870 benches ended in .0 - that doesn't seem very likely to me.

 

[BRiT]

I would have thought with less than a week until all is revealed that Silly Season would be over, but it's not...

Carry on.

 

[thop]

Do you want to figure out the flaw in your theory yourself, jimbo75?

 

[ZerazaX]

That's far more games than most (all?) reviews use.

Doesn't mean that most reviews are good reviews

And again, a lot of those are same games, repeated settings... (and how much are cherry picked)

I mean, look at Batman:AA... I thought AA was disabled (in game) on the Radeons for example!

 

[jimbo75]

Do you want to figure out the flaw in your theory yourself, jimbo75?

Found it, it's all the same games that end in .0;s.

Well it was good while it lasted imo, and believe me I've dreamt up worse theories than that.

 

[Rangers]

Those scores make $499 480 a little more justifiable.

The Crysis Warhead scores are disappointing though. Looks like 5XXX is the winner whenever VRAM isn't a limitation. I always pay special attention to Crysis scores since it's one of the few games that actually needs high end GPU power. Whether you get 100 or 95 FPS in COD and most other games doesn't matter much.

 

[annihilator]

Those scores make $499 480 a little more justifiable.

The Crysis Warhead scores are disappointing though. Looks like 5XXX is the winner whenever VRAM isn't a limitation. I always pay special attention to Crysis scores since it's one of the few games that actually needs high end GPU power. Whether you get 100 or 95 FPS in COD and most other games doesn't matter much.

yeah, Crysis scores look bad unfortunately. Metro 2033 scores look to favor Fermi (%30 difference) but when one is 16 and the other is 23FPS, you can't really talk about a winner there... just a game that needs to be patched.

 

[aaronspink]

yeah, Crysis scores look bad unfortunately. Metro 2033 scores look to favor Fermi (%30 difference) but when one is 16 and the other is 23FPS, you can't really talk about a winner there... just a game that needs to be patched.

be interesting to know what settings they used in 2033. If they used VH then we're about a generation out from any card really being able to handle those settings.

 

[nathan1988]

does anyone know whan its going to be released ?

 

[digitalwanderer]

Rumor has it as Friday around 7pm, somewhere time

 

[Dave Baumann]

All in all, if we are talking about drivers, Fermi is more likely to gain future performance enhancements - 5870 is architecturally much more similar to 4870 than GTX 480 is to GT200's. Plus Cypress was released some eight months before.

Cypress has DX11 that hasn't been seen on any previous product, so there are new workloads being thrown at the architecture and hence there are new avenues to explore. There will continue to be improvements there, irrespective of anyones percieved notion of architectural similarities; much the same as there were continual performance improvements for RV7xx.

 

[Ailuros]

Cypress has DX11 that hasn't been seen on any previous product, so there are new workloads being thrown at the architecture and hence there are new avenues to explore. There will continue to be improvements there, irrespective of anyones percieved notion of architectural similarities; much the same as there were continual performance improvements for RV7xx.

AMD is beyond doubt ahead in driver maturity for its DX11 products as it stands. I would even suggest that the driver portion that addresses the Evergreen family has been coded nearly if not completely from scratch. And no that definitely doesn't mean that AMD will sit back and enjoy its laurels

 

[KimB]

that one smells of nvidia marketing i'm not trusting that since there's no way any reviewer's gotten 2 480s and a 470 by now

Well, it does say "NVIDIA confidential" right across the slide

 

[KimB]

What makes me wonder about those slides is the amount of .0's.

Look at how many fps totals end in .0 The fact that there are a few that don't make me think the "benchmarker" added them as an afterthought, ie he thought he better add some 0.7's and 0.9's to add credibility.

But yeah..66% or more of them end in .0 and that makes them all look suspicious. In that last slide 11/14 5870 benches ended in .0 - that doesn't seem very likely to me.

Many game benchmarks don't have decimal places. Note that the scores that are consistently '.0' are from the same games.

 

[jimmyjames123]

As I said: We're considering it right now. But a, no two - we added 4870 X2 too - bad decisions from half a year back don't necessarily mean, we have to keep with that until all time ends. After all, we've quit benchmarking Quake 3 also without being called names.

You guys are on the right track in my opinion. If one insists on comparing a $499 GTX 480 to a $599 HD 5970 (the latter which costs 20% more, and is subject to certain limitations inherent in a multi-GPU setup), then why not compare a $599 HD 5970 to a $698 GTX 470 SLI (the latter which costs only 16% more, and is also subject to certain limitations inherent in a multi-GPU setup).

What NVIDIA should do to combat ATI/AMD's high end dual GPU card is to position the GTX 470 SLI against the HD 5970. Hell, their partners could even offer GTX 470 (or 480) SLI package deals. Sure, power consumption will be significantly higher with GTX 470 SLI, but even more importantly in this segment, the performance will be significantly higher with GTX 470 SLI.

One thing is clear to me though: Make no mistake, ATI/AMD's next major GPU [strike]refresh[/strike] revision will have a significantly larger die size than RV770 and RV870. So, ironically, I believe that ATI/AMD is moving in the direction of increasing die size on their high end GPU. In turn, I do believe that NVIDIA's performance/area gap will get smaller compared to ATI/AMD.