The GT300/Fermi Thread

[mao5]

Hmmm...


http://forum.beyond3d.com/showpost.p...&postcount=524

Edit- On a side note: This guy manages to mention PhysX when talking about tessellation in the Uningine...
http://www.youtube.com/watch?v=gkI-ThRTrPY&NR=1

shame on NVIDIA

[Sam_oslo]

The PDF actually says:

"The Hair demo (left) uses tessellation, geometry shaders, and physical simulations."
P.15

I still don't know what technology is behind this hair, but do you think nVidia would mess up with own terms, and call the PhysX for "physical simulations"?

[neliz]

Here's the Siggraph 2008 presentation where the HAIR demo comes from: http://developer.nvidia.com/object/s...2008-hair.html

(the second slide deck.)

[Sam_oslo]

Here's the Siggraph 2008 presentation where the HAIR demo comes from: http://developer.nvidia.com/object/s...2008-hair.html

(the second slide deck.)

Do me a favor, go to that page, click on .. Find on this page .. and type PhysX. How many match doe you get? nVidia would never let PhysX to fall off, if it has been used in that coo l demo.

[zalbard]

@ neliz
Thanks!

[Solus Corvus]

You can see that our "simulated" results for the HD 5870 are actually within a few FPS of AT's. So, vindication right there.

If AT's numbers at 1920x1200 are a vindication then what explains the huge discrepancy at 2560x1600?

[SKYMTL]

I think the most impressive part about the Hair demo is that it was done TWO YEARS ago.

[JohnZS]

Interesting. So this means GF100 is extremely good at handling Tessellation?

Ironically I never even saw that coming.
I predicted that ATi would be better at the tessellation and nVidia would be better at the computational stuff and the FSAA + AF.
Big surprise to see they are also good at the tessellation especially considering ATi have been meddling with tessellation for years now.
John

[Sam_oslo]

I think it is too early to get into the details about performance +/- few FPS. This GF100 looks really good based on early results, (unless nVidia is really bluffing, but i don't think so).

We have to wait for some official test, and several games. otherwise we are going to hang on hairs, LOL.

[SKYMTL]

If AT's numbers at 1920x1200 are a vindication then what explains the huge discrepancy at 2560x1600?

That has me scratching my head but more regarding AT results than mine. I've never seen a case in FC2 where cards would loose so little in terms of framerates when going form 1920 to 2560 at 4xAA. Even at Very High settings as seen here.

[Solus Corvus]

That has me scratching my head but more regarding AT results than mine. I've never seen a case in FC2 where cards would loose so little in terms of framerates when going form 1920 to 2560 at 4xAA. Even at Very High settings as seen here.

I see what you mean and maybe it has something to do with the particular bench run, some system detail, or poor methodology.

What I mean is that if you are calling AT's 2560x1600 numbers into question then you shouldn't claim the 1920x1200 numbers as vindication. Either use both the numbers or don't use either - accepting one number and denying the other because of personal expectations is cherry picking.

[Dainas]

let's say GF100 IS A LOT FASTER THAN HD5970!! BIG DEAL!!! we'll now play games with 250 fps instead of 200 FPS!!!!

Oh believe me there are games out there that need the power these cards bring. 4 year old oblivion with the better cities mod can bring the geometry performance of todays cards to their knees. Poly performance is in big need of a massive boost, all our performance gains in the last 6 years are pretty much just shader and AA/Anis.

[Solus Corvus]

Shattered Horizon needs the power of these new cards.

[kadozer]

Arma2 and the latest stalker will.

[Biker]

More Fermi coverage over @ PCPER...

Pretty similar in terms of format and content to Hardwarecanucks but hey.... we like moar coverage, don't we?

[tdream]

Hopefully this means Tessellation will be used to the max and be the main differentiation between DX10 and DX11 games. At the moment you can't even tell them apart.

[NapalmV5]

^ moar coverage ??

here you go.. 53+ news articles so far

http://news.google.com/news/more?um=...2-_QHpGbjZ8XdM

[Biker]

^ moar coverage ??

here you go.. 53+ news articles so far

http://news.google.com/news/more?um=...2-_QHpGbjZ8XdM

MOAAAAARRR!!!

[BenchZowner]

That has me scratching my head but more regarding AT results than mine. I've never seen a case in FC2 where cards would loose so little in terms of framerates when going form 1920 to 2560 at 4xAA. Even at Very High settings as seen here.

I'll try to steal some time off some other things that I have to do and run some benchmarks tomorrow with a 5870 for starters.
FWIW your numbers look ok ( and nice work on the fermi coverage dude )

[Sam_oslo]

I don't think those 50-60% over HD5870 is going to be the average and typical result for most games.

All nVidia needs to do it to get 20% ahead of HD5870 with a dissent power composition. Even if they could get much more out of the GF100, they would probably limit the performance of the retail GPU just about 20% over HD5870, I would think.

[BenchZowner]

I don't think those 50-60% over HD5870 is going to be the average and typical result for most games.

All nVidia needs to do it to get 20% ahead of HD5870 with a dissent power composition. Even if they could get much more out of the GF100, they would probably limit the performance of the retail GPU just about 20% over HD5870, I would think.

Two takes here.
1) What you just said
2) GTX 360 = 10-20% better than the 5870, just a tad higher price than the 5870 & GTX 380 close to the 5970 with a bit better ( lower obviously ) price.

[Sam_oslo]

Two takes here.
1) What you just said
2) GTX 360 = 10-20% better than the 5870, just a tad higher price than the 5870 & GTX 380 close to the 5970 with a bit better ( lower obviously ) price.

+1 ,
I agree the same rules applies to HD5970 too

EDIT: on this note, nVidia don't need to beat the performance of the HD5970, just matching it with a single GPU would be enough.
But Im not so very sure about the pricing .. i hope you are right. If I'm not mistaking, nVidia has traditionally priced such a "superior" single GPU a good deal over the HD5970.

I believe, the prices would depend more on ATi and how they manage to compete. I hope ATi drops a refresh, or even new GPU at the same time, then we can get really good price on GF100.

[SKYMTL]

What I mean is that if you are calling AT's 2560x1600 numbers into question then you shouldn't claim the 1920x1200 numbers as vindication. Either use both the numbers or don't use either - accepting one number and denying the other because of personal expectations is cherry picking.

Sure I can. IMO, there is something wrong with their 2560 result. Why? Take this into consideration:

1680 -> 1920 = 540,000 pixels difference = 7.1 FPS drop

1920 -> 2560 = 1,792,000 pixels difference = 3.4 FPS drop


Their 1920 x 1200 results look good.

[NapalmV5]

MOAAAAARRR!!!

lol now 56 news articles

GTX 360 = 10-20% better than the 5870, just a tad higher price than the 5870 & GTX 380 close to the 5970 with a bit better ( lower obviously ) price.

but benchDowner wouldnt fermi be out in retail already ??

[SKYMTL]

I don't think those 50-60% over HD5870 is going to be the average and typical result for most games.

All nVidia needs to do it to get 20% ahead of HD5870 with a dissent power composition. Even if they could get much more out of the GF100, they would probably limit the performance of the retail GPU just about 20% over HD5870, I would think.

Benchmarks change as do which games benefit certain manufacturers. However, FC2 at Ultra High details remains one of the most demanding games in any category.

On the flip side of that coin, this was a PR event by NVIDIA which means they showed what they wanted to show and no more. I don't think average performance will be close to what was shown but then again, considering drivers can mature and clock speeds will increase, you never know...