do we get gtx480/470 benchmarks today?

[yacoub]

You should expect to see everything (i.e. specs and benchmarks).

Sweet!
I hope to see something from this new line of DX11 GPUs at the $150 price point to replace this stupid HD5770. I am tired of all the little glitches and annoyances with ATI's GPU products/drivers. Wish NVidia hadn't taken forever and a day to get out some DX11 cards.

 

[MrK6]

Chinese review of 5870 VS GTX480

Wow, if so, ouch. Thanks for posting!

 

[Shilohen]

Wow, if so, ouch. Thanks for posting!

Why ouch, these numbers are actually better than I thought they would be and decent for the pricing. Noise and OC potential remains to be seen, but stock FPS is really not as horrible as I was fearing.

 

[Daedalus685]

Wow, if so, ouch. Thanks for posting!

I require context being unable to view images!

 

[Tempered81]

I require context being unable to view images!

haha! okay.

-heaven benchmark, maxed framebuffer, 30" with 8xAA,16xAF gtx480 is 880% faster than 5870.

-Crysis maxed at 1680x1050, gtx480 is 5% slower than a 5870.

-dirt2, avp, farcry2, ut3, gtx480 is 30-40% faster than a 5870 at normal resolutions

-stalker cop, quakewars, gtx480 is 5% faster than 5870 at normal res

-3dmark vantage performance 480 = 5870, extreme 480 is 10% faster than 5870.

480 consumes 130 watts more at load, 45 watts more at idle than 5870, fan runs at 72% temps 90-96C load

 

[1h4x4s3x]

~30% fps over 5870

 

[RussianSensation]

-dirt2, avp, farcry2, ut3, gtx480 is 30-40% faster than a 5870 at normal resolutions

Ya those numbers are a lot better than the BS 5-10% faster claims. I did say many times I expected +30%. It's just odd that Crysis and some other games like STALKER: Call of Pripyat the advantage is only 5-10%. However, 5870 already ran the 4 games you mentioned above well.

I am more interested in BF:BC2, Metro 2033 since they are most intensive games out right now.

 

[Daedalus685]

haha! okay.

-heaven benchmark, maxed framebuffer, 30" with 8xAA,16xAF gtx480 is 880% faster than 5870.

-Crysis maxed at 1680x1050, gtx480 is 5% slower than a 5870.

-dirt2, avp, farcry2, ut3, gtx480 is 30-40% faster than a 5870 at normal resolutions

-3dmark vantage performance 480 = 5870, extreme 480 is 10% faster than 5870.

480 consumes 130 watts more at load, 45 watts more at idle than 5870, fan runs at 72% temps 90-96C load

TY!

lol at 880%, never thought we'd be smashing into the frame buffer cap this soon again. If the performance in the popular titles is true that is a hell of a lot faster than i thought it would be.

 

[MrK6]

I require context being unable to view images!

Will do .

Why ouch, these numbers are actually better than I thought they would be and decent for the pricing. Noise and OC potential remains to be seen, but stock FPS is really not as horrible as I was fearing.

I was kind of figuring 10-15% across the board, so I'm not surprised really, the thing that gets me is the inconsistency across the board. I'm trying not to read into the results too much since we don't know the details of the system or the testing environment (for example, were the Cat 10.3a's used?), but here's my interpretation of the results:
Unigine - not surprised. This benchmark was bought and sold by NVIDIA, and this is just a clear indication of the tesselation power of the GTX4xx's. Not surprised, and you can even see when 5870 runs out of VRAM at 2560x1600 with 8x AA. It's a tesselation benchmark, nothing more.
Neurotone Demo - Never heard of it, although it's interesting that the 5870 is ahead (~10%) of the GTX480.
Dirt 2 - Very interesting, considering it's an "AMD game." That's why I wonder if this is with the 10.3a's or not, which gave a 15%+ performance boost to AMD cards ion Dirt 2. If it is, the GTX480 is clearly faster in this game, but it's interesting that NVIDIA is still having trouble once you crank the AA (loses at 2560x1600 8xAA).
Battleforge - Clear win for the GTX480, across the board, but I could care about the game so maybe I'm biased against this result.
STALKER: CoP - typical 10-15% performance increase. The higher percentage at 2560x1600 4xAA is due to the 5870 running out of VRAM. I can confirm this on both my previous 5870 and my 5850's. The game is a RAM hog, and the 1.5GB on the GTX480 helps it, although neither result is what I would call playable.
3DMarkVantage - Pretty much dead even here, which is odd (being GPGPU based, I would have thought the GTX480 would be much faster in synthetic tests). I could care about Vantage though.
Crysis - the GTX480 actually loses by 2-5% until the 5870 hits a VRAM wall, but at 2560x1600 4xAA neither card gives playable framerates.
Enemy Territory: Quake Wars - The 5870 is overall 5-10% faster, but I don't really care about this game either .
UT3 - The GTX480 is much faster across the board, but IIRC, NVIDIA always dominated this game. Besides, either card gives more than playable framerates.
Power usage - HOLY JESUS the GTX480 uses another 140W at load (326W for the 5870, 466W for the GTX480). Most 5870 users could crank their 5870s well past 1GHz and still use less power, and that could definitely change the playing field.

My $0.02, take it for what it's worth .

 

[sandorski]

My inside sources tell me that Freecell has very impressive Framerates with Fermi!!

 

[jackstar7]

My inside sources tell me that Freecell has very impressive Framerates with Fermi!!

I know that guy. He's just a fanboy!

 

[nitromullet]

Power usage - HOLY JESUS the GTX480 uses another 140W at load (326W for the 5870, 466W for the GTX480). Most 5870 users could crank their 5870s well past 1GHz and still use less power, and that could definitely change the playing field.

nm... I think you're right about this.

 

[Tempered81]

full review:
http://translate.google.com/transla...10/0326/865/000000865774.shtml&sl=zh-CN&tl=en

 

[Udgnim]

if the Chinese review numbers are remotely true, then I'd say Nvidia has done a decent job competitively pricing Fermi against AMD's 58XX line.

although, the Chinese review is using Cat 10.2 instead of Cat 10.3a, which is disappointing. Bad Company 2 numbers are also wacky. GTX 285 > GTX 470, 5850, and 5870

some of the translations are hilarious though. have "Batman: Forrest Gump" instead of "Batman: Arkham Asylum."

 

[Rhezuss]

If those numbers are true, the GTX 470 is very tempting!

 

[Daedalus685]

Very interesting. Nothing in that review about power other than a word or two about how there is no dual GPU card.

One thing that strikes me is how all over the place the numbers are... Is this going to be a few months of where we will really have to pick the GPU that plays our favourite games best?

 

[tviceman]

One benchmark shows the gtx480 losing in crysis, while another shows it winning in Crysis Warhead.

Did texture memory management change between the two titles?

 

[Daedalus685]

One benchmark shows the gtx480 losing in crysis, while another shows it winning in Crysis Warhead.

Did texture memory management change between the two titles?

A few things did change, though I'm not sure what exactly. The translated review looks pretty legit though, likely that no other review is going to show us crysis 1 (as in not warhead) results anyway.

 

[Tempered81]

Bad Company 2 numbers are also wacky. GTX 285 > GTX 470, 5850, and 5870

GTX285 was benched in DX10, while cypress & fermi were benched in DX11. Seems like the higher the DirectX, the larger the performance hit.

 

[Daedalus685]

I really wish they had not done metro with physX enabled...

 

[cbn]

full review:
http://translate.google.com/transla...10/0326/865/000000865774.shtml&sl=zh-CN&tl=en

Thanks for the review. It looks like Fermi is pretty strong in BFBC2, Dirt2, Far Cry2, Street Fighter 4 and Metro 2033 (with Physx turned on).

I would like to have seen the results for 2033 with Physx turned off also.

 

[MrK6]

The GTX470 @$350 looks like it could be competitive; it doesn't lose to the HD5850 much and many times it's nipping on the heels of the 5870 (a couple of benchmarks actually show it as being faster). However, it still has to overclock somewhat well to stay in the game considering the overclocking potential of the 5850 and 5870. The power consumption is also "meh," but I realize that doesn't matter to everyone as long as it still overclocks well. The GTX480 is a joke though, wow.

I was kind of disappointed with the PhysX benchmarks - no crap running the same PhysX effects coded for GPU hardware via software/CPU is going to result in crap FPS. Also, I have no idea how they got their Metro 2033 results - I run the game with "Advanced PhysX" checked and I have no issues whatsoever.

Thanks for throwing these all up Tempered81! :thumbsup:

 

[Piano Man]

We're starting to see preorder prices for the GTX 470/480 at $345/$485 respectively. If these numbers hold up, and there is good OC potential, ATI is going to have to adjust prices.

 

[OCGuy]

Come on Anand, lets see what you got!

 

[tviceman]

The GTX470 @$350 looks like it could be competitive; it doesn't lose to the HD5850 much and many times it's nipping on the heels of the 5870 (a couple of benchmarks actually show it as being faster). However, it still has to overclock somewhat well to stay in the game considering the overclocking potential of the 5850 and 5870. The power consumption is also "meh," but I realize that doesn't matter to everyone as long as it still overclocks well. The GTX480 is a joke though, wow.

I was kind of disappointed with the PhysX benchmarks - no crap running the same PhysX effects coded for GPU hardware via software/CPU is going to result in crap FPS. Also, I have no idea how they got their Metro 2033 results - I run the game with "Advanced PhysX" checked and I have no issues whatsoever.

Thanks for throwing these all up Tempered81! :thumbsup:

Speaking of Metro2033, since you own the game can you say which has a more noticeable impact on image quality: DOF or tessellation?