NVIDIA's GeForce GTX 580: Fermi Refined
by Ryan Smith on November 9, 2010 9:00 AM ESTCrysis: Warhead
Kicking things off as always is Crysis: Warhead, still one of the toughest game in our benchmark suite. Even 2 years since the release of the original Crysis, “but can it run Crysis?” is still an important question, and the answer continues to be “no.” One of these years we’ll actually be able to run it with full Enthusiast settings…
Right off the bat we see the GTX 580 do well, which as the successor to what was already the fastest single-GPU card on the market is nothing less than we expect. At 2560 it’s around 16% faster than the GTX 480, and at 1920 that drops to 12%. Bear in mind that the theoretical performance improvement for clock + shader is 17%, so in reality it would be nearly impossible get that close without the architectural improvements also playing a role.
Meanwhile AMD’s double-GPU double-trouble lineup of the 5970 and 6870CF both outscore the GTX 580 by around 12% and 27% respectively. It shouldn’t come as a shock that they’re going to win most tests – ultimately they’re priced much more competitively than the GTX 580, making them price-practical alternatives to the GTX 580.
And speaking of competition the GTX 470 SLI is in much the same boat, handily surpassing the GTX 580. This will come full circle however when we look at power consumption.
Meanwhile looking at minimum framerates we have a different story. AMD’s memory management in CrossFire mode has long been an issue with Crysis at 2560, and it continues to show here with a minimum framerate that simply craters. At 2560 there’s a world of difference between NVIDIA and AMD here, and it’s all in NVIDIA’s favor. 1920 however roughly mirrors our earlier averages, with the 580 taking a decent lead over the GTX 480, but falling to multi-GPU cards.
Facebook
Twitter
RSS
160 Comments
View All Comments
AnnonymousCoward - Wednesday, November 10, 2010 - link
I'm with you, that AMD still has a superior performance per power design. But with the 580, nvidia took Fermi from being outrageous to competitive in that category, and even wins by a wide margin with idle power. Looking at the charts, the 580 also has a vastly superior cooling system to the 5970. Mad props to nvidia for turning things around.FragKrag - Tuesday, November 9, 2010 - link
Still no SC2? :(Ryan Smith - Tuesday, November 9, 2010 - link
Honestly, I ran out of time. I need to do a massive round of SC2 benchmarking this week, at which time it will be in all regular reviews and will be in Bench.ph3412b07 - Tuesday, November 9, 2010 - link
There is always some debate as to the value of single gpu solutions vs multi gpu. I've noticed that the avg/max framerate in multi gpu setups is in fact quite good in some cases, but the min fps paints a different picture, with nearly all setups and various games being plagued by micro-stutter. Has anybody else come across this as reason to go with a more expensive single card?eXces - Tuesday, November 9, 2010 - link
Why did u not include some overclocked 5970? Like u did with GTX 460 when u reviewed 6800 series?Ryan Smith - Wednesday, November 10, 2010 - link
If you don't recall from our 5970 review, we disqualified our 5970 when running at 5870 clocks. The VRMs on the 5970 cannot keep up with the power draw on some real world applications, so it does not pass our muster at those speeds by even the loosest interpretation.529th - Tuesday, November 9, 2010 - link
I knew OCCT was a culprit of causing problems.Ph0b0s - Tuesday, November 9, 2010 - link
Was very interested to look at the review today to see how the new GTX580 and other DX11 card options are in comparison to my GTX 285 SLI setup. But unfortunately for the games I am playing BFBC2, Stalker etc and would base my descition on, I still don't know as my card is not represented. I know why,, becuase they are DX11 games and my card is DX10, but my card still runs them and I would want to know how they compare even if one is running DX10 and the other running DX11. Even Anandtech's chart system gives no measure for my cards in these games . Please sort this out. Just becuase a card does not run the latest version of directx does not mean it should be forgotten. Escpecially since the people most likley to be looking at upgrading are those with this generation of card rather than people with DX 11 hardware...mapesdhs - Wednesday, November 10, 2010 - link
Don't worry, I'll have some useful info for you soon! 8800GT vs. 4890 vs. 460, in all
three cases testing 1 & 2 cards. You should be able to eaisly extrapolate from the
results to your GTX285 vs. 580 scenario. Send me an email (mapesdhs@yahoo.com)
and I'll drop you a line when the results are up. Data for 8800 GT vs. 4890 is already up:
http://www.sgidepot.co.uk/misc/pctests.html
http://www.sgidepot.co.uk/misc/stalkercopbench.txt
but I'm adding two more tests (Unigine and X3TC).
Ian.
juampavalverde - Tuesday, November 9, 2010 - link
NVIDIA exceed AMD with this... as long as the barts should have been 6770, this fermi slight improvement just in this universe can be called 5xx series. it is just the gf100 done right, and should have been named properly, as gtx 490.