We've all heard the rumours about Dirt 2 benchmarks being altered to favour nVidia's newest cards, but I wasn't sure if I really believed it. Dirt 2 was pretty much the only game that Bit-Tech found that Fermi really had a huge performance lead over the 5 series. So I was intrigued when Hexus brought out their 470 GTX review this morning, and they had changed their Dirt 2 benchmarks to separate DX9 / DX 11 tests. DX9: Fermi is conclusively faster in DX9 mode. DX11: Fermi is no faster in DX11 than the 5-series.
I don't get it... From those graphs it's pretty clear the 480 is quite a bit faster than the 5870 up to 2560, where there's very little in it.
I suppose you could say that it is still comfortably faster than the 5870 but it is surprising given that fermi was meant to be better future proofed...
The difference isnt as great as many hoped, and in its current state the fermi cards arent a smart buy. even less so when you factor in the heat and power issues, driver updates will help performance but its always going to be hotter than its competition Realistically speaking the only way fermi could have won was if it matched/bettered the 5970 in all tests.
^ Or GTX 470 @ £180-200 and GTX 480 @ £275-300 Then they would be worth while... they are good preforming DX11 cards. Just so damn expensive and run very hot.
At those prices it would be a difficult choice really - would the extra power and heat be worth a small performance gain? I'm not sure it would really.
there is no chance of nvidia being able to offer these prices - not if they want to make any money anyway.
What id like to know is how nvidia managed to justify the power and heat issues when they were testing the card themselves, I know companies make their own benchmarks look amazing but surely somebody at nvidia must have thought "hang on a minute, its using a lot more power than i thought, and its a little on the toasty side"
at those prices, ATI would also drop their prices. that's the hype talking, and i think it's the hype that killed Fermi. it's actually a very good product, if you look at the their tessellation performance.
They aren't that efficient at tesselation, it seems (although there's only negligable amounts in Dirt2, I think there's flags and puddles that are affected), looking at the difference in fps. It drops 30-40fps in dx11, from dx9. It could be down to the drivers that have only just bolted from the stable doors (one would hope, anyway). We all know how 'good' drivers can be in their early stages.. (think when Vista first came out!) I know it's fantasising, but a more efficient, cheaper 470 with better drivers would have been perfect. As they are, the 470 and 480 are a huge disappointment for me.
Except the tessellation performance they demonstrate in the Heaven benchmark uses up the majority of the processing power, leaving next to nothing for actual environment rendering. I'm interested in seeing benchmarks in games that actually make good use of tessellation, because it won't be as good as the benchmarks make it look.
Did some indepth retesting on monday morning - there's nowt wrong with our Dirt 2 numbers. We force 16xAF in the driver, which hits the ATI card's performance much more than the GTX 480. Blog post incoming with some figures!
Theoretical benchmarks and games are two different things. I doubt we'll see that many games use extensive tesselation because consoles don't feature it. Dirt2 is an exception because ATI plowed a lot of money into it but compared to say, Mass Effect 2 or BFBC2, Dirt2 sales are much, much less. Tesselation is like PhysX or whatever else in some regards, it's a layered feature we might or might not see.
A 480 batter a duel gpu chip some of you really are stupid If the 5970 was beaten by the 480 ati would be in huge trouble As it is the 5970 is still a waste of money period what is love to see is 5870CF 5850CF 470sli 480sli that would be fun
I want to see that too rolo. Hardocp tested the 480SLI which was faster than the HD5970 and concluded the SLI settup was "better". Surely if they were being fair they should of tested the HD4870CF too. After all, The HD5970 isn't quite two 5870's strapped together. Those benchmarks posted are useless. Wheres the mins? No point having a high average if its dropping below 30fps at points. Id far rather have a smooth card rather than one that goes between two very large ranges.
For me its the noise!! How did they allow that through testing? They have some videos on HardOCP and the 480's in SLI sound like a jet fighter about to tak off, unless thats what Nvidia was aiming for...
