ATI launch drivers vs Catalyst 10.6's, performance in detail.

[happy medium]

http://www.hardocp.com/article/2010/07/01/amd_ati_catalyst_driver_performance_comparison/2

quote from conclusion.

" We always hear this common phrase: "give it some time, and newer drivers will improve performance." Well, here we are nine months later and while performance has generally improved, it isn’t as earth shattering as one would hope though. This is good news in one way, as the driver and performance we showed at launch is still relevant today in some games and it seems AMD came out of the gate swinging with a really solid driver to begin with. On the other hand, it is also kind of disappointing not to see some drastic improvements by now. Of course, there are some improvements, but none allow us to change our gameplay settings or experience compared to the launch drivers."

I'll say it again "Shot there load" j/k guys.

 

[Scali]

I suppose this is partly because the 5000-series is based on the 4000-series architecture. So they probably already had pretty mature drivers at launch time.

I recall the early days of my GeForce 8800GTS, with Vista x64 and Crysis in DX10.
Everything was new, the hardware, the drivers, the API, the OS, the game...
Initially I could only play it in DX9... DX10 was very slow, even at low settings.
After a lot of updates of drivers, OS and the game, eventually DX10 became quite playable. I went from single-digit framerates to 20+ fps nearly everywhere.

 

[ModestGamer]

I updated to 10.6 recently on my card. It certainly improved performance a bit but nothing earth shattering.

 

[Genx87]

I'd be interested in seeing how Fermi stacks up to the 5000 series 3 and 6 months after launch with newer drivers.

 

[tincart]

I'd be interested in seeing how Fermi stacks up to the 5000 series 3 and 6 months after launch with newer drivers.

Agreed. Given the gains to Fermi in the last major "performance" release, it seems that nVidia will have a good showing in such a comparison.

 

[happy medium]

I updated to 10.6 recently on my card. It certainly improved performance a bit but nothing earth shattering.

I seen improvements with the 10.3 drivers, that I remember, since then nothing to write home about.

 

[ModestGamer]

I'd be interested in seeing how Fermi stacks up to the 5000 series 3 and 6 months after launch with newer drivers.

Fermi is a great card in terms of performance. but it really needs to come with a LWHP reactor.

 

[Genx87]

Fermi is a great card in terms of performance. but it really needs to come with a LWHP reactor.

I have a 470 and it is quieter and my case is cooler than the 4850 it replaced.

/shrug

 

[ModestGamer]

I have a 470 and it is quieter and my case is cooler than the 4850 it replaced.

/shrug

Oh the last generation GPU's all ate power like hogs. They are all bad like that.

 

[thilanliyan]

I have a 470 and it is quieter and my case is cooler than the 4850 it replaced.

/shrug

Could that be because the 4850 was single slot and didn't exhaust air out of the case?

 

[Grooveriding]

I have a 470 and it is quieter and my case is cooler than the 4850 it replaced.

/shrug

I got two great deals recently in a sale and purchase of new cards, cost me $80 to go from 5870 CF to 480 SLI.

Never owned a 4850, but the 480s are really, really noticeably hotter and much louder than my 5870s were. I don't care about heat, noise and power consumption personally, but fermi cards are clearly loud and really hot.

 

[Genx87]

Could that be because the 4850 was single slot and didn't exhaust air out of the case?

Yes, that is the reason. But what would I care if the card runs at 150C if the overall result is a cooler case due to a good cooling solution?

 

[Ares202]

Yes, that is the reason. But what would I care if the card runs at 150C if the overall result is a cooler case due to a good cooling solution?

Its not fair to compare a 4850 to GTX 470, i mean the 4850 launched at $199 therefore cooler costs are kept to a bare minimum

if you wanted an exhaust cooler the 4870 was just fine

 

[Xarick]

um.. isn't this almost always the case with drivers? Some performance improvements, but nothing major. I personally think the new nvidia drivers are likely to be the most increase we will see.

 

[Voo]

The only thing I noticed about the 10.6 drivers (had 10.3 before) is that hardware accelerated flash is horrible on my 4870, thankfully deactivating it fixed that issue - though I hope that's some problem with my flash version or something, because otherwise that thing is unuseable.. well not really interested in trying to fix that issue since my CPU handles it just fine.

 

[Chiropteran]

Thanks for the post, captain obvious.

Isn't this always the case, except when there is some serious problems with the initial driver release?

Should I feel bad because the video card I bought was running close to it's full potential since day one? Gee, I really wish ATI's drivers were worse on release, it would be so much more fun to start with poor performance and see it gradually improve with each driver update.

 

[ShadowOfMyself]

Should I feel bad because the video card I bought was running close to it's full potential since day one? Gee, I really wish ATI's drivers were worse on release, it would be so much more fun to start with poor performance and see it gradually improve with each driver update.

You mean like certain companies which release "special" drivers with a catchy name on the release date of the competitors cards?

 

[Lonyo]

Wow, there's loads of detail.
7 tests, 1 card. 1 lot of settings in most of the games.

Also, did anyone notice the minimums?
17 -> 26 in AvP
16 -> 25 in Splinter Cell
18 -> 23 in BFBC2 second test
And a few fps in the rest of them.
10% performance increase in AvP average, 10% in Dirt 2, 10% in BFBC2 second test.

I'm not going to claim that doing more testing in more games or with more cards would make the results in any way better, I was just laughing at the idea that this testing could be considered in any way detailed.


Also, based on the conclusion posted in the OP, higher minimums doesn't improve gameplay experience. You heard it here first. Higher minimums don't mean anything.

 

[MrK6]

Wow, there's loads of detail.
7 tests, 1 card. 1 lot of settings in most of the games.

Also, did anyone notice the minimums?
17 -> 26 in AvP
16 -> 25 in Splinter Cell
18 -> 23 in BFBC2 second test
And a few fps in the rest of them.
10% performance increase in AvP average, 10% in Dirt 2, 10% in BFBC2 second test.

I'm not going to claim that doing more testing in more games or with more cards would make the results in any way better, I was just laughing at the idea that this testing could be considered in any way detailed.

Excellent points, and that's part of it. However, the other thing to note is who the hell expects game-changing performance out of driver updates? If some driver update drastically increases performance, it's because the driver was screwed up in the first place, not because you unlocked some hidden magic in the card. These companies aren't stupid, they do a good job tweaking their software for their hardware (despite all the bitching that goes on in forums like this). I think 10% increases in performance are pretty good, and it shows they really tweaked the hell out of their architecture. I'm not sure what else [H] or anyone else was expecting.

 

[Bill Brasky]

Excellent points, and that's part of it. However, the other thing to note is who the hell expects game-changing performance out of driver updates? If some driver update drastically increases performance, it's because the driver was screwed up in the first place, not because you unlocked some hidden magic in the card. These companies aren't stupid, they do a good job tweaking their software for their hardware (despite all the bitching that goes on in forums like this). I think 10% increases in performance are pretty good, and it shows they really tweaked the hell out of their architecture. I'm not sure what else [H] or anyone else was expecting.

Your common sense is not welcome here.

 

[BTRY B 529th FA BN]

does it still crash BF BC2?

 

[Genx87]

Excellent points, and that's part of it. However, the other thing to note is who the hell expects game-changing performance out of driver updates? If some driver update drastically increases performance, it's because the driver was screwed up in the first place, not because you unlocked some hidden magic in the card. These companies aren't stupid, they do a good job tweaking their software for their hardware (despite all the bitching that goes on in forums like this). I think 10% increases in performance are pretty good, and it shows they really tweaked the hell out of their architecture. I'm not sure what else [H] or anyone else was expecting.

I wouldnt say the driver was screwed up. Only that optimizations for that particular game hadnt been written or written very well.

 

[GaiaHunter]

I wouldnt say the driver was screwed up. Only that optimizations for that particular game hadnt been written or written very well.

Or the optimizations for the canned benchmark of that game...

Bet someone testing canned would see slightly more than [H].

 

[bryanW1995]

Could that be because the 4850 was single slot and didn't exhaust air out of the case?

yeah, I didn't like my 4850 b/c of the crappy single slot cooler. gtx 260 is much nicer.

 

[edplayer]

so how does a 5770 compare now to a 4870? I recall that the mass sentiment at release was that they were equal sometimes to about maybe 10% slower