HD 4870X2 QuadFire vs GTX 295 Quad SLI Compared

[Paxi]

4870 Trifire has better Performace than 4870 Quadfire in many games. But the real problem is still the Microstuttering...

[bigcesar]

the platform

[16floz470ml]

Microstuttering is an issue. This Far Cry 2 graph with 9.2 driver is a great example. Every dip you see is the second card dropping out of crossfire. At the end the second card had dropped out entirely. If you did not have this graph then the benchmark fps would seem good. I also observe the power draw go down on my watt meter when the fps go down.

[okorop]

Omg what a configuration!!!!!! so with 3d mark vantage nvidia goes better than ati but in the 06 is the opposite
is the 3d mark vantage done with physx on or off?

[Paxi]

Have you also got such a graph from microstutterin with the 2 x 295?
And very nice review man!

[kemo]

I do have a question for the 30" users , can you guys tell the difference between 4xAA and no AA during hardcore gaming ?

[r1rhyder]

In the heat of the battle, hell no. I hardly ever use AA, looks good without it. And I agree about the Far Cry 2 results being way off.

[NameTaken45]

i would of liked to see the X3 Bench compared.

[RAID]

So two people have called out my results and although I still stand by my original testing and posted scores, I'll entertain the possibility of error on my part and conduct more benchmarks.

Since my Far Cry 2 scores are "way off" I suppose that's a good place to start. I'd like to note that none of the members who claimed my results to be wrong run the same set up I used for my testing. But I am interested in seeing their FC2 benchmarks run with the same settings I used.

Far Cry 2
2560x1600 60Hz
Ranch Long - 3 Loops
Anti-Aliasing - 8x
VSync
Direct3D 10
Fire - Very High
Physics - Very High
Real Trees - Very High
Vegetation - Very High
Shading - Ultra High
Terrain - Ultra High
Geometry - Ultra
Post FX - High
Texture - Ultra High
Shadow - Ultra High
Ambient - High
HDR
Bloom





After running Ranch Long bench again (and again), the results are similar to the original score:

Original - 17.91 fps

Second - 18.52 fps

Third - 16.80 fps

That's a total of 9 loops with similar average fps results. Not sure how these results are "way off" but I look forward to an explanation from the people that claim so.

[Xello]

I do have a question for the 30" users , can you guys tell the difference between 4xAA and no AA during hardcore gaming ?

Increasing resolution negates the usefulness of AA as long as your screen size isn't increasing relative the the resolution, because at 30" and 2560x1600 pixels are small enough that we can't see aliasing like we can on lower resolutions.

It's more useful for those of us with 40"+ screens @ 1920x1080.

[r1rhyder]

Raid, could you put a check mark in the "enable ati overdrive" box and rerun the benchmark. See if that makes a difference. I have passed my copy of far cry 2 off to a friend and don't have the benchmark available at the moment. I will get it when I see him soon and reinstall it. I can't find the benchmark for download anywhere, ubisoft has removed it from the web it seems. I will get back to you on this.


Edited: and please run the benchmark with vsync off. A cap of 60fps will lower the average framerate. I remember playing this game at over 100fps most of the time at just 3Ghz.

[16floz470ml]

I am not at my computer right now to show benches but Raid is right on the Far Cry 2 with 8xaa. 4xaa with ATI works but when you use 8xaa with Far Cry 2 it almost seems like the second cards drops out totally. Maybe ATI forgot to program that in or something.

I and most other people I am sure are also interested in the Crysis Warhead results. That is the only bench that seems off to me now.

[RAID]

I and most other people I am sure are also interested in the Crysis Warhead results. That is the only bench that seems off to me now.

From reading through your posts in this thread, I discovered that my listed settings for Crysis Warhead in my original post is not fully accurate. My GTX 295 runs did allow 16X AA as listed but the 4870X2 CrossFire benches only allowed a maximum of 8X AA, which is what I used for my orignal results. I have noted this in the first post but I suppose the proper action to take is to conduct the GTX 295 SLI bench in 8x and post those results for Crysis. I'll do that at a later time since I currently have my ATI cards installed. So in short:

Crysis Warhead Bench - GTX 295 SLI
2560x1600
Ambush
Anti-Aliasing - 16x (Max)
VSync
DirectX 10

Crysis Warhead Bench - 4870X2 CrossFire
2560x1600
Ambush
Anti-Aliasing - 8x (Max)
VSync
DirectX 10

Here are the follow up benchmark results with the ATI cards.







Original - 12.15 fps

Second - 13.96 fps

Third - 11.48 fps

Fourth - 10.27 fps

The last run was a ranked benchmark that ended up being the slowest of the bunch but you can check it out for yourself in the online ranking results. Anyways, I feel that the results are close enough and that the original score is accurate.

[r1rhyder]

By r1rhyder at 2009-03-07



Could you put a check mark in this CCC box circled in red and then hit apply.

[16floz470ml]

I have 4850x2's and I have the box checked for overdrive. It makes no difference for me. ATI cannot do 8x aa at high resolutions and remain playable.

[r1rhyder]

This is what I was saying, no vsync and overdrive enabled, not just unlocked. Video cards were running at 750/900, not 507/500. If you want, I could do the three loop run, but I don't see the point. And this is not photochopped.



By r1rhyder at 2009-03-08

[16floz470ml]

Very nice results. I am currently using the 9.2 drivers. What drivers are you using? In Far Cry 2 and Crysis Warhead I cannot get 8x aa to run acceptable. Raid seems to have the same problem. In Battlefield 2 8x aa works fine but it is an older game. Are you doing anything else to make it apply a crossfire profile?

[r1rhyder]

Im using 9.2 on Vista 32 SP2, with all updates, Directx nov update. CCC overdrive unlocked and enabled, fans at 50%, AI on/standard, 8xAA enabled/box, crossfire enabled. Sometime's, setting AA in CCC to match AA in game makes a big difference. Did not rename any .exe's. I wouldn't say it run's smooth as butter with 8xaa. Background logging with Everest show's it does use all four gpu's at 750/900. I used to have two 3870x2's, and they ran pretty damn good too. And they overclocked alot better than these two 4870x2's I have now.

[gatecrasherlok]

Yeah I thought the results on the first page were a bit off for farcry.

[gamervivek]

I have 4850x2's and I have the box checked for overdrive. It makes no difference for me. ATI cannot do 8x aa at high resolutions and remain playable.

4850x2 doesn't have the bandwidth for higher resolutions and 8xAA,you can check reviews where it would run close to 4870x2 till 1900x1200 and then drop off abruptly at higher resolutions.

as for OP-i am not sure of warhead but in original crysis benching tool,8xQ is the true 8xMSAA,16xAA and 8xAA are nvidia supported modes that are marginally worse or better than 4xAA on ati cards and do not show the real picture.
16xAA should be less performance and memory demanding than 8xQ on nvidia cards too,16xQ is the higher AA mode from 8xQ.

This is what I was saying, no vsync and overdrive enabled, not just unlocked. Video cards were running at 750/900, not 507/500. If you want, I could do the three loop run, but I don't see the point. And this is not photochopped.



By r1rhyder at 2009-03-08

zomg 100fps on 2560x1600

can you also do the ranch long?only a single loop should suffice,though FC2 is quite abrupt with different runs.

[16floz470ml]

I do not believe that bandwidth is the problem for me. I can bench higher on my system than these other guys benches if I use the settings that they use. My thing is that for you to use aa as it is intended for the game you have to leave the CCC option for aa to app controlled. This is what Raid did as the op. If I move the aa slider up in CCC and ignore the game settings the game will not look the same. If we are going to compare benches then we all need to use the same 3d settings. The last bench that was posted also was using standard ai instead of advanced. I thought the whole point of this thread was to turn the settings all up to eleven and compare benches.

I can see your point about the Nvidia aa settings though. If within the game itself the 8x aa is really only for Nvidia that would explain everything. I mean Crysis has the Nvidia logo in the loading screen. Is Far Cry 2 sponsored by Nvidia also?

[clayton]

Hot review. Hope next-generation single cards perform anywhere near that.

[slim142]

I like this review, thanks RAID, very informative.
It seems like ATI needs to improve.

Now, if you could add 1280x1024 for "reference", it would help to have a base.

[buddatech]

Nice review, thanks! So in a single card config which would be the way to go?

[JOCKTHEGLIDE]

Nice review, thanks! So in a single card config which would be the way to go?

GTX295 FTW