Interesting Crysis Multi GPU scaling results

X

xp3nd4bl3

2[H]4U
Joined
Sep 25, 2004
Messages
2,259
I recently abandoned my 780i dual G92 GTS SLI setup for the setup in my sig. I did this knowing very well that "Crysis sucks on ATi cards and doesn't scale well". I found this to be very true on my setup in DX10. But I found a very different story when trying the game in DX9 mode. Check out these results:

All tests done Crysis 1920x1080 with everything on HIGH

DX10

4 ATi 3870 (CrossFireX) 34.4fps
2 Nvidia 8800GTS 512 SLI 38fps

DX9

1 Nvidia 8800GTS 512 22fps
2 Nvidia 8800GTS 512 39fps (77% improvement over 1 GPU)
1 ATi 3870 20.2fps
2 ATi 3870 (x2) 36.5fps (79% improvement over 1 GPU)
4 ATi 3870 (2 x2) 52.5 fps (44% improvement over 2 GPUs 160% improvement over 1 GPU)

Ultimately I'm really happy with how the game runs in DX9 and I'm quite pleased with my decision to switch. As a long-time nVidia fanboy I'm sure I'll switch back to the GT200 vs R700 but for now I'm a happy ATi user.

Just to test compatibility I also loaded up Bioshock, Call of Duty 4, Call of Juarez, Frontlines, and Rainbow Six Vegas. All the games ran very well with CrossFireX and I found no issues or bugs. The big pleaser was R6: Vegas and GoW which run significantly faster than they did on my SLI setup.
 
Thanks for the results. I really appreciate those results. I am thinking of my next purchase will be 2 of those ATI monsters when I upgrade my mainboard later this year.
 
You must be using watercooling right? the heat output on that thing!!! Well, I say FPS supersedes everything else.

Master Yoda says: To achieve gaming nirvana, crossfire you must.
 
Cool I'm glad this was beneficial.

ScourggeFX said:
You must be using watercooling right? the heat output on that thing!!! Well, I say FPS supersedes everything else.

Master Yoda says: To achieve gaming nirvana, crossfire you must.
Click to expand...

LOL, no H2O, just an Ultra120 Extreme. I discovered that by putting 2 fans on it (push/pull) my temps dropped 3-4C and my OC potential went way up. I can actually hit 4GHz but it won't stay stable with 4 sticks of RAM.

I'm definitely liking CrossFire. I miss playing around with SLI profiles but CrossFire has the easy factor going for it. Plus it doesn't take 3 days to turn on and off in Vista like SLI does. :D
 
make sure you get the Cuban Ultima Doom Sword's mod from here. It sounds oh so silly but it's got incredible IQ ( dx10 quality in dx9) and it's easy on the computer.
 
@ 1920x1200 with everything on High I can get betwen 30 and 50 fps as well on the DX9 code path. unfortunately, I have the powerplay bug and the game hitches all the time.
 
-=PESTILENCE=- said:
@ 1920x1200 with everything on High I can get betwen 30 and 50 fps as well on the DX9 code path. unfortunately, I have the powerplay bug and the game hitches all the time.
Click to expand...

I had never even heard of the powerplay bug until you mentioned it. I'm going to look into it.
 
38xx series card apparently clock down to 2D speeds during 3D gameplay. Yikes!:eek:
 
skadebo said:
make sure you get the Cuban Ultima Doom Sword's mod from here. It sounds oh so silly but it's got incredible IQ ( dx10 quality in dx9) and it's easy on the computer.
Click to expand...


Low quality shaders and textures + sun-ray bursts = win!!!
 
xp3nd4bl3 said:
Cool I'm glad this was beneficial.



LOL, no H2O, just an Ultra120 Extreme. I discovered that by putting 2 fans on it (push/pull) my temps dropped 3-4C and my OC potential went way up. I can actually hit 4GHz but it won't stay stable with 4 sticks of RAM.

I'm definitely liking CrossFire. I miss playing around with SLI profiles but CrossFire has the easy factor going for it. Plus it doesn't take 3 days to turn on and off in Vista like SLI does. :D
Click to expand...

Nice setup you should post some pics of it not alot of members have Quad CrossfireX setups. I'm curious about the push pull thing is that 2 fans facing opposite directions on either side of the heatsink?
 
kadozer said:
Nice setup you should post some pics of it not alot of members have Quad CrossfireX setups. I'm curious about the push pull thing is that 2 fans facing opposite directions on either side of the heatsink?
Click to expand...

Nope, two fans facing the same direction on either side of the heatsink. Here's a pic. Sorry for the scotch tape mod, haven't had time to fab up a bracket yet.

x38setup.jpg
 
Why is there such a large difference between your results and [H]'s results? Is it because of the 1.2 patch?

The idea of 50+ fps on high has me drooling right now. I also play at 1920x1080, and even with my 8800 Ultra, I have to play on all low settings in order to get the frames I need for competitive multi player.
 
Raxxath said:
Why is there such a large difference between your results and [H]'s results? Is it because of the 1.2 patch?

The idea of 50+ fps on high has me drooling right now. I also play at 1920x1080, and even with my 8800 Ultra, I have to play on all low settings in order to get the frames I need for competitive multi player.
Click to expand...

Not that I doubt it...I sure as heck wanna believe it... a Crysis Benchmarking Tool screenie would be nice.


Nice rig, by the way...
 
Nope, two fans facing the same direction on either side of the heatsink. Here's a pic. Sorry for the scotch tape mod, haven't had time to fab up a bracket yet.
Click to expand...
Very nice especially with the 3rd fan exhausting it out. Gotta get me a thermalright now.
 
Ogilvy said:
Looks legit to me...based on that screen, it looks like he ends up at least 50 fps.


Nice!
Click to expand...

the in-game benchmark is notoriously generous when it comes to frame rate. at the very least he should have tried a custom time demo.
 
rennyf77 said:
the in-game benchmark is notoriously generous when it comes to frame rate. at the very least he should have tried a custom time demo.
Click to expand...

At the very least I could have not shared my results and contributed nothing of value, like you're doing right now.

I realize that the canned benchmark is not indicative of observed gameplay performance. I also realize that many people on this forum don't think timedemos provide ANY valuable information. I personally believe that timedemos are great for comparative analysis of how one configuration performs vs another in a repeatable method of testing.

For example if graphics card A gets 50fps in the timedemo graphics card B gets 40fps, that doesn't necessarily mean it'll get 50 and 40fps respectively in every situation in the game, but it DOES tell me that graphics card A is faster than graphics card B which is valuable information to me.

Bottom line; if you don't like my testing or results then skip it; don't read my post and stay out of my thread.
 
Wow nice tests and please post some more pics of that sexy build. I like how you have the wiring and really do want to see more!
 
Raxxath said:
I get over 100 fps average in the GPU benchmark, yet have a hard time staying above 50 or so in-game.
Click to expand...

At what res? 320x240? The highest ive seen so far in the gpu benchmark under dx9 @ 1280x1024 was around 75-77 fps, and that was utilising a quad 9650 at 4.5 ghz and crossfire 3870 x2's overclocked.
 
Chris_B said:
At what res? 320x240? The highest ive seen so far in the gpu benchmark under dx9 @ 1280x1024 was around 75-77 fps, and that was utilising a quad 9650 at 4.5 ghz and crossfire 3870 x2's overclocked.
Click to expand...

Like I said above, 1920x1080, but on all low settings.
 
punbx360 said:
can we please get some crysis benchmarking tool screens after all test have been run with the min max and average displayed thanks
Click to expand...

You can't estimate those things from the screenshot?

You haven't seen that timedemo benchmark for yourself at least 100 times by now?



Or do you think he faked his results?
 
Ogilvy said:
You can't estimate those things from the screenshot?

You haven't seen that timedemo benchmark for yourself at least 100 times by now?



Or do you think he faked his results?
Click to expand...

i don't think he has faked the results (i could care less) but i would still like to see a screen whats the problem with taking a screen after its finished is that too much to ask for (and if it is then don't do it)

on another note xp3nd4bl3 please consider posting some ingame screens in the crysis screenshot thread i would love to see how incredible the game must look running on your setup

http://www.hardforum.com/showthread.php?t=1251098
 
Nice posts of your results.
Looks like a very nice setup.

For contrast, my computer will play Crysis DX10 all settings on very high with roughly 40 FPS sustainable.
Yes its not entirely apples to apples, but you have 4 GPUs to my 3 so that may make it worthwhile; and I use a 680i board.
You also may have spent a bit less money.:D
 
magoo said:
Nice posts of your results.
Looks like a very nice setup.

For contrast, my computer will play Crysis DX10 all settings on very high with roughly 40 FPS sustainable.
Yes its not entirely apples to apples, but you have 4 GPUs to my 3 so that may make it worthwhile; and I use a 680i board.
You also may have spent a bit less money.:D
Click to expand...

would you also consider posting some ingame screens in the crysis screenshot thread i think you'd be the first tri gtx sli

http://www.hardforum.com/showthread.php?t=1251098
 
well having used both the maximus forumla w/ 2x 3870 x2's and the evga 780i and 2x8800gtx's I would say overall they are about the same in terms of game performance. I though opted for the 780i right now as its doing the job I want it to do and I honestly never knew the whole performance difference anyhow. As for board stability.. I have the overnight RMA policy on this board and a backup board right now as well just in case anything comes up. Right now overall the 3870 x2's definitely shine in DX9 games, but I tend to stick with nvidia cards right now as overall they are doing what I need them to do.
 
You must log in or register to reply here.
Back
Top