• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

crossfire 6850 - what am I doing wrong?

T

turbogeek

Limp Gawd
Joined
Nov 13, 2008
Messages
179
I recently jumped on the crossfire bandwagon by adding a 2nd 6850 to my setup (both XFX dual fan cards, new card is rev. 1 and is apparently built on the 6870 pcb, has 2 6-pin power connects vs the older style single 6-pin power connect).

I am getting poor results (0 scaling) when gaming with these cards crossfired and have spent several days scouring the internet for answers.

Tonight I ran a lengthy bench test comparing FPS and GPU% usage across several programs to get some solid data. Judging by the results, I feel like it is something simple that I'm missing.

Rig Specs:
Asus P5E
E8400 (oc 3.9 GHz)
8 GB G.Skill
Kingston 64GB SSD x2, raid-0 (OS & games)
TT Toughpower 700w PSU
1080 resolution
XFX 6850 x2, crossfire
Catalyst Driver 11.11b
CAP2
used driver sweeper, etc.

Data collected through Fraps and Afterburner.
VSync was OFF for all games.
Test results:

Furmark
-----------------------
Single GPU
FPS AVG - 37
GPU1% - 99%
GPU2% - N/A (0%)

Crossfire
FPS AVG - 72
GPU1% - 99%
GPU2% - 99%
--------------------------------------------------------

Heaven Benchmark
-----------------------
Single GPU
FPS MIN - 11
FPS MAX - 55
GPU1% - 99%
GPU2% - N/A (0%)

Crossfire
FPS MIN - 20
FPS MAX - 99
GPU1% - 99%
GPU2% - 99%
---------------------------------------------------------

Shogun 2
-----------------------
Single GPU
FPS MIN - 21
FPS MAX - 48
FPS AVG - 33
GPU1% - 99%
GPU2% - N/A (0%)

Crossfire
FPS MIN - 17
FPS MAX - 52
FPS AVG - 32
GPU1% - 25-50%
GPU2% - 25-50%
--------------------------------------------------------

DiRT 3
-----------------------
Single GPU
FPS MIN - 49
FPS MAX - 74
FPS AVG - 63
GPU1% - 99%
GPU2% - N/A (0%)

Crossfire
FPS MIN - 49
FPS MAX - 74
FPS AVG - 64
GPU1% - ~50%
GPU2% - ~50%
(usage while racing, menus appeared to bump up to nearly 99%/99%)
--------------------------------------------------------

Skyrim
-----------------------
Single GPU
FPS MIN - 20
FPS MAX - 67
FPS AVG - 37
GPU1% - 99%
GPU2% - N/A (0%)

Crossfire
FPS MIN - 8
FPS MAX - 60
FPS AVG - 28
GPU1% - 30-50%
GPU2% - 30-50%
--------------------------------------------------------


If anyone can read into these benches and lend some insight. I don't think it's the driver. The 11.11 drivers are the newest and the a/b/c revisions are only for current game issues. As I understand it, older games like DiRT 3 and Shogun 2 have been fixed to work correctly with crossfire. However, I understand that sometimes new driver updates break previously working titles, etc.
Clearly, something is causing games to limit to roughly single GPU performance, just distributed across twice as much $$$. (but for some reason, benchmark tests are scaling properly.)
 
Last edited:
again, this is only happening in games. benches and game menus are scaling.

what is the likelihood that my old E8400 is the problem? bottlenecking?

I see a lot of forum posts writing off the issue saying "you need quad core. C2D is ancient. i7 is the only cure. blah blah blah."
then I read this http://alienbabeltech.com/main/intel-e6850-bottleneck-investigation
showing that even an underclocked Intel E6850 is enough to pass the bottleneck to the GPU.

I could understand an i7/i5 giving improved performance, but for the E8400 to bottleneck so bad that i am stuck at nearly exactly 50/50 split load (single card performance) doesn't seem to add up. If I threw a 6970 at my e8400 I highly doubt I'd be seeing only 6850 numbers. I know it's an old CPU, but it should still work and give some improvement right?

The P5E board also gives full x16 lanes for both cards in crossfire.
 
Last edited:
turbogeek said:
what is the likelihood that my old E8400 is the problem? bottlenecking?

I could understand an i7/i5 giving improved performance, but for the E8400 to bottleneck so bad that i am stuck at nearly exactly 50/50 split load doesn't seem to add up. I know it's an old CPU, but it should still work and give some improvement right?

The P5E board also gives full x16 lanes for both cards in crossfire.
Click to expand...

I was thinking your CPU is whats holding you back. What resolution are you using?
 
1920x1080

even though I would like to upgrade to SB soon, I don't want to just throw money at this crossfire problem and hope it goes away. know what I mean?
 
Most likely CPU if adding more GPU power has no increase in games.
 
well, tonight I'm going to run some more benches similar to the link I posted earlier to see if I can determine if my CPU is capped out. (Underclock x% CPU test vs. Underclock x% GPU/Crossfire test, and see which hits my FPS worse. If I still have headroom on my CPU, then the GPU underclock should have a much more significant effect on FPS and something else is the culprit. In theory.)

Meanwhile, if anybody else wants to chime in, I'm all ears.
 
I also think you're probably at the max your going to get out of that C2D. I would expect the higher IPC of the SB architecture would yield better performance at a lower clock speed.

Also that you do get better performance out of the benchmarks leads me to think your CPU is the bottleneck. Especially since those tests are pretty much GPU only tests.
 
When playing games, check to see what your CPU load is in the Task Manager. Also you might want to try setting profiles for your CFX, the newer driver releases should allow you to do that.
 
I ran a few more tests last night and found that my CPU was about 90-100% on both cores regardless if the gfx settings were ultra-low or ultra-high, and that the crossfire saturation was still only about 50% in both scenarios. Some games got ok frame increase with low grafix and some didn't get much at all. soooo....

I think I'm about to just go ahead and pull the trigger on an i5 upgrade. I've just been procrastinating on it anyways. The z68 boards are cheap enough and ddr3 almost free. I like the Asus lineup, even the midrange P8xxx boards have all the nice stuff to make for solid overclocking, just without a bunch of extra bells and whistles.
 
interesting...
I just installed the Beta 12.1 drivers and, at least in DiRT3, I saw a nearly total GPU saturation in crossfire. @ 'Ultra High' settings with full AA was getting 85-100% on both cards when racing whereas before I was getting a pretty solid 50% per card regardless gfx settings.

Also, I was getting much better GPU usage in Shogun 2 at maximum settings. Only about 65-70% tops, but better than the dismal 25-50% I was previously getting.

As far as skyrim, there wasn't much change but that may still be borked. Will attempt to tweak some more (don't have any CAPs installed for instance, may not matter)

Aside from the improved (fixed?) GPU saturation, I noticed that my CPU usage was dynamic during games and was not constantly stuck at 100%. Before, every game had my CPU cores pegged. HOWEVER, it is clear from Skyrim and Shogun 2 that my CPU does cause the game to hitch when it gets overloaded and the hitching occured with hi and lo gfx.

I think (I hope) there's more to this yet and I might just be able stretch my dinosaur e8400 a little longer. lol.
 
Last edited:
skyrim's another beast of its own xD how do u know its not the crappy engine ?
Did you have the large address awareness patch ? cause the game tens to chop on its own!
but yeah Skyrim gives better performance with higher cpu frequency
specialy if you got shadows to ultra

12.1 you can make your own profiles for any game tho u gotta do them manualy >.>
and if a cap doesn t work you can override them and make one and see if that makes the difference.

Fus ro dah!
 
CFX support for Skyrim has been pretty horrible in general. It seems to take AMD awhile to put out driver updates for CFX issues in games. I am glad the driver update fixed some of the issues though, better than spending all that money for an upgrade if you don't need it.
 
NoOther said:
CFX support for Skyrim has been pretty horrible in general. It seems to take AMD awhile to put out driver updates for CFX issues in games. I am glad the driver update fixed some of the issues though, better than spending all that money for an upgrade if you don't need it.
Click to expand...

Yeah...
I went ahead and jumped on the SB bandwagon with a z68 board and 2500k.
I've been supremely impressed with my i3-2310M Toshiba Satellite and I know the updated architecture will be night and day better with the 2500k over the E8400.
In fact, I'm betting it'll walk circles around it.
Should have the new gear on Thursday. Just enough time to drool all over it before I leave town for a week and a half for the holidays. lol.
 
Last edited:
turbogeek said:
Yeah...
I went ahead and jumped on the SB bandwagon with a z68 board and 2500k.
I've been supremely impressed with my i3-2310M Toshiba Satellite and I know the updated architecture will be night and day better with the 2500k over the E8400.
In fact, I'm betting it'll walk circles around it.
Should have the new gear on Thursday. Just enough time to drool all over it before I leave town for a week and a half for the holidays. lol.
Click to expand...

Heh, yeah my latest computer build I got all the parts in...right after I had an emergency that kept me busy for 3 weeks. It stunk having to look at all those glorious parts and not being able to build the computer yet..
 
You must log in or register to reply here.
Back
Top