Crossfire not working

C McCarter

Limp Gawd
Joined
Mar 13, 2007
Messages
183
Hi,

It seems that my CF setup is not working. In GPUZ the first 6870 is CF enabled, while the 2nd is disabled. I have seen no improvement in frame rate in Lotro ~40-80fps (outside), ~20fps (busy town). I have downloaded and reinstalled 11.9 and CAP, but it did not help. I have CrossfireX enabled in AMD Vision and it does show both GPUs. No information is displayed in the AMD CrossfireX Diagnostics.

My specs are in sig.

I am aware that my PCIE 2.0 are running at 16x and 4x.

Any help would be great!
 
It's normal in GPU-Z to display the second as disabled. Click the sensors tab, then switch to the second card ("disabled"), and watch the GPU load category. Fire up a few games, leave GPU-Z running in the background, and either alt-tab, or look after the game is done, and see if it has been graphing anything in the "GPU load" category. That's a quick and easy way to tell. Even Kombuster uses 2% (just shows that it recognizes the second card, not really utilizing it at all).

For the PCI-E, My board is running at 8x+8x.. It's odd that it runs at 16x and 4x, but that might just be a difference with your chipset. Read some reviews of your board and see if they talk about what is or is not normal for that board to run crossfire with.
 
what resolution?

I just got my 2nd 6850 today and had the same experience. I was kinda bummed, but then I bumped up the details and my frames went up.. On meduim settings in BFBC2 @1920x1080 no AA or AF, i was getting 46-50 FPS with xfire, then i put everything High with 4AA and 4 AF sitting at a nice 80fps with lowest to ~65 here and there.
 
I am running 1920x1200. I forgot to mention that I have a second monitor (not the same size) but only run games on my main.

I have looked in games at GPU usage and I get maybe 3% for LOTRO on the second GPU.
 
To be certain, D/L Afterburner, and use it to monitor your GPU useage and FPS. It's good software.

The X4 slot may be a problem.

I'd try reinstalling the driver and CAP again just for good measure.

Also do others have good success with CF in the games you are interested? Maybe the game doesn't support CF?? Google it to make sure.:D
 
Quick question regarding Crossfire, do you use 2 bridges or 1? I always assumed one (and only used one bridge when i SLI my 8800GT's) but i saw a picture a few weeks ago of a guys rig on [H] and he had 2 6950's with 2 bridges attached between both of the cards.

To clear up i am asking for myself. I have no idea if you are supposed to use 2 or 1, hence i am asking everyone.
 
Last edited:
Both cards seem to run (35%) while a single card is ~88% while playing BFBC2 @ Max Graphics x4AA 1920x1200. My FPS was 60-70 with a single card while CF was 60-90FPS. I have not had time to do extensive testing but I might load Crysis to see how it does.

I am using only 1 bridge fyi, but thanks for pointing it out as a possible source of error.
 
Quick question regarding Crossfire, do you use 2 bridges or 1? I always assumed one (and only used one bridge when i SLI my 8800GT's) but i saw a picture a few weeks ago of a guys rig on [H] and he had 2 6950's with 2 bridges attached between both of the cards.

You will find some difference of opinion on that.
Realistically, there are two connectors for the Triple and quad-fire conformations.
You only need one brigde to connect two cards. (ie the minimum)

But......I have always used two and see no problem with it, currently I can only use one due to my setup. I did the same when I had SLi.

Some will argue that two bridges introduces stuttering, but I have never seen that published scientifically. Nor have I seen where two are better than one......it's just what I do.
 
I think one bridge is fine

It is true, the proof is in the pudding, fire up kombustor, a bitmining client or a folding client, anything that will 100% load the card, then watch utilization and temperature.

What does CCC say??
 
Both cards seem to run (35%) while a single card is ~88% while playing BFBC2 @ Max Graphics x4AA 1920x1200. My FPS was 60-70 with a single card while CF was 60-90FPS. I have not had time to do extensive testing but I might load Crysis to see how it does.

I am using only 1 bridge fyi, but thanks for pointing it out as a possible source of error.

Was that BFBC2 single player or multiplayer?
I have read that CF isn't so hot in MP.
I don't play MP in BFBC2, so I can't help you much there.
I would think in Crysis SP you will see an advantage.

If your cards are only running 35% in CF, thats not really well optimized. Most games if well optimized in CF will run both cards at 80-90% most of the time.
 
Was that BFBC2 single player or multiplayer?
I have read that CF isn't so hot in MP.
I don't play MP in BFBC2, so I can't help you much there.
I would think in Crysis SP you will see an advantage.

If your cards are only running 35% in CF, thats not really well optimized. Most games if well optimized in CF will run both cards at 80-90% most of the time.

+1
I havent used CF but i have read a lot of posts about it and in all of them that talked about BC2 they mention CF is horrible in BC2, it even downscales for some. (aka they get better frame rates with CF turned off).
 
Just ran Kombustor (extreme) and got 1548 with CF on and 1570 with CF off..... not sure what is happening...Utilization on cores is 100% with running CF and 100% on GPU2 when not running CF.
 
Hmm thats strange.. Wonder if that 4x is actually choking the card?

Do you have anything in the PCI4x slot? If so that might be a problem to according to neweggs info

1 x PCI Express x16 slot, running at x16 (PCIEX16)
1 x PCI Express x16 slot, running at x4 (PCIEX4)

Note: The PCIEX1_1 and PCIEX1_2 slots share bandwidth with the PCIEX4 slot. When PCIEX4 slot is populated with a x4 card, the PCIEX1_1 and PCIEX1_2 slots become unavailable.
 
My MoBo is 1 PCIE 16x and 1 PCIE 4x so it could be the 4x. However, I have seen multiple tests where 4x does not reduce performace of CF by much (I think around 10% from 8x8x or 16x16x).
 
I was able to force PhysX CPU and my Kombuster score went to 4700. ~60% increase. Could this be the problem? The games are using the second GPU for physics instead of the CPU?
 
I was able to test it with Crysis. With CF 30-70 FPS everything Very Hight 4x AA. Same settings without but not real difference between FPS rates.
 
I was able to force PhysX CPU and my Kombuster score went to 4700. ~60% increase. Could this be the problem? The games are using the second GPU for physics instead of the CPU?

You can't run Physx on AMD cards.:eek:

Just turn that off completely.

Honestly, if all you have populated are the PCI-e #1 and 2 slots, I think the 4X is choking your second card.

Are you sure each one works right?

Take turns running each card alone, running Kombuster on each card singly.
 
Just ran Kombustor (extreme) and got 1548 with CF on and 1570 with CF off..... not sure what is happening...Utilization on cores is 100% with running CF and 100% on GPU2 when not running CF.

Thats FUBAR......CF should be much higher than one card alone!
 
That is the feeling I am getting as well. I am going to try a BIOS update and switching CF Bridges. If that does not work, time for a new mobo. Lucky I plan on building a HTPC this winter so I can just use by current mobo for that. Thanks for all the help with this!
 
I'm checking out LOTRO and it actually runs worse in fullscreen mode (crossfire enabled).

I'm on 11.9 CAT and 11.9 CAP4.
 
I had the same problem last night when I installed my two 6950's........ I just reinstalled the drivers again. I read somewhere you have to do it maybe a couple times and it will get crossfire working.

I didn't uninstall the drivers just reinstalled again and it worked.

Good luck!
 
1. Get MSI Afterburner - latest version can be gotten fromm guru3d.com
2. Extend your display to the second monitor
3. Open up taksmanager, go to the performance tab and put it on your second monitor
4. Open up MSI Afterburner and put it on your second monitor

5. Run whatever games you want and watch the CPU and GPU load.

My bet is that you are being limited by your CPU and/or your RAM throughput in CF mode.
 
Just loaded Crysis. 25(min) 75(max) FPS CPU Usage ~60-80% 3.25GB ram usage. CPU and Ram do not seem to be limiting.
 
Just loaded Crysis. 25(min) 75(max) FPS CPU Usage ~60-80% 3.25GB ram usage. CPU and Ram do not seem to be limiting.

What did MSI Afterburner show for GPU usage on both cards? GPU1 and GPU2 should have both shown up.
 
I will give that a shot soon.

Quick question, would it be possible that the x4 lane is decreasing performance?
 
I will give that a shot soon.

Quick question, would it be possible that the x4 lane is decreasing performance?

Not that significantly theres been tests done on [H] that show the difference in performance with x16/x16, x8/x8, x16/x4. To be relatively minor. It probably wouldn't be noticeable outside of benchmarks.
 
OCd to 3.7 @ 1.488V and did not see much if any gain in any test (tried Crysis, LOTRO, Kombustor). However my temps are ~60-68C
 
What is the stock voltage for that chip? You really shouldn't have to bump it up that much for only a 300Mhz increase.

You may want to try overclocking via bus speed to get the north bridge speed up as that will affect your overall speed a lot more than just bumping up the multiplier. Or at least bump up your north bridge speed if that setting is available in BIOS.

As for Crysis... it was known for bad multi-gpu scaling back in the day.. but.. with my setup (in sig - only running one 4870x2) I am never getting below 50fps - 1920x1080.. everything maxed except for shadows which are on medium and no AA.

I haven't played LOTRO or used kombustor.
 
Tried uping my northbridge speed to 2600 from 2000...nothing... not sure what to do...
 
DLed 3Dmark11 (basic) and saw a 34% increase with CF on. Seems it is working just not well.
 
Would a new motherboard solve my problem? I.e. one that is CFX supported.

That would make nor difference whatsoever. CF is enabled, so your board supports Crossfire.

DLed 3Dmark11 (basic) and saw a 34% increase with CF on. Seems it is working just not well.

From the different benches I have looked at, 3dMark doesn't show anywhere the size of increase in speed that would make sense when using CF or SLI.

Did you install the AMD chipset drivers for your motherboard?

Have you tried overclocking the PCI-E speed a little.. say to 105Mhz? It will probably make a difference IF the x4 slot is actually causing the slowdown.. but I doubt it is.
 
Reinstalled the chip set drivers from Gigabyte. No Effect.

OCd the PCIE frequency to 105. No Effect.
 
Back
Top