2 vid cards in mobo, not crossfired...lower 3dmark score? What gives?

[Debeucci]

So I had a strange thing happen to me recently. I purchased a 2nd 5850 for my computer and ran a few benchmarks...ie...3dmark11, to see which one was faster. They were essentially the same.

Then I slotted both of them to crossfire and then realized that my crossfire cable was too short. I left the 2nd card in while I ordered the longer crossfire cable and waited. Just for kicks, I ran 3dmark11 again and my score was much lower. Looking at the 3dmark11 numbers, my primary card seems to be running much slower, it seems like it downclocked or something. What gives? The only difference is that I have 2 cards in there right now, but shouldn't the 2nd card just be sitting there doing nothing? Because the crossfire cable isn't there, could it be making my primary card slower?

Thanks

 

[SpamShadow]

What motherboard do you have? Some motherboards (my ASUS for instance) require a PCI-e Switch card in order to run a single slot at full speed (switch card should have come with your mobo if you need one).

 

[Debeucci]

I have an old 680i. I'm not aware of the pci-e switch card. When I was just running one card, it seemed to run at full speed fine.

i have an EVGA 122-CK-NF68-AR LGA 775 NVIDIA nForce 680i SLI ATX Intel Motherboard

 

[SpamShadow]

Are you planning on using hacked drivers? You can't Crossfire on an SLI board, its not supported. I hear you can hack your drivers in order to make and SLI board Crossfire, but I've also heard that it doesn't work very well.

Is the card still plugged in and running? Having an extra PCI-e device running not doing anything will slow your system down.

 

[Debeucci]

I asked the forums about that and it seems not to be true. I thought you need an SLI mobo to SLI, but you don't need a crossfire mobo to crossfire, You can crossfire on any motherboard. That's what I've read at least.

And yes...i'm waiting for sandy bridge before I order my new mobo. I'm just taking advantage of a rebate from newegg and got the 2nd card early.

 

[wolfram23]

Pretty sure you can crossfire on any board, although I don't have experience with Nvidia chipsets so I can't be certain, but I know that's the case for all other brands.

Anyway, it's possible that the performance suffered from bandwidth being cut in half, although that shouldn't have been an issue. I'm curious how much the difference is?

 

[Debeucci]

TBH, i'm not sure. my 3dmark score was about 10-15% lower. My old single card setup was running around 28-3100 with the settings I had. After having both cards, it ran 25-2600ish. I have an x6800 core 2 duo. It's funny because the reason I reran the test was because I OC'd it from 2.93 to 3.3. Imagine my shock when my score came back even lower.

Now I remember, it came back 2630 with the OC, and after I set it back to default, it was 2500...so roughly 10% lower.

The 5850 runs at 650 clock speed default, and I think 3dmark showed it running at something insanely slow like 150.

 

[wolfram23]

150 core, 300 mem is the idle clock speed. It's possible 3D mark was just reading your 2nd card at idle.

Stock clocks should be 700/1000 tho iirc.

But anyway, I had some odd issues with my 5850 3DMark Vantage scores changing. I got a 31000 GPU score once and then I just couldn't get that high again. Don't know why it was so good for that one run. I tried overclocks on both CPU and GPU but it didn't help hit that score again.

You could try messing around with a few settings in your BIOS like PCIe frequency or PCIe amplitude if your mobo has features like that. Otherwise I wouldn't worry too much about it.

BTW if you're interested I have a thread with a bunch of single vs CF results, and I meerly disabled CF rather than removing the card so you can see if the results are similar (although mine were at an OC setting but you could match it)

http://www.tomshardware.com/forum/289809-33-single-5850-furmark-crysis-stalker-dirt#t2150795