Bitcoin Forum

I added a 2nd GPU to my rig today :)

When I run cgminer, I receive a GPU 0 failure disabled error message, running bitminter java applet I am getting at least 200Kh/s less on the original 7950 tahiti card, but at least both are running.

GPU1: MSI R7950 series
GPU2: MSI R7870 series
MB: MSI 970A-G46 (MS-7693)
PSU: Cooler Master M850
Windows 7 Enterprise N

Please let me know what info you guys need to help.

TIA


https://i.imgur.com/FRz3gj1.png
https://i.imgur.com/iVpTWD0.png

Edit: added MB and PSU info

Motherboard model?

Edit: Either way your card may not be getting enough power; what's your PSU wattage?
Your MB might not be able to supply full speed for both PCIe slots at the same time as well.

MB: MSI 970A-G46 (MS-7693)
PSU: Cooler Master M850

Bump

I have nothing that even comes close to a solution but I can report the exact same problem. Luckily for me I don't really care as I run intensity 13 on both cards while I run VM's as well. My only advice would be to try raising the core and mem frequency on the 7950 to around 1k/1.5k to match the higher frequency of the 7870. Make sure both cards run the same frequencies, esp if crossfired.

Anybody have any ideas on this?

How do you guys get 3,4,5 or even 6 GPU's running without this issue?

Had the same in my rig, dummy plug into the card solved the problem.

Noticed you're also running different GPU speeds and Mem speeds.  You're 7800 is running at 960Mhz. I normally have mine running around 1150 to get is max performance while maintain stability.

I have the exact same issue with all of my rigs, but have found that it is only with Windows.

As an example here are my rigs with 7970s. Each has Windows 7, afterbuner 3.00x beta 10, catalyst 13.3, + the following:

Rig 1. 2x7970 - ASRock OC Formula, 2500k, 8GB Ram, SSDs, + 1000 watt PSU
Rig 2. 3x7970 - ASRock 970 Extreme4, Sempron, 4GB Ram, SSD, + 1200 watt PSU
Rig 3. 2x7970 - ASRock 970 Extreme4, Sempron, 4GB Ram, SSD, +1100 watt PSU

These rigs have unfortunately had 7970s shuffled between them too many times due to hardware related issues.

Rig 1. For comparison (2x Visiontek)
1090mv / 1080e / 170m = 649.x Mh/s per GPU

Rig 2. With 4x7970 (2x Visiontek, 1x Powercolor, 1x Sapphire)
1169mv / 1110e / 170m = 645.x Mh/s per GPU (Visiontek)
1094mv / 1050e / 170m = 611.x Mh/s per GPU (Visiontek)
1056mv / 980e / 170m = 568.x Mh/s per GPU (Powercolor)
1169mv / 1110e / 170m = 645.x Mh/s per GPU (Sapphire)

Rig 2. With 3x7970 (2x Visiontek + 1x Sapphire)
1169mv / 1110e / 170m = 659.x Mh/s per GPU (Visiontek)
1094mv / 1050e / 170m = 623.x Mh/s per GPU (Visiontek)
1169mv / 1110e / 170m = 659.x Mh/s per GPU (Sapphire)

Rig 2. With 3x7970 (2x Visiontek + 1x Powercolor)
1169mv / 1110e / 170m = 652.x Mh/s per GPU (Sapphire)
1094mv / 1050e / 170m = 619.x Mh/s per GPU (Visiontek)
1056mv / 980e / 170m = 585.x Mh/s per GPU (Powercolor)

Rig 3. With 2x7970 (1x Visiontek + 1x Sapphire)
1094mv / 1050e / 170m = 621.x Mh/s per GPU (Visiontek)
1169mv / 1110e / 170m = 656.x Mh/s per GPU (Sapphire)

-------------------------------------------------------------------------------------------------------------------------

The weird thing about it is that it follows no pattern. The same Sapphire 7970 with two other 7970s in that system running generates 14 Mh/s more than if I add the Powercolor card into the system. But, if I put it in a system with only one other card running it generates -3Mh/s. It is something whacky with Windows and 7 series GPUs.

The first rigs I ever put together were 4x5870 rigs with xubuntu 11.04. Each GPU ran at 980e / 170m and generated 433 Mh/s. I recently had to split some of my rigs due to not having enough risers. Using the same 11.04 cd, catalyst 11.6, and SDK 2.5, two 5870 at the exact same settings till got 433 Mh/s. If that were Windows they'd somehow magically be getting 450 ish Mh/s.

I'd use linux for all of my rigs, but I need to have the ability to undervolt my 7970s to keep them cool as I live in a nuclear inferno. The only thing I can think of is that it has to have something to do with either MSI Afterburner, bios, or Windows power saving feature. At this point I've tried powered risers, moving power supplies aroun, testing voltage input from the wall, different catalyst versions, etc... I wish someone could figure this out as it is extremely annoying.

power saving mode? didnt even consider that <palmface/>

FYI,
For others that may encounter this issue in the future, I plugged my LCD into one and a dvi/hdmi*->vga plug into it and am now getting ~650MH/s between the two of them. The original card is still lower than what it was but at least it's not in the 250 range like it was.


*dvi/hdmi, I can never remember the difference.

I have an older server that I have not used for a while and may just isolate them each in their own box.

I had this exact same problem, on ubuntu. When mining on 1 card it did full hashrate, when plugging in another card both of them were doing like 75-80% hashrate.
Reinstalled linux, upgraded cgminer, tried other cards, nothing changed.

When I updated the ati driver, it magically solved the problem, so maybe you should try it with another driver.

I had a similar problem with my dual HD7850s (OC 1090/1200 stock volts) -- I'm running BFGMiner and dialing up the "intensity" setting to 4 on both cards leveled the cards out at a higher rate than I was getting on the default settings (went from <610 to >630)

I had the same problem. When I put in the second card it was stifling the airflow on the first card and making it run at a much higher temp. Which slowed it down.

Have you noticed the running gpu temp increase on your first gpu since you put the second in?

I fixed my problem by adding in a fan blowing air directly into the top card.

Update, for some reason I thought that maybe moving the GPU's power to a different spot on my PSU would help....

And it has!

Gone from 750MH/s to 900MH/s+

https://i.imgur.com/33Bnpip.jpg