[LTC] HW errors from PCI-e slot with cgminer

[ironstove]

I was running 2 7950s on my mining rig and recently added a 3rd one on... However when I run all 3 at the same settings that the 2 were working on before, the GPU that I put into the last PCI-e slot keeps giving me HW errors, when I reduce the intensity down to 13, I no longer get HW errors...

It seems to specifically be whichever GPU is connected to the specific PCI-e slot as switching around the GPUs makes whichever GPU was swapped into the PCI-e have HW issues as well, so I know it isn't the GPU.

I have 2 other mining rigs that have identical hardware and they do not have this issue, they are all able to run 3 GPUs at the same settings just fine.

I have tried reinstalling drivers and I'm having some minor issues, I can reinstall catalyst 13.1, but APP SDK 2.8 keeps failing during installation, not sure if this is related.

I have already set the system vars for gpu_max_alloc, but I did it again to be sure...

I guess for the time being I'm going to have to run the GPU @ 13 intensity as I am leaving town for a week, when I get back I'll try to resolve the issue, and update this topic when I make a discovery.

(So far, it seems to me that it's a software issue, maybe has to do with the operating system and drivers.... I previously had the same issue of getting HW errors @ high intensity when I was first setting up my cards a long time ago, and setting the GPU_MAX_ALLOC_PERCENT corrected this issue. It seems almost as if the command is not working, or maybe the computer isn't reading enough system RAM? I am on windows 7 x64 btw, and I have 6 GB of system RAM, and as I said there are 2 other rigs set up exactly the same but they are able to run all 3 GPUs on high intensity without any problems).

[megablue]

I had similar issues... After a lot of trial and errors i finally figured out what caused it.

Dont run all your GPUs under one instance of cgminer.

Run separated instances instead, let each instance control a single GPU.
--gpu-reorder --device 0 --remove-disabled
--gpu-reorder --device 1 --remove-disabled
--gpu-reorder --device 2 --remove-disabled

Let me know if this works
Donation will be appreciated if you find my reply is helpful

[ironstove]

Hmm interesting. I'll give it a shot when I get the chance I'm at the airport ATM haha.

This doesn't explain why I can run 3 GPUs on my other miners though. All miners have identical hardware.. So it has to be something off with software, if your method works I'll prob still end up reformatting and starting from scratch.

[digitalindustry]

Hmm interesting. I'll give it a shot when I get the chance I'm at the airport ATM haha.

This doesn't explain why I can run 3 GPUs on my other miners though. All miners have identical hardware.. So it has to be something off with software, if your method works I'll prob still end up reformatting and starting from scratch.

CPU?

Motherboard?

what are they?

cards are in 1x? 16? 8x?

 

[ironstove]

Hmm interesting. I'll give it a shot when I get the chance I'm at the airport ATM haha.

This doesn't explain why I can run 3 GPUs on my other miners though. All miners have identical hardware.. So it has to be something off with software, if your method works I'll prob still end up reformatting and starting from scratch.

CPU?

Motherboard?

what are they?

cards are in 1x? 16? 8x?

 

CPU: Sempron 145
Mobo: GIGABYTE-970-DS3
All cards are in the x1 through risers.

[IrishTR]

I am having the exact same issue.  I can run 3 cards full bore no problem, add in the 4th card and it gets nothing but hardware errors or I have to lower the intensity from 18 to 13 and remove any overclocking on it, while the other 3 maintain overclock and 18 intensity just fine.


Here is the setup.

ASRock 970 Extreme 4 Current Bios v2.60
8GB DDR3 1600mhz RAM (2x4)
AMD Sempron 145 stock no OC/Core Unlock (tried and it didnt happen easily so didnt bother)
(4) 6950 PowerColor 2GB
Running AMD Catalyst 13.11 Beta9.5 (forget off top of head but its the most current Beta Drivers posted as of yesterday)
PSU is (1) 850W for 3 cards and (1) 600W for the 4th card (Even tried on the one 850W since it has 8 PCIe 8pin power connectors but split power to rule out it as the problem).

As far as connection goes, I have 3 of them solid on 16x non powered risers in the 16x slots.  They each get 485khash per card.  When I add the 4th card via a 1x to 16x Powered Riser (tried both with and without power connector) same results if I get into Windows and am able to fire up CGMiner, with all 4 on same CGMiner the card will get HWs and very low hashrates like double digit low or start at 200s and drop down to eventually get called SICK.

And if I tried to just run the 1x card by itself in a solo CGMiner session it gets the same results (no other cards running).  I'm just so frustrated on what to do as I have a card that is basically sitting useless.

I also tried to rule out the Riser as the problem and plugged it into a 16x slot and it worked fine.  So that tells me its not the video card (as I changed them around as well) and the riser is working.  I checked the BIOS there is nothing in bios in relation to the PCIE slots other than the voltage setting and spread spectrum.

Really at my whits end with trying to make this work when everyone/where else I see posted with practically the exact same spec albeit 7950s for video cards working just fine. HELP please