• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Second GPU Runs at Half Speed.

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Guff

Registered
Joined
Jan 31, 2014
Hey guys,

I have two MSI 7970 Twin FrozrIII's and I can't seem to get them to both hash properly in CGMiner.

It's been three days of tweaking settings and changing drivers and I still can't get both cards to hash at 700-750 at the same time.

So, at first, both cards would get to 750ish and then one card would begin drop off. It would keep dropping until it hit 200-300, about 8-10 minutes later.

So I did a ton of Google searching and tried a million things but nothing really worked.

Eventually I updated to the new beta CCC (14.1?) and now my issue is that when I start it with my bat file, one gpu hashes at 580~ and the other hashes at 250-300. The load is maxed out, and it's running the exact same settings as the other card obviously, but it just refuses to work properly.

My .bat is, "-I 13 -g 2 -w 256 --thread-concurrency 8192"

k6n767y.jpg

^These are my current clock speeds, hash rate, and GPU1 temps.

Stock Clock with a unmodified .bat is getting me 425KH/s on GPU0 and 250 KH/s on GPU1

Stock Clock on a .bat with "-I 13 -g 2 -w 256 --thread-concurrency 8192" (which is what I usually use) is getting 510KH/s on GPU0 and 280KH/s on GPU1.

Same deal, GPU1 is always slower than GPU0.

Temps are fine, GPU1 is 78C and GPU0 is 72C. VRM Temps are 90C and 83C, respectively.

I tried swapping cards, and the card in GPU1's bank is always the slow one. So maybe there's something wrong with the board? I have a powered 1x riser on the way, so I can see if its the 3.0x16 slot acting up or something to that effect...

It can't be the card though because it happens to either of them.

I have also noticed that the slow card is has a jumpy VDDC Current, whereas the other one has a steady one. It jumps 20A from 100A to 120A constantly, which certainly seems like an issue.

HJJdg7x.gif

That would point towards a PSU issue, but I am running two 600w PSUs and it doesn't matter which PSU the card is connected to, it's always the card in the GPU1 bank that has the Voltage fluctuation.

No matter what I try I just can't get the damn cards to behave and just mine at the same speed.

Any suggestions?
 
couple things to know:
- did you max alloc it ? (see getting started if not)
- windows 32 or 64bit ?
- system ram ?
- anything using cpu intensely ? one cpu thread should be dedicated to cgminer
 
couple things to know:
- did you max alloc it ? (see getting started if not)
- windows 32 or 64bit ?
- system ram ?
- anything using cpu intensely ? one cpu thread should be dedicated to cgminer

-Yes
-64 Bit
-8GB ram
-Nothing using the CPU...
 
hmmm.. it may be throttling the upper card, try bumping fans to 100 just for a while, let em both cool down as much as possible and see what happens
 
I fixed it changing script but I don't remember what did I change then. I remember that both my cards were running at 50% at bad script.
 
hmmm.. it may be throttling the upper card, try bumping fans to 100 just for a while, let em both cool down as much as possible and see what happens

The first picture on the OP was with the slow card running at 100% fan for some time. It never makes a difference because the card is slow even on a cold start.

That being said, the 'fast card' should be running faster, so maybe it is throttling...
 
well funny thing is both cards not running is easier to fix, one running full one half is harder, its gotta be something throttling it, should be heat
could have been max alloc, but that was done, same with ram space to play with

btw Guff, id still try setting up a config for cgminer and setting individual settings for each card
easiest way is to hit "S, then W" while cgminer is running
 
well funny thing is both cards not running is easier to fix, one running full one half is harder, its gotta be something throttling it, should be heat
could have been max alloc, but that was done, same with ram space to play with

btw Guff, id still try setting up a config for cgminer and setting individual settings for each card
easiest way is to hit "S, then W" while cgminer is running

I tried setting up a config, but I didn't mess with it too much. I'll make another one when i get home today and see what I can coax out of the cards...
 
may be worth trying diff settings, conf file makes it so much easier + it has more options
 
Guff

do this and report back, on cgminer, press G, then press i, choose GPU 1, then set itensity to 20
 
hed have to set g1 first or he'll just crash, but ya may try a bat with g1 i20 tc 24000 for both cards, worth a try
 
hed have to set g1 first or he'll just crash, but ya may try a bat with g1 i20 tc 24000 for both cards, worth a try

I'm thinking that the 2nd card may be running in g 1

I'd set my .bat this way

-I 13,13 -g 2,2 -w 256,256 --thread-concurrency 24000,24000 -d 0 -d 1
 
doubt it, cgminer doesn't support diff -g settings
btw 24000tc with g2 ? :) that wont quite work
Id say he try g1 tc24000 i20, maybe his cards will simply react better to g1
 
g1, tc 24000, i 20, got me to 660KH/s on GPU0 and 330KH/s on GPU1. Also, it lagged the **** out of the computer and the fans were spiking up and down... So perhaps I 20 isn't the best of ideas?
 
g1, tc 24000, i 20, got me to 660KH/s on GPU0 and 330KH/s on GPU1. Also, it lagged the **** out of the computer and the fans were spiking up and down... So perhaps I 20 isn't the best of ideas?

Might as well try and go back to basics, uninstall the AMD Drivers and the AMD SDK with CCleaner or driver cleaner, then re-install them and see what happens.
 
I might try rolling back to the current CCC (13.12) instead of the beta (14.1). When i was on that, I would at least get 750 in GPU0 on the same .bat...
 
it may after all really be the board .. do this, try single card
 
Well, well. Going back to the old CCC seemed to have done something! I used the " -I 13 -g 2 -w 256 --thread-concurrency 8192" .bat and now...

Both GPUs are running at ~550 KH/s!

SCsoAcQ.jpg

That being said, it's about 200 KH/s lower than i want it to be...

Also, The temps on GPU1 are waaay too high. 102C VRM Temp is not acceptable... That can be fixed perhaps with some undervolting and some risers to get it out of it's place between the CPU and the other GPU... Also, the VVDC is still wonky. GPU0's is fine though...

I'll try out a couple of different settings and check out the Hardware Comparison Wiki to see if I can't find some better settings...


Edit: I forgot to mention, GPU1 starts off at less than 100 KH/s and then I have to do the old "Pull the DVI Cable out and Plug it back in" trick to get it up to speed. Any way I can avoid that? Is it bad in any way?
 
Back