One of three 7950s is running slower?

[aznewsh]

I have three 7950s and two consistently mine together around 520 Mh/s
The other one is always about 70 behind - around 450 Mh/s

This happens when I am mining Bitcoin, if I mine scrypt like Litecoin the all run together at around 630 Kh/s

They all have the same settings and are all the identical card MSI R7950 Twin Frozr.

The one that is mining slow shows it is only using about 80% GPU and has a temp of 74 so not too hot, actually cooler than the others.

Can somebody explain why this is happening?

<Edited to correct typo on card model - Thanks to TheSpiral for helping with that>

[BlazinBeaches]

Is the slow one plugged into a monitor

[TheSpiral]

9750? Not for maybe a year.

[aznewsh]

No - none are plugged into a monitor

[aznewsh]

9750? Not for maybe a year.

7950 - had a brain fart - Am I supposed to give you a cookie or something now?

[TheSpiral]

9750? Not for maybe a year.

7950 - had a brain fart - Am I supposed to give you a cookie or something now?

I'd take one if you have one, sure.

[aznewsh]

Back on topic - anybody got any thoughts on this?

[glendall]

Are you using any pcie risers?

[aznewsh]

Are you using any pcie risers?

Not yet still waiting for them to arrive, interestingly the cards all run at around 638 kh/s when mining scrypt - the difference only happens when mining bitcoin

[SirMintALot]

Is the one running slower than the others older than the others? Maybe removing dust and replacing the thermal grease might help. My Asus HD7950-DC2T-3GD5 (that I bought used) is also only hashing about 440 MHash/s, after removing dust and optimizing airflow. I can't set Powertune over -9 without overheating it in a very short time. Before I removed dust and added extra fans -13 was the best. My last hope is that the thermal grease is the problem. I have some Thermaltake TG-3 ordered, I let you know if it helps.

[glendall]

I have the same problem on one of my rigs. My guess was that it was because I am using unpowered pcie riser as opposed to powered, and my mobo is putting out enough power.

It isn't the card (in my case). It works fine in other systems. But my 3 third card of the same make on my one rig, connected with a PCIE extender cable, only gets about 75% the performance (450 k/hash as opposed to 600 for the other 2) .  7950's & scrypt mining.

I've been a bit too busy to fully troubleshoot it. I think I'm going to save myself the headache and just swap some mobo's around (with spare slots) so I don't have to use any extender cables.

[ThatDGuy]

Are you using any pcie risers?

Not yet still waiting for them to arrive, interestingly the cards all run at around 638 kh/s when mining scrypt - the difference only happens when mining bitcoin

Got me on this one...

Have you tried pulling them out and switching that card to a different position? See if it is the card or the connection?

[camaro69327]

I have seen this weirdness before.

You don't say what you are using for mining or what the settings are.

But..when i have seen this in cgminer after starting a miner up, (one card not performing at 100%) I change the Intensity from say 9 to 7. Then change it right back to what it was (9) and then the card will run at 100%.

[SirMintALot]

Replacing the grease did help a lot. First of all the was to much grease on the chip and it was cheap grease. After replacing it with a small amount of TG-3 the card is now about 11° C cooler (51°C instead of 63°C) but sadly it didn't help much on the hashrate. I can get now 5 MHash/s more out of the card.
The trick with changing the intesity doesn't help with my card.

[aznewsh]

The temperature remains cool on the slower card, actually cooler than the others around 74

I will try swapping slots and messing with intensity though I don't think I set that in my bat file anyway, I think I just let bfgminer or cgminer handle that - I tend to use bfgminer more since cgminer seems to act very flaky for me - sometimes runs fine - other times starts then closes immediately - no idea why.

[aznewsh]

OK so I moved slots, I changed all cards into different ones and each card kept the same cables to the PSU

The issue still remains and it is still on OCL 0

Since the PSU and the GPUs have effectively been taken out of the equation, what does this leave; MB or Software?

Reminder: all cards are identical and new
this issue does not happen when mining scrypt, all cards get 640 kh/s, with sha 256: OCL 0 gets say 440 mh/s whilst OCL 1&2 get 500 mh/s

This is my start up, as you see nothing complex there

bfgminer -o stratum+tcp://stratum.mining.eligius.st:3334/ -O [myusername is here] --temp-target 80 --temp-overheat 85 --temp-cutoff 90 --auto-fan --auto-gpu --gpu-engine 1050 --gpu-memclock 1400

[Charles999]

This is not uncommon as it's happenning to 3 of my rigs.  I have 3 7970 cards in each box and one of the card is always running at about 75% of the other two. 

This also happens to my two rigs with 7990 cards in it.  Each basically has 4 Tahiti in it.  3 runs fine and 1 is slow.  WIERD  all my units have 16GB of ram in it.  Running Windows 7 64 bit with nothing else.

[aznewsh]

I did notice that at start up bfgminer reports:

OCL 0  (Thread 1) being disabled

It does not report this for OCL 1 & OCL 2

I do not have any idea what this means though?

[aznewsh]

Apparently it was the intensity causing the problem; I had it set to dynamic and as soon as I changed them all to the same value 9 - they all ran the same rate

Hope this helps somebody down the road.

Thanks for all input

[Beans]

I have a 7970 that runs slow if the monitor is not plugged in when the miner is started. Not sure why it acts goofy, my other rigs don't do it.