Mining LTC and GPUs keep going to SICK, then DEAD

[Wolf0]

NaN.

[thef]

Are your overclock settings stable? That's what usually causes it.

[ethought]

Turn those overclocking settings down a bit. Try Intensity 19 or even 18..

[lazydna]

I'm mining LTC on Windows 7 with CGMiner 3.2.1. I have three Sapphire HD 7950 Dual-X cards (11196-16-20G), and they all keep going to SICK, then DEAD soon after I start CGMiner. My batch file follows:

Code:

C:\Users\Wolf\Desktop\CGMiner-bin\cgminer.exe --scrypt --gpu-platform 0 --failover-only --url stratum+tcp://us.wemineltc.com:3333 --url stratum+tcp://coinotron.com:3334 --userpass xxx:xxx --userpass xxx:xxx --worksize 256,256,256 --gpu-threads 1 --intensity 20,20,20 --lookup-gap 2,2,2 --thread-concurrency 16384,16384,16384 --gpu-powertune 20,20,20 --gpu-fan 0-100,0-100,0-100

Any help?

Change tc to 24000 power tune and clocks set to default. See if any hardware errors, set fan limit to 85

[Singlebyte]

Lower intensity to 18

[mm8000]

Delete the .bat files in the cgminer folder. Restart computer.

[sunce86]

You overclocked them too much

[mm8000]

And try without the thread-concurrency flag.

Also have you run the setx commands? Those really helped me.

[mm8000]

Delete the .bat files in the cgminer folder. Restart computer.

What are you smoking?

You overclocked them too much

Happens at default clocks.

And try without the thread-concurrency flag.

Also have you run the setx commands? Those really helped me.

Of course I used the setx commands. I'll try without the TC flag, though.

Oops. I mean .bin files.

They should look something like:
scrypt130511Tahitiglg2tc21312w128l4.bin
poclbm130302Tahitigv1w64l4.bin

[mm8000]

Delete the .bat files in the cgminer folder. Restart computer.

What are you smoking?

You overclocked them too much

Happens at default clocks.

And try without the thread-concurrency flag.

Also have you run the setx commands? Those really helped me.

Of course I used the setx commands. I'll try without the TC flag, though.

Oops. I mean .bin files.

They should look something like:
scrypt130511Tahitiglg2tc21312w128l4.bin
poclbm130302Tahitigv1w64l4.bin

Yeah, I know what they look like, and they're actually on my desktop. Tried that.

You running the latest drivers? Does it work on older versions of cgminer?

[Xanthe]

I'm assuming this is your first try mining with 3 cards in this particular system? Have you ever run the system with just one or two cards? I'm kind of wondering if the issue could be power related. Is your PSU a single rail 850W+ from a good manufacturer? Grab a program like GPU-Z (http://www.techpowerup.com/downloads/2244/techpowerup-gpu-z-v0-7-1/) and watch the +12V when you start up cgminer. If it drops down to something like 11.5V or fluctuates wildly in swings of .1V+ then you have some power issues. Normally a power supply would simply shut down if you try to pull too much current from it -- that's not happening in this case. Nonetheless, sometimes PSU manufacturers skimp to save on costs and leave out OCP (over current protection) which can then lead to the kind of symptoms you're experiencing.

It may turn out that your problem is not PSU related at all, but I figured I'd throw this out there since you've tried so many other things without solution.

Good luck

[Xanthe]

Heh, about that... it's a Corsair TX750M. I have to undervolt the cards to 1.1v or the PSU shuts down.

Ohhh, interesting!

The Corsair TX750M is a good unit. It's manufactured by CWT (a decent PSU OEM), I believe . I'm running a TX850v1 and a TX850M myself at the moment. That said, it may still be that you're overtaxing the poor thing.

You're probably right at the point before the OCP kicks in but after the +12V rail falls wildly out of spec. This can certainly cause the issue that you're experiencing (but don't take this as a 100% positive diagnosis). Again, if possible try monitoring your +12V when you start cgminer and see if it's going wild. Also, if it's not too much trouble, change nothing except remove 1 card from the system and try firing up cgminer. If it runs perfectly then I'm going to say it's time to upgrade that PSU.

[blockzett]

Ohhh, interesting!
The Corsair TX750M is a good unit. It's manufactured by CWT (a decent PSU OEM), I believe . I'm running a TX850v1 and a TX850M myself at the moment. That said, it may still be that you're overtaxing the poor thing.

/sign :-( that's possible.

I'm using a XFX 1000W Pro Series 80PLUS Platinum. With a little OC and voltage@1,12v 3 cards are pulling ~750w (power meter) out of the wall while scrypt mining.
This system based on a Sempron 145, 8GB RAM, 2 fans and a SSD.

Larger oc@1,2v = ~810w

I get similar problems if I don't give them enough core voltage. Had one 7950 with bad ASIC quality, it got SICK/DEAD really quick.

Good luck!

[dentldir]

What temps are they running at when they do run?

[Xanthe]

Disabled one GPU, ran CGMiner. Display driver just crashed, taking down the first GPU.
EDIT: And there goes the second display driver crash, taking out the remaining GPU.

Hmmmph. Well, I guess maybe the problem lies elsewhere then. Which drivers are you running? I'm currently running my 3x 7950 system on Catalyst 12.8 (best performance) but also had it running fine on version 13.1, 13.3beta and 13.6beta.

The 11.78V on the +12V is technically still within spec (+-5%), but sounds a bit low. It's obvious your PSU is under some major stress.
As you said it seems this indeed might not be your only problem though.

I'm going to go with blockzett on this one. Maybe 1.1V isn't enough core voltage for your cards. You could try again with only two cards, but raise their voltage up a little and see what happens. If the 1.1V that you state is inside of Sapphire Trixx, keep in mind that your true voltage will be lower. My 7950s have a vdroop of around 60mV which means that when I set them to 1100mV in Trixx, they actually run at 1040mV as measured by GPU-Z.

Other than this I'm a bit stumped now, sorry.

[hermann1983]

Well, adding voltage seems to work, but it kills the hashrate on one of the cards as it throttles once it goes above 100C...

you push it too much. start with default settings and increase speed by bit only when stable for few hours