CCminer issues - Hardware stack error & Illegal memory issue.

[neujc]

Hi,

I'm having a few issues regarding my miner, which ive been trying to problem solve for days now by searching online without a solution. I've narrowed it down to 2 certain GPU's that are causing these two issues. At first the miner was running fine for the first 4 days without any issues using ccminer alexis 1.0 and the newest ccminer by tpruvot 2.2.4.

Before I continue my specs are:

MOBO: Asus h270-Prime PLUS
GPU: 4 X 1070 ti EVGA SC
CPU: Core i3 - 6100
RAM: 4GB Kingston
PSU: Corsair HX1200
OS: Windows 10 home
Software: CCminer 2.2.4 & CCminer Alexis 1.0
Batch file script: ccminer.exe -a x17 -i 18 -o stratum+tcp://yiimp.eu:3737 -u 'wallet address'


ISSUE ONE - AN ILLEGAL MEMORY ACCESS WAS ENCOUNTED:

Out of nowhere I started getting an error from one of the 1070ti cards which was;

GPU#3: An illegal memory access was encountered
'Cuda error in func 'quark_blake512_cpu_setBlock_80' at line 547 : an illegal memory access was encountered.

At first I thought it was an OC issue, so I put everything back to stock including the ccminer intensity down to 18 and am still receiving the issue.
After reinstalling drivers (CUDA 9 and 7.5, NVidia 388.13 and the newest 390 version), changing ccminer's version, expanding the virtual memory to 20GB's, changing PCI slots, cables and extender boards I came to no solutiuon.  I started removing cards and mining with one at a time and narrowed it down to one card causing the problem, as soon as I remove the card the issue is gone and everything works fine (keep in mind its not a PSU power issue as when the card is the only card in the rig it has the memory issue regardless). So I then put the card into my gaming pc and still the same issue with the ccminer (both versions)

ISSUE TWO - HARDWARE STACK ERROR

After a few days running the rig again, after I removed the card that was causing the memory issue another card started playing up... getting an error;

GPU#0: hardware stack error
'Cuda error in func 'quark_blake512_cpu_setBlock_80_sp' at line 577 : hardware stack error.

Again I did the above to try and resolve it but no solution and narrowed it down to another particular 1070ti causing the problem. I'm out of options. My rig is not overclocked and running on stock with 18-21 intensity in ccminer.

Any suggestions at all? as I'm all out of ideas.

Thanks!

[1715005460]

1715005460

[1715005460]

1715005460

[1715005460]

1715005460

[ylpkm2]

Could be issues with the memory on the card. I have 2 1080s that can only use 4 algorithms without crashing. Gonna fix them soon, but so far it appears to be a card problem.
Using a different pc results in the same issues, illegal memory no matter the clocks.

[neujc]

Could be issues with the memory on the card. I have 2 1080s that can only use 4 algorithms without crashing. Gonna fix them soon, but so far it appears to be a card problem.
Using a different pc results in the same issues, illegal memory no matter the clocks.

That's what I was thinking with my two cards.. which are brand new. Maybe something is stuffing them up cause it seems to be a big coincidence that two brand new cards have failed after working days apart, possibly a faulty motherboard?

[gfigg42]

I think x17 is bugged on ccminer. Atleast on the windows version I tried. All of them encountered this error with x17 but work fine with any other algo I've tried. Try the linux version if you can.

[ivakar]

Try to swap rizers between working card and a bad one, if bad one will start to work without errors and good one will start working with error, then you found the source of the problem, it is bad rizer
If not, try to see how many rizers sits on one sata cable from your psu, it shoud not be more then 2 cards on one line.
If all of this is not help, guess some cards are bad