Unstable rig when mining x17

[dragonmike]

Was wondering if other people were experiencing similar sorts of issues.

When mining equihash using DSTM (or even EWBF), neoscrypt, lyra2v2, lyra2z, Nist5 using various versions of ccminer, my rig of 1080's is perfectly stable.

When I mine x17, it just keeps crashing. I've tried ccminer Alexis78 and ccminer-Cuda9. At the beginning they'd just keel over after a few minutes.

The miner displays an error message: GPU #0: an illegal memory access was encountered... and then crashes.

I've reduced overclock on memory, no change. Reduced core speed, no change. Reduced intensity, no change. Only by setting "force P2 state" to OFF in nvidiaprofileinspector was I able to achieve more stability. The miner will still crash several times a day nonetheless.

Temperatures are all fine (GPU's in mid-60's), PSUs have lots of overhead margin.

If anyone has any idea, I'd be all ears.

[joshuaj]

Maybe you hae to lower down the intensity, some algorithm cannot stand with high intensity.

[dragonmike]

I've lowered intensity by 1.5 but it crashes just as much.

I'm out of ideas, tbh. Ccminer just throwing out this error and crashing. Pfff...

[darkneorus]

lower the intensity. something like 18-19 will be fine for x17.

[helladave]

im using this alexis build on x17 with 2 1080ti at 21.5 intensity stable multiple days in a row now... before other miners were not for me.. good luck
getting this hashrate off the the 2 cards....

1BTC =  15207.87 USD


   Type: NVIDIA

Miner         Algorithm       Speed BTC/Day BTC/GH/Day USD/Day Pool
-----         ---------       ----- ------- ---------- ------- ----
Ccmineralexis X17        32.97 MH/s 0.00463    0.14047  70.440 ahashpool-Unknown


https://github.com/nemosminer/ccminer-Alexis78/releases/download/ccminer-alexis78/ccminer-alexis78-ms2013-cuda7.5.7z

[Kompik]

First lower your overclock, especially the core. X17 does not stand high overclocks. Lower your intensity to 17.5-19. Cap your power limit.

Then make sure that your swap memory is high enough. For 6 cards at least 32GB.

[cashen]

I have a version of 1080TIs which will not run X17s for anything.

I gave up.

[VyprBTC]

I have a version of 1080TIs which will not run X17s for anything.

I gave up.

Oh yeah, which version is that, because I pretty much have every single model of 1080 Ti. Works just fine.

[dragonmike]

First lower your overclock, especially the core. X17 does not stand high overclocks. Lower your intensity to 17.5-19. Cap your power limit.

Then make sure that your swap memory is high enough. For 6 cards at least 32GB.

Overclocks lowered substantially, intensity lowered to 18. Still crashing.
I've now increased page file size from 32 to 64 GB.

This thing better work now, it's getting ridiculous...

[dragonmike]

For the records I'm on Cuda9, and latest driver installed a couple of days ago.
I'd always be interested to have more granular info on error messages. These "illegal memory access" crashes are just stupid to troubleshoot...

[dragonmike]

Well, after another 24 hours of testing I can safely say the damn thing still crashes all the time. Looking at the benches posted on zpool I just cannot believe these top hashrates were sustained for more than a few minutes. I've kinda given up on x17 now, can't spend all day checking that one rig all the time. Waste of time (and hash).