Gridcoin Mining: Generating heat and computational errors
Hi everyone,
After many failed attempts to start crunching data with my HD7990 on Ubuntu 16.04 and 14.04 (details here and here), I decided to make another step back and downgrade my system to Ubuntu 12.04. I successfully install the fglrx driver, and the Boinc client recognized the GPU with no effort.
After troubles with the "No CPU" configuration on the Gridcoin Pool (the pool is sending a XML file with a tag not recognized by the old version of the Boinc client), I manage to synchronized the pool with my unit, and the client started to download work packages for the CPU and the GPU.
I configured the Gridcoin Pool to send work for two projects that are only using the CPU (World Community Grid and TN-Grid Platform) and two projects that are supporting the ATI GPUs (PrimeGrid and Milkyway@Home). Unfortunate, the GPU work units ended up with massive number of computational errors.
Yes, it is generating heat, but the data processed is useless:
Curious is that for the PrimeGrid project, not all work units processed by the GPU are wrong, but only the ones using the Genefer application.
I will keep the current configuration few more days to experiment with different projects, but I plan to move back to Ubuntu 16.04 and try with different open source drivers (thank you @tomasbrod, for the hint regarding the opencl-mesa driver and the open source radeon driver!).
Just want to say , I run 1.2gW worth of systems ( 2 tablets + 4 phones + desktops + blades ) and part of development for the project admins and coders if invalid/corrupt WU's from errors while processing and or downloading etc. This is key in making things work wide spread such as new/old drivers and new/old kernels and kernel/module flags :) . Rosetta hates 1 quad core tablet and 1 oct core phone but loves the rest in my setup. This is a totally normal thing , but can also be caused by heat on the the south bridge and voltage issues to the GPU not just Linux core system and modules.
You should check stderr.txt files for the failing tasks, maybe it would give us some clues. Critical errors are usually logged at the end of the file.
Presumably the MilkyWay errors are part of the server problem (now fixed) they had last week.
Nah, they wouldn't error out so quickly then. Server problem last week caused many tasks to be discarded during the validation process, not after just 2 secs of crunching. Not to mention that Primegrid GFN tasks are also erroring out.
Thanks for mentioning me!
Genefer does not run for me under opencl-mesa either.
Under opecl-mesa, from primegrid only pps_sieve of would run.
Milkyway worked only under opencl-amd (which is userspace part of closed amdgpupro, with radeon opensource kernel module).
I have not tested Primegrid on opencl-mesa-git nor on opencl-amd.
I never succeeded in installing fglrx nor full amdgpupro drivers.
Thank you for the valuable inputs.
img credz: pixabay.com
Nice, you got a 87.0% @gogogadgetupvote upgoat, thanks to @steemmillionaire
Want a boost? Minnowbooster's got your back!
The @OriginalWorks bot has determined this post by @steemmillionaire to be original material and upvoted it!
To call @OriginalWorks, simply reply to any post with @originalworks or !originalworks in your message!
For more information, Click Here!
Special thanks to @reggaemuffin for being a supporter! Vote him as a witness to help make Steemit a better place!