Welcome to MilkyWay@home

Posts by koschi

1) Message boards : Number crunching : GPU upgrade time (Message 70259)
Posted 22 Dec 2020 by koschi
Post:
Well, driver 20.45 computes Einstein FGRP1G WUs within the usual run times, but fails entirely at Milkyway.
I have no problem with dkms or leftovers of eg. 20.40.
2) Message boards : Number crunching : GPU upgrade time (Message 70257)
Posted 21 Dec 2020 by koschi
Post:
For Linux users that want to crunch Milkyway "Big Navi" is problematic.
The 20.45 driver, which is the first to officially support Radeon 6800/6900(XT,) is no longer using the PAL OpenCL implementation for Vega and newer cards.
20.45 switched to ROCm as OpenCL backend, for Vega 10/20, Navi and Big Navi.

My Vega 56 crunched Milkyway on 20.40 just fine, but on 20.45 fails with OpenCL errors:
Error creating command queue (-6): CL_OUT_OF_HOST_MEMORY
Error getting device and context (-6): CL_OUT_OF_HOST_MEMORY


I tried Milkyway on a 6900XT, same issue. When I tested the free Mesa/Clover and also ROCm OpenCL implementations last year, the same error occurred on this project.

So with Big Navi becoming more wide spread over coming years and people updating GPU drivers on their Vegas and Radeon VIIs, more people bump will into this issue.
3) Message boards : Number crunching : New App status (Message 6082)
Posted 12 Nov 2008 by koschi
Post:
*psst* *psst* (avoicefromtheshadows)

From my sources (that don't want their names made public) I can tell you that the performance tuning of the new app is finished, but it's not ready to be released.
Right now two major items are not finished yet:
1. native support for GPU computing on the soon-to-be-released Bitboys Avalanche 3D chips
2. port to GNU Hurd (upcoming release in 2009)

Development is progressing fast, though some issues are still open. QA testing should start in december. Expect the final release somewhere in january or february. I was told there is some nice easteregg in the app, but can't say more, sorry...
4) Message boards : Application Code Discussion : Possible Bug in stPbxConvolved (Message 5418)
Posted 9 Oct 2008 by koschi
Post:
Then international guys come and complain, hey, don't f**k up international statistics, switch to the country you come from ;-)

Guys, check the profile of Milksop at try and check out what boxes they are running and how big the improvements by some code clean up actually is!
It is up to 75* faster than the projects standard app...

Since I learned to know how inefficient the current official app is, I removed my C2D and C2Q from the project.
Just an old AthlonXP remains to complete the 100.000 Credits, than it will also be removed if there is no progress from the project leaders side...

edit:

I was way to slow...
5) Message boards : Number crunching : Compute errors (Message 4666)
Posted 10 Aug 2008 by koschi
Post:
Hello everyone!

I attached to Milkyway just one hour ago and commanded my quad core first too do some work, but the 8 units it received resulted in compute errors immediately.
Those broke after 0 seconds :-(
http://milkyway.cs.rpi.edu/milkyway/results.php?hostid=25394

Then I also attached a small C2D and it runs fine, at least up to now there was no error ;-)

The Quad (Q6600) is running Kubuntu 8.04 with some additions from intrepid, say kernel 2.6.26 and some other stuff...
The dualcore (E7200) is a Debian testing plus some packages from unstable.

Any idea what might cause these immediate errors on the quad? Maybe some library issue or like that?

This host is crunching QMC, SETI v8, Einstein SSE2 and a lot more projects without any problems, so I can't imagine that this is a hardware issue...

Every idea will be appreciated :-)

edit:

after a little reading here in the forums I found crunchers staticly linced version for Suse and maybe others, but the site is not reachable anymore, can some one please send me the file?!? That could be very helpfull maybe =)
mail adress would be m-a_r_t-i-n_.k-o-s-_c-_h_m-i-d-er@_w-_e-b.de (please remove the - and _ :) Thanks a lot in advance!




©2024 Astroinformatics Group