Posts by [A.S.a.M.]koschi
|
1)
Message boards :
Number crunching :
Linux OpenCL Support ?
(Message 2022468)
Posted 9 Dec 2019 by  [A.S.a.M.]koschi
Post: Not all projects work with ROCm, eg. Milkyway fails straight away with CL errors as well. Einstein and others work, though I found ROCm OpenCL is still a bit slower than the AMDGPU-PRO OpenCL. I'm running Ubuntu 19.04, which is also unsupported by the official package. I created http://kerbodyne.com/boinc/create_amdgpu-ocl-pkg.sh (based on an Arch build script I found) to extract the OpenCL libs from the official driver archive and build a package that just installs these few files. Maybe it can serve as a reference for your own Slack version, so you can finally get the "real driver" working on your machine. |
|
2)
Message boards :
Number crunching :
Developing AMD GPU Utilities
(Message 1998955)
Posted 20 Jun 2019 by [A.S.a.M.]koschi
Post: Hi Rick, today we configured an RX580 to BOINC on Linux Mint 19.1. clinfo works, BOINC utilizes the GPU, however, amdgpu-utils don't find the card. amdgpu-monitor showed "No AMD GPUs detected, exiting..." num_amd_gpus must have been 0 there. I couldn't fully understand how gpu_list.num_gpus() is defined to check what it was looking for. Being Linux Mint, the official driver doesn't install, so I used the extracted OpenCL parts from AMDGPU-PRO 19.10. I'm doing that on Ubuntu 19.04 as well, there the tools work without problems. What are you checking in that moment, where does that data come from? Thanks! |
|
3)
Message boards :
Number crunching :
Developing AMD GPU Utilities
(Message 1985748)
Posted 18 Mar 2019 by [A.S.a.M.]koschi
Post: Thanks, the ls and monitor now work for root and my user. pac works for my user, but segfaults for root: root@host:~/amdgpu-utils# ./amdgpu-pac Unable to init server: Could not connect: Connection refused Unable to init server: Could not connect: Connection refused AMD Wattman features enabled: 0xffff7fff Command '['dpkg', '-l', 'amdgpu']' returned non-zero exit status 1. Warning: amdgpu drivers not may not be installed. 1 AMD GPUs detected, 1 may be compatible, checking... 1 are confirmed compatible. (amdgpu-pac:4395): Gtk-CRITICAL **: 12:36:00.559: _gtk_style_provider_private_get_settings: assertion 'GTK_IS_STYLE_PROVIDER_PRIVATE (provider)' failed (amdgpu-pac:4395): Gtk-CRITICAL **: 12:36:00.559: _gtk_style_provider_private_get_settings: assertion 'GTK_IS_STYLE_PROVIDER_PRIVATE (provider)' failed (amdgpu-pac:4395): Gtk-CRITICAL **: 12:36:00.559: _gtk_style_provider_private_get_settings: assertion 'GTK_IS_STYLE_PROVIDER_PRIVATE (provider)' failed Segmentation fault (core dumped) No big deal though :-) |
|
4)
Message boards :
Number crunching :
Developing AMD GPU Utilities
(Message 1985744)
Posted 18 Mar 2019 by [A.S.a.M.]koschi
Post: Thanks, the message disappeared, now it complains about amdgpu package missing: user@host:~/amdgpu-utils# ./amdgpu-ls AMD Wattman features enabled: 0xffff7fff Command '['dpkg', '-l', 'amdgpu']' returned non-zero exit status 1. Error: amdgpu drivers not installed, exiting... Inspired by an Arch Linux package build script, I extracted the OpenCL libraries from the 18.50 driver, then packaged these into a DEB file that can be easily installed/removed. https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=opencl-amd It works, WUs that got downloaded with Mesa/Clover and ROCm had shown "coproc missing" after I removed those again. Then installing the package/libraries, BOINC restart and it worked again. The card shows up just like it would with full AMDGPU-PRO drivers. Sun 17 Mar 2019 16:14:07 CET | | OpenCL: AMD/ATI GPU 0: Radeon RX 580 Series (driver version 2766.4, device version OpenCL 1.2 AMD-APP (2766.4), 8169MB, 8169MB available, 5161 GFLOPS peak) Wattman interface is accessible, so its just that check against the amdgpu package that is blocking. For people not seeking OpenCL support, they won't need the official driver, the mainline amdgpu is sufficient. Maybe that check can be dropped, or configured to be overwritten with a CLI flag. Thanks! |
|
5)
Message boards :
Number crunching :
Developing AMD GPU Utilities
(Message 1985695)
Posted 17 Mar 2019 by [A.S.a.M.]koschi
Post: Hi Rick, thanks for creating these tools, very useful so far! I tried Ubuntu 19.04 this weekend, it ships kernel 5.0. This seems to be problematic, at least when it comes to the amdgpu-utils: root@host:~# ./amdgpu-utils/amdgpu-ls Using Linux Kernel 5.0.0-7-generic but benchMT requires > 4.17. Error in environment. Exiting... root@host:~# ./amdgpu-utils/amdgpu-monitor Unable to init server: Could not connect: Connection refused Unable to init server: Could not connect: Connection refused Using Linux Kernel 5.0.0-7-generic but benchMT requires > 4.17. Error in environment. Exiting... Most likely easy to fix though :-) Thanks! |
|
6)
Message boards :
Number crunching :
From FX to Ryzen
(Message 1856006)
Posted 17 Mar 2017 by [A.S.a.M.]koschi
Post: Which one is the currently recommended optimized version for MB on Linux 64bit, http://lunatics.kwsn.info/index.php?action=downloads;cat=48 ? Will start testing earliest tomorrow evening. Currently I'm testing long running Einstein (Gravitational Wave search) WUs with 8 and 16 threads. 3.6GHz with the Wraith Spire is realistic. My R7 (8y old Enermax Pro82+ Bronze & 32GB DDR4) manages that at 1.1625V, consuming 120W with Einstein, 145W with WCG or mprime. The Spire should be able to cool that. It had no problem with 3.2GHz and way to high auto voltage of 1.35V. |
|
7)
Message boards :
Number crunching :
From FX to Ryzen
(Message 1855841)
Posted 16 Mar 2017 by [A.S.a.M.]koschi
Post: Which one is the currently recommended optimized version for MB on Linux 64bit, http://lunatics.kwsn.info/index.php?action=downloads;cat=48 ? Haven't done much SETI lately, but am right now trying to get an overview with which Intel CPUs my R7 @ 3.6GHz roughly compares with. Thanks! |
|
8)
Message boards :
News :
SETI@home 8 released for Raspberry Pi.
(Message 1778374)
Posted 12 Apr 2016 by [A.S.a.M.]koschi
Post: Yes that's all that was needed... |
|
9)
Message boards :
News :
SETI@home 8 released for Raspberry Pi.
(Message 1778098)
Posted 11 Apr 2016 by [A.S.a.M.]koschi
Post: The 64bit Ubuntu 16.04 offered by Hardkernel... |
|
10)
Message boards :
Number crunching :
Raspberry Pi results
(Message 1778095)
Posted 11 Apr 2016 by [A.S.a.M.]koschi
Post: 103.000-126.000 seconds on the Raspberry Pi 3 at 1.2GHz https://setiathome.berkeley.edu/results.php?hostid=7974154 78.000-80.000 seconds on the Odroid C2 at 2GHz https://setiathome.berkeley.edu/results.php?hostid=7974151 |
|
11)
Message boards :
News :
SETI@home 8 released for Raspberry Pi.
(Message 1777932)
Posted 10 Apr 2016 by [A.S.a.M.]koschi
Post: Odroid C2 (4x2GHz) 78000-80000 seconds Raspberry Pi 3 (4x1.2Ghz) ~105000 seconds The app creates fftw wisdom files in the slot directories (not coming from the project directory). Are these created when the app starts, through built in routines that mimic fftwf-wisdom wisdom file generation? Since they differ between hosts, they should be already "self-optimized", right? |
|
12)
Message boards :
News :
SETI@home 8 released for Raspberry Pi.
(Message 1777502)
Posted 9 Apr 2016 by [A.S.a.M.]koschi
Post: Thanks |
|
13)
Message boards :
News :
SETI@home 8 released for Raspberry Pi.
(Message 1777484)
Posted 9 Apr 2016 by [A.S.a.M.]koschi
Post: Does the app support/use Neon? |
|
14)
Message boards :
Number crunching :
What CPU will this board handle?
(Message 770036)
Posted 18 Jun 2008 by [A.S.a.M.]koschi
Post: If it recommends a 356W psu, please buy a 380/400W model and you are fine :) That page doesn't sum the consumption values and lets the calculation of efficiency and stuff to you. This is actually a value you should rely on. Why I'm I so sure about this? Simply because I'm powering an Intel C2Q 6600@3,2GHz, 2GB of overclocked PC6400, Geforce 7300, one hard disk and one DVD-RW with a simple Seasonic SS-400ET 400W psu. The system will even run @3.6GHz on cold winter days, but that is a different story, my cooling can hardly handle those 1.42v :-( So please buy a good quality PSU from a major brand and you'll save money and nerves :) edit: That calculator recommends me a 396W power supply. Everything else would be a waste of energy... |
|
15)
Message boards :
Number crunching :
Linux port of Alex v8 code
(Message 764406)
Posted 7 Jun 2008 by [A.S.a.M.]koschi
Post: Hi everyone :) It was only today, that I stumbled upon the V8 release for Linux. After V8 was released for Windows, I set up 32bit WinXP in a kvm (kernel based virtual machine) on 64bit Kubuntu. The system is powered by a Q6600 @ 3,2GHz (8*400) The crunch times were awesome compared to the old 2.4 under Linux: ~ 2600s for a 50.xx WU ~ 3370s for a 7x.xx WU before the 73 credit WUs took 5600s, the 53 credit ones 5200 seconds. Now that V8 is running natively on the system, times are the following: ~ 2630s for a 50.xx WU ~ 3160s for a 7x.xx WU So the crunching time for the big WUs is 4,5min lower, but the middle range WUs times stay on the same level. While testing the virtualization in January, I found out that it was 3% slower running Spinhenge@home than natively under Windows. When I subtract these 3% from 2600 seconds, that would give 2522s, which means that the middle credit WUs are faster under Windows. Doing the same on the 70 credit units would result in 3270 seconds, which is still 2 minutes longer than under Linux. So altogether, both versions have their advantages and also disadvantages. As we can't choose which work units the server will give us, I think both versions would result in the same RAC, when run on the same machine. I was also amazed how the V8 SSSE3 app performed on my new build C2D E7200 @ 3,45GHz. I expected it to be only little faster (as it has smaller caches) than the Q6600 @ 3,2GHz, but thats not just a little bit, it's a very nice gain :-D ~2800s for ~7x.xx WUs ~2350s for a 5x.00 WU Thanks to every one who was involved in this release! You did good work. regards from Brno, Czech Republic |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.