PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
1) Message boards : Number crunching : Cannot run GFN/GPU Tasks on my AMD R9 290x Graphics Cards (Message 91076)
Posted 1256 days ago by Profile Crystal PelletProject donor
Lost my way, any help would be great.

Is BOINC running as a service on that machine?
2) Message boards : Generalized Fermat Prime Search : GFN Status by n-Range (17-22) (Message 90521)
Posted 1279 days ago by Profile Crystal PelletProject donor
N=15 (32768) Has been switch from OCL3 to OCL2.

Surprisingly, the loss of speed was smaller than I was expecting. My tasks went from about 65 seconds to about 108 seconds.

No more Genefer 32768 v3.11 (atiGFN15) on my AMD Radeon HD 7770 anymore:

v3.10 OCL3: 180 seconds wall time (55 seconds CPU)

v3.11 OCL2: 864 seconds wall time (80 seconds CPU)

864 seconds was a fast one. The second task needed 910 seconds.
3) Message boards : Number crunching : LLR checkpoint frequency (Message 89914)
Posted 1306 days ago by Profile Crystal PelletProject donor
From looking at the code (albeit 3.8.8 which is pretty recent but not the latest and greatest) the coded-in default is 30 minutes. However, it looks like this can be overridden in the llr.ini file. Those .ini files in my current (again, not the most up-to-date) boinc directories reset the value to 10 minutes. Look for llr.ini.*, and look at the value for "DiskWriteTime" in them. Your prpnet instances may not include that value, so it would fall back to 30 minutes in those cases.

--Gary

Not looking at the code, but what's happening, it seems the '10' in llr.ini seems to be 10 seconds and not 10 minutes.
4) Message boards : Problems and Help : app_config help needed (Message 69497)
Posted 2088 days ago by Profile Crystal PelletProject donor
Unfortunately, neither of both helps. It keeps downloading 8 SOB units. Plus when all TRPsieve units run out, it won't download any unless you abort 4 waiting SOB units.

Unresolvable, huh? :(

The only action you can take, is when you have 4 SoB's running, adjust PrimeGrid's preference here and deselect SoB until the 4 are almost ready.
5) Message boards : Number crunching : Does GPU usage have a throttle? (Message 68708)
Posted 2125 days ago by Profile Crystal PelletProject donor
eFMer TThrottle
6) Message boards : Generalized Fermat Prime Search : Source code of Genefer for OpenCL is available. (Message 68695)
Posted 2126 days ago by Profile Crystal PelletProject donor
Do other people see this or is this just my host?

I set the priority of the process geneferocl-windows.exe to lowest and keep all 8 threads 100% busy with TRP_sieve (so total 9 BOINC tasks running).
I don't have such a fast card, so no results ready, but after 13 hours running the cpu usage is only 4.1% and that low usage has no negative influence
on the total elapsed time of the task. Task manager shows most of the time 0%.
7) Message boards : Generalized Fermat Prime Search : Source code of Genefer for OpenCL is available. (Message 68684)
Posted 2127 days ago by Profile Crystal PelletProject donor
That's why I asked Husu* to try to run it with the priority set manually to idle priority. If it doesn't affect performance, we could let GeneferOCL, when running on Nvidia, run at low priority instead of normal priority.

Michael, are you also interested in the performance running a WU with idle (Windows -15) priority on a ATI-card with all 8 CPU-threads 100% busy?
8) Message boards : Generalized Fermat Prime Search : Source code of Genefer for OpenCL is available. (Message 68628)
Posted 2129 days ago by Profile Crystal PelletProject donor
Each WU (3.1.2-5 64bit) needs about 3.6 Gig of memory. And the speed of the memory counts!

The 3.1.2-6 solves this problem with ATI driver.

I started a real WU on this machine with the newest geneferocl 3.1.2-6 (Windows 32-bit OpenCL)

It seems geneferocl-windows-x64 was a tiny bit faster, but this version allocates only 168MB Virtual memory.
Extrapolating the elapsed time of the current running WU with % progress done so far the total elapsed time will be 95,666 seconds om my ATI 7770.
9) Message boards : Generalized Fermat Prime Search : Source code of Genefer for OpenCL is available. (Message 68607)
Posted 2130 days ago by Profile Crystal PelletProject donor
Could you check the memory allocated by the application when a large number is tested?

Thanks, Yves

Running a task with geneferocl 3.1.2-5 (Windows 64-bit OpenCL). Estimated run time 26.2 hours.
Seems to be a bit faster than geneferocl 3.1.2-1 (Windows 64-bit OpenGL) That task elapsed time was 27 hours.
I do not reserve a core for the GPU anymore; seems to make no difference.
Memory usage:
10) Message boards : Generalized Fermat Prime Search : Source code of Genefer for OpenCL is available. (Message 68587)
Posted 2130 days ago by Profile Crystal PelletProject donor
Don't know how to interpret, but the developers should know whether this output is OK from my ATI 7770:

geneferocl 3.1.2-5 (Windows 64-bit OpenCL)
Copyright 2001-2013, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2013, Iain Bethune, Michael Goetz, Ronald Schneider

Command line: geneferocl-windows-x64.exe -b


Running on platform 'AMD Accelerated Parallel Processing', device 'Capeverde', version 'OpenCL 1.2 AMD-APP (1084.4)' and driver '1084.4 (VM)'.

Generalized Fermat Number Bench
2199064^8192+1 Time: 118 us/mul. Err: 0.2344 51956 digits
1798620^16384+1 Time: 119 us/mul. Err: 0.2266 102481 digits
1471094^32768+1 Time: 144 us/mul. Err: 0.2344 202102 digits
1203210^65536+1 Time: 264 us/mul. Err: 0.2656 398482 digits
984108^131072+1 Time: 522 us/mul. Err: 0.2188 785521 digits
804904^262144+1 Time: 1.24 ms/mul. Err: 0.2188 1548156 digits
658332^524288+1 Time: 2.35 ms/mul. Err: 0.2188 3050541 digits
538452^1048576+1 Time: 5.13 ms/mul. Err: 0.2266 6009544 digits
440400^2097152+1 Time: 10.3 ms/mul. Err: 0.2266 11836006 digits
360204^4194304+1 Time: 21.5 ms/mul. Err: 0.2031 23305854 digits
294612^8388608+1 Time: 46.3 ms/mul. Err: 0.1895 45879398 digits
Genefer Mark = 19.


Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2019 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 2.89, 3.01, 2.86
Generated 19 Jun 2019 | 21:41:28 UTC