| Author |
Message |
|---|
 Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
                               
|
|
Discuss the new GeneferOCL app here.
(The old discussion about the development of GeneferOCL can be found here.)
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Is there a list of suitable ATI GPU models, Please, or at least tell me if my HD5450 can run it?
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
Is there a list of suitable ATI GPU models, Please, or at least tell me if my HD5450 can run it?
The thread right next to this one is List of Compatible ATI/AMD GPUs.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Help! :0
Stderr output
<core_client_version>7.0.27</core_client_version>
<![CDATA[
<message>
process exited with code 1 (0x1, -255)
</message>
<stderr_txt>
geneferocl 3.1.2-7 (Linux 32-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: ../../projects/www.primegrid.com/primegrid_genefer_3_1_2_7_2.07_i686-pc-linux-gnu__atiGFN -boinc -q 230006^1048576+1 --device 0
No OpenCL device found.
22:24:33 (8745): called boinc_finish
</stderr_txt>
]]>
I'm using Linux Mint 13 LTS (x64), ATI 7970 using driver 12.14 downloaded from ATI. No overclocking.
Is there a OpenCL package I'm missing?
Thank you
____________
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
Help! :0
I'm using Linux Mint 13 LTS (x64), ATI 7970 using driver 12.14 downloaded from ATI. No overclocking.
Is there a OpenCL package I'm missing?
Thank you
Are you able to run PPS Sieve on that GPU?
Are you able to run any BOINC projects on that GPU?
____________
My lucky number is 75898524288+1 |
|
|
RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
                
|
[quote]Help! :0
Is there a OpenCL package I'm missing?
Try installing the AMD SDK with a supported version of Catalyst:
http://developer.amd.com/tools-and-sdks/heterogeneous-computing/amd-accelerated-parallel-processing-app-sdk/downloads/
Not saying its required, just something to try to get it to work with Linux Mint.
If it works for you I am tempted to try Mint. Looks cool. Windows 8 is Meh. |
|
|
|
|
[quote]Help! :0
Is there a OpenCL package I'm missing?
Try installing the AMD SDK with a supported version of Catalyst:
http://developer.amd.com/tools-and-sdks/heterogeneous-computing/amd-accelerated-parallel-processing-app-sdk/downloads/
Not saying its required, just something to try to get it to work with Linux Mint.
If it works for you I am tempted to try Mint. Looks cool. Windows 8 is Meh.
PPS sieve works fine Mike.
I will try the SDK asap.
____________
|
|
|
|
|
|
That was it!
Thanks Roger! |
|
|
|
|
|
6:04:01 to do a shorty on my 7970.
Nice!
Now to tackle a WR.
____________
|
|
|
|
|
6:04:01 to do a shorty on my 7970.
Nice!
Now to tackle a WR.
Wow that's fast. Hopefully it will encourage TPTB to run another GFN-short challenge in 2014 (I know, it's early to be discussing such things, but I thought I'd throw that out there). There's also a tiny boost from the recent availability of the AVX version (and SSE3 now too) of the CPU app.
--Gary |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
6:04:01 to do a shorty on my 7970.
Nice!
Now to tackle a WR.
Wow that's fast. Hopefully it will encourage TPTB to run another GFN-short challenge in 2014 (I know, it's early to be discussing such things, but I thought I'd throw that out there). There's also a tiny boost from the recent availability of the AVX version (and SSE3 now too) of the CPU app.
--Gary
Indeed it is, but look at this short workunit: http://www.primegrid.com/workunit.php?wuid=355286040
5:48:08 from a Tahiti-class ATI GPU and the wingman's Titan did it in 4:16:54. Both were OpenCL.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
As you mentioned that the OpenCL version is running faster than the CUDA version on a Kepler GPU I tried to crunch some short WUs. But all are crashing on my host while the CUDA version seems to run fine.
GPU is a GTX 670, OS is WinXP 32.
Here's an example.
Stderr shows
couldn't start app: CreateProcess() failed - (unknown error)
Any ideas? |
|
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 842
ID: 164101
Credit: 306,514,171
RAC: 5,382
|
OS is WinXP 32.
Stderr shows
couldn't start app: CreateProcess() failed - (unknown error)
Any ideas?
On my computer, OpenCL.dll depends on MSVCR100.dll (Visual Studio 2010 C run-time).
You can try to install the redistribution package (the latest one is "Microsoft Visual C++ 2010 Service Pack 1 Redistributable Package MFC Security Update"). |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
OS is WinXP 32.
Stderr shows
couldn't start app: CreateProcess() failed - (unknown error)
Any ideas?
On my computer, OpenCL.dll depends on MSVCR100.dll (Visual Studio 2010 C run-time).
You can try to install the redistribution package (the latest one is "Microsoft Visual C++ 2010 Service Pack 1 Redistributable Package MFC Security Update").
If that doesn't work (please try Yves' suggestion first), try running this version of the program:
EDIT: The XP-built app is now the official app, so there's no longer a separate download for this version.
____________
My lucky number is 75898524288+1 |
|
|
|
|
If that doesn't work (please try Yves' suggestion first), try running this version of the program:
http://www.primegrid.com/download/geneferbeta/geneferocl-windows-x86-XP.exe
1. I re-installed Microsoft Visual C++ 2010 Service Pack 1 Redistributable Package, but app was still crashing.
2. Then I downloaded the linked executable and made an app_info.xml for it ... and IT'S RUNNING!
So I reckon it's an issue with the official app and XP? |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
So I reckon it's an issue with the official app and XP?
Yes. The question is whether the built-for-XP version of the app breaks anything with newer versions of Windows.
And if we go even further back, I suspect there's also problems with W2K which will be more difficult to fix.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
|
For XP, it should no longer be necessary to use the special version of geneferocl and app_info. That version is now the official app and will sent from the server as 2.09.
I have replaced the Windows geneferocl app with a version that should work on all versions of Windows from XP and Server 2003 onwards. With the help of Jim and Randy, I've done at least rudimentary testing with Windows XP, Windows 7, Windows 8, Windows Server 2003, Windows Server 2008, and Windows Server 2012.
Windows XP is as far back as the support will go, however.
I strongly suspect this will not work on Windows 2000 or even earlier versions of Windows. Of course, finding an OpenCL capable GPU that can be plugged into a computer that came with Windows 2000 is probably pretty rare. In the last sixth months, only 75 Windows 2000 hosts have connected to PrimeGrid, and only two of those have a GPU recognized by BOINC. Both of those GPUs are Nvidia and neither has the required double precision hardware. There's also 5 hosts running 98 or ME, but none have GPUs.
____________
My lucky number is 75898524288+1 |
|
|
Scott BrownVolunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2416
ID: 1178
Credit: 19,978,285,056
RAC: 19,305,139
                                               
|
I strongly suspect this will not work on Windows 2000 or even earlier versions of Windows. Of course, finding an OpenCL capable GPU that can be plugged into a computer that came with Windows 2000 is probably pretty rare.
And there are two more reasons that Win2k support is not needed. NVidia drivers have CUDA support that usually works on Win2k, ATI never officially supported a Win2k driver version that worked with their compute capable cards (even the old HD 2xxx series). |
|
|
|
|
|
In the old dev thread the question was asked about "amd and nvidia in the same machine, which does it choose". I was unable to answer at the time as my machine was still being built.
The answer is that amd opencl wr decides to run on the nvidia card.
On this machine http://www.primegrid.com/show_host_detail.php?hostid=408638, the amd card is in pci slot 1, and has the display attached. The task started running with --device 0, and chose the nvidia card.
BOINC manager shows the task as requiring an ATI, and continues running other nvidia tasks at the same time.
Logs will be forthcoming when i get home from work.
Regards, Andrew
____________
  |
|
|
|
|
|
From the stderr txt:
geneferocl 3.1.2-7 (Windows 32-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: projects/www.primegrid.com/primegrid_genefer_3_1_2_7_2.09_windows_intelx86__atiGFNWR.exe -boinc -q 12040^4194304+1 --device 0
Priority change succeeded.
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 760', version 'OpenCL 1.1 CUDA' and driver '326.80'.
From the sched_reply:
<app_version>
<app_name>genefer_wr</app_name>
<version_num>209</version_num>
<file_ref>
<file_name>primegrid_genefer_3_1_2_7_2.09_windows_intelx86__atiGFNWR.exe</file_name>
<main_program/>
</file_ref>
<platform>windows_intelx86</platform>
<plan_class>atiGFNWR</plan_class>
<avg_ncpus>0.170704</avg_ncpus>
<max_ncpus>0.170704</max_ncpus>
<flops>35417298943.728439</flops>
<coproc>
<type>ATI</type>
<count>1.000000</count>
</coproc>
<gpu_ram>402653184.000000</gpu_ram>
</app_version>
...
<command_line>-boinc -q "12040^4194304+1"</command_line>
If i modify the command line in the reply with "-amd" will that cause the task to pick it up? (i went and looked at the assembla repo) Or does it need to be specified elsewhere? (don't want to go full app_config as i have some SoBs slowly ticking away i don't want to accidentally nuke)
____________
|
|
|
|
|
|
And finally, to answer myself:
if (*fp || *if_bench || *if_bench2 || *if_bench3 || *if_test || *if_residuals || *if_limits || *if_version || strcmp(platformName, ""))
{
fprintf(stderr, "-boinc used with incompatible option (filename, -b, -b2, -b3, -t, -r, -l, -v, -V, -nvidia, -ati, or -amd).\n");
boincFatal(GFN_INVALID_ARGS);
}
I must not specify -amd/-ati as this is mutually exclusive with BOINC mode and will error. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
If i modify the command line in the reply with "-amd" will that cause the task to pick it up? (i went and looked at the assembla repo) Or does it need to be specified elsewhere? (don't want to go full app_config as i have some SoBs slowly ticking away i don't want to accidentally nuke)
It's a configuration error on the server. There's supposed to be a configuration file that's sent along with the task, and for Windows tasks (both WR and short) the configuration file isn't being sent.
The Nvidia/AMD selection had to be done via a configuration file rather than command line option due to BOINC restrictions.
I'll correct this sometime later today. I'm not going to fix it now as I'm rushing out the door and don't want to make an even bigger mistake. Sorry for the inconvenience.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
|
NVIDIA/AMD selection on Windows hosts should now be working.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
No worries, thanks!
So, is there a way i can insert the config file into the slot folder, or would i be better to abort the task and wait for another? (it hasn't progressed passed 0% yet, it keeps being pre-empted by other and tasks in my list as the manager decided it needs to high-priority clean them out)
Cheers, Andrew |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
No worries, thanks!
So, is there a way i can insert the config file into the slot folder, or would i be better to abort the task and wait for another? (it hasn't progressed passed 0% yet, it keeps being pre-empted by other and tasks in my list as the manager decided it needs to high-priority clean them out)
Cheers, Andrew
If it's at 0%, the easiest thing is just to abort it.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
opencl_platform.txt added to slot folder, with contents AMD
stderr now shows the task running on Tahiti (a 7950), GPU-Z shows 100% utilization. Estimated runtime 80hrs. Wow. 100hrs faster than the 760 in the same machine (with cuda). |
|
|
|
|
|
hello, Nvidia OpenCL app won´t work on OS X 10.8.4 or 10.8.5 with GTS 450 and GTX 650. always crunch errors behind start.
the collate OpenCL app works well |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
hello, Nvidia OpenCL app won´t work on OS X 10.8.4 or 10.8.5 with GTS 450 and GTX 650. always crunch errors behind start.
the collate OpenCL app works well
Is the server not sending any tasks to your computer?
If not, what does your log show when you request work? Are you able to get PrimeGrid CPU tasks? Can you get PrimeGrid GPU tasks?
If you are getting tasks but the the tasks are failing, what error message appears in the result page?
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
hi, thx for reply, here its the output:
| CUDA: NVIDIA GPU 0: GeForce GTX 650 (driver version 5.5.25, CUDA version 5.5, compute capability 3.0, 2048MB, 1646MB available, 813 GFLOPS peak)
| OpenCL: NVIDIA GPU 0: GeForce GTX 650 (driver version 8.16.74 310.40.00.10f02, device version OpenCL 1.1, 2048MB, 1646MB available, 813 GFLOPS peak)
Stderr output
<core_client_version>7.2.11</core_client_version>
<![CDATA[
<message>
process exited with code 1 (0x1, -255)
</message>
<stderr_txt>
dyld: DYLD_ environment variables being ignored because main executable (/Library/Application Support/BOINC Data/slots/2/../../switcher/switcher) is setuid or setgid
geneferocl 3.1.2-7 (Apple x86 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: primegrid_genefer_3_1_2_7_2.07_i686-apple-darwin__OCLcudaGFN -boinc -q 242144^1048576+1 --device 0
No OpenCL device found.
15:36:13 (587): called boinc_finish
</stderr_txt>
]]> |
|
|
|
|
15. What is the status of OpenCL* on Intel® Processors on Apple* Mac OS* X?
As of today, Apple* support OpenCL* as part of their Mac OS* X. We suggest you ask Apple for direct comment on specific features and compatibility.
|
|
|
|
|
|
apple won´t support OpenCL on intel cpu´s (gpu) jet. only support ati and nvdia.
i think that the app won´t work as it should because the sonntags collatz OpenCL app works well with my nvidia gts 450 and gtx 650 on os x 10.8.5 and the latest boinc client |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
hello, Nvidia OpenCL app won´t work on OS X 10.8.4 or 10.8.5 with GTS 450 and GTX 650. always crunch errors behind start.
the collate OpenCL app works well
You are able to run the the GeneferCUDA app on those systems, correct?
____________
My lucky number is 75898524288+1 |
|
|
|
|
apple won´t support OpenCL on intel cpu´s (gpu) jet. only support ati and nvdia.
yes, but nvidia software don't promote opencl... |
|
|
|
|
|
true, but os x promote the nvdia and ati OpenCL gfx card |
|
|
|
|
|
yes, but i wan´t to use the faster OpenCL app for my nvidia kepler card |
|
|
|
|
|
see, amd/ati have opencl sdk, nvidia does not have it. for nvidia gpu should be install third-party sdk, for example intel opencl sdk. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
hello, Nvidia OpenCL app won´t work on OS X 10.8.4 or 10.8.5 with GTS 450 and GTX 650. always crunch errors behind start.
the collate OpenCL app works well
You are able to run the the GeneferCUDA app on those systems, correct?
Mac intel,
I believe I know what the problem is. I need you to run the GeneferOCL program from the command line with the "-b" parameter, and post the entire results here. I need to see what the platform name is. I suspect it's "Apple" instead of "NVIDIA" or "AMD".
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
hello, here its the my primegrid_genefer_3_1_2_7_2.07_i686-apple-darwin__OCLcudaGFNWR benchmark output:
bash-3.2# /Library/Application\ Support/BOINC\ Data/projects/www.primegrid.com/primegrid_genefer_3_1_2_7_2.07_i686-apple-darwin__OCLcudaGFNWR -b
geneferocl 3.1.2-7 (Apple x86 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: /Library/Application Support/BOINC Data/projects/www.primegrid.com/primegrid_genefer_3_1_2_7_2.07_i686-apple-darwin__OCLcudaGFNWR -b
Running on platform 'Apple', device 'GeForce GTX 650', version 'OpenCL 1.1 ' and driver '8.16.74 310.40.00.10f02'.
Generalized Fermat Number Bench
2199064^8192+1 Time: 249 us/mul. Err: 0.2656 51956 digits
1798620^16384+1 Time: 402 us/mul. Err: 0.2500 102481 digits
1471094^32768+1 Time: 577 us/mul. Err: 0.2656 202102 digits
1203210^65536+1 Time: 1.04 ms/mul. Err: 0.2656 398482 digits
984108^131072+1 Time: 2 ms/mul. Err: 0.2383 785521 digits
804904^262144+1 Time: 3.78 ms/mul. Err: 0.2305 1548156 digits
658332^524288+1 Time: 6.61 ms/mul. Err: 0.2324 3050541 digits
538452^1048576+1 Time: 13.9 ms/mul. Err: 0.2344 6009544 digits
440400^2097152+1 Time: 24.9 ms/mul. Err: 0.2148 11836006 digits
360204^4194304+1 Time: 41.8 ms/mul. Err: 0.2109 23305854 digits
294612^8388608+1 Time: 71.9 ms/mul. Err: 0.1953 45879398 digits
Genefer Mark = 9.
bash-3.2#
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
hello, Nvidia OpenCL app won´t work on OS X 10.8.4 or 10.8.5 with GTS 450 and GTX 650. always crunch errors behind start.
the collate OpenCL app works well
The good news: This should now be fixed and you (and everyone else) should now be able to successfully run GeneferOCL on Mac computers with either NVIDIA or AMD GPUs.
The bad news: To make this work, I had to remove the ability to select between Nvidia or AMD GPUs. For most people, this won't matter. However, if you have a Mac with BOTH an Nvidia and an AMD GPU, the program isn't going to know which one it's supposed to run on and the results may be unpredictable.
If you have only one GPU in your Mac (or only one brand of GPU is you have more than one GPU), this should work fine.
____________
My lucky number is 75898524288+1 |
|
|
kiskaVolunteer tester
Send message
Joined: 13 Apr 12
Posts: 47
ID: 138397
Credit: 242,095,027
RAC: 98,142
               
|
|
Hello:
I am back! And I have an error that I am unable to understand.
here is the wu link: http://www.primegrid.com/result.php?resultid=488875576
feedback is appreciated. I am running a Intel Core i5-3337U, along with a GTX 740M doing Genfer OCL.
Kiska
____________
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
hello, Nvidia OpenCL app won´t work on OS X 10.8.4 or 10.8.5 with GTS 450 and GTX 650. always crunch errors behind start.
the collate OpenCL app works well
The good news: This should now be fixed and you (and everyone else) should now be able to successfully run GeneferOCL on Mac computers with either NVIDIA or AMD GPUs.
The bad news: To make this work, I had to remove the ability to select between Nvidia or AMD GPUs. For most people, this won't matter. However, if you have a Mac with BOTH an Nvidia and an AMD GPU, the program isn't going to know which one it's supposed to run on and the results may be unpredictable.
If you have only one GPU in your Mac (or only one brand of GPU is you have more than one GPU), this should work fine.
Has anyone been able to run GeneferOCL successfully on a Mac? I see several tasks in progress, but none have been returned successfully.
____________
My lucky number is 75898524288+1 |
|
|
|
|
hello, Nvidia OpenCL app won´t work on OS X 10.8.4 or 10.8.5 with GTS 450 and GTX 650. always crunch errors behind start.
the collate OpenCL app works well
The good news: This should now be fixed and you (and everyone else) should now be able to successfully run GeneferOCL on Mac computers with either NVIDIA or AMD GPUs.
The bad news: To make this work, I had to remove the ability to select between Nvidia or AMD GPUs. For most people, this won't matter. However, if you have a Mac with BOTH an Nvidia and an AMD GPU, the program isn't going to know which one it's supposed to run on and the results may be unpredictable.
If you have only one GPU in your Mac (or only one brand of GPU is you have more than one GPU), this should work fine.
Has anyone been able to run GeneferOCL successfully on a Mac? I see several tasks in progress, but none have been returned successfully.
Out of all the work at PG the only thing that works on my old mac pro (8 core with a PNY Signature GTX 460 1GB) is Wieferich on PRPNet. Nothing from boinc works, cuda/ocl, llr/sieve. I think this card is jinxed.
____________
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
Hello:
I am back! And I have an error that I am unable to understand.
here is the wu link: http://www.primegrid.com/result.php?resultid=488875576
feedback is appreciated. I am running a Intel Core i5-3337U, along with a GTX 740M doing Genfer OCL.
Kiska
Unfortunately, we do not understand what the cause of this crash was. I assume this is a laptop (because of the mobile GPU)? If so, is it possible the laptop went to sleep? We're theorizing that perhaps this played a role in the problem, although going to sleep really shouldn't cause a problem.
Aside from that, does this problem occur with all GFN tasks or was this a one-time occurrence?
____________
My lucky number is 75898524288+1 |
|
|
kiskaVolunteer tester
Send message
Joined: 13 Apr 12
Posts: 47
ID: 138397
Credit: 242,095,027
RAC: 98,142
|
|
Just this one task I downloaded as soon I saw the error, I cancelled the rest. It doesn't go to sleep as per user defined preferences. So maybe a bug in the program?
____________
|
|
|
|
|
|
hello, sry for my late reply, my genefer version 208 (OCLcudaGFN) is still working.
what i could say right now is that the GPU load is much higher than the genefer Cuda app. and it looks like 30 % faster on my Nvidia GTX 650
my first OCL WU was finished with an error. i don´t know why, could it be because the high GPU load or sleep issue? could i reduce the GPU load with an app.xml?
ps:
at the weekend i installed os x 10.9 GM and i could tell you that OCL for Intel GPU´s above HD4000 are now supported. i test it with Luxmark v2.1 beta 2 and boon report is:
OpenCL: Intel GPU 0: HD Graphics 4000 (driver version 1.2(Sep 19 2013 22:31:23), device version OpenCL 1.2, 1024MB, 1024MB available, 154 GFLOPS peak)
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
my first OCL WU was finished with an error. i don´t know why, could it be because the high GPU load or sleep issue?
This is from that task:
maxErr exceeded for 247382^1048576+1, 0.5000 > 0.4500
MaxErr exceeded may be caused by overclocking, overheated GPUs and other transient errors.
Waiting 10 minutes before attempting to continue from last checkpoint...
Calculation failed after 6 tries, please test your GPU or CPU carefully to ensure it can reliably compute GFN tests.
This is the same problem we've seen since the beginning with the CUDA version of Genefer on Nvidia GPUs. With the CUDA version it's a problem related to overclocking or overheating, and I assume the OpenCL version has the same problem (which it should since this is a hardware design problem, not a software bug). The following steps may help.
1) Make sure you have excellent cooling for the GPU. If you can, manually set the GPU fan to run at 100%. Open up the case and having a desk fan blowing into the case helps. One extreme setup had an air conditioner ducted directly into the computer case, which allowed the GPU to run correctly even at very high overclocking.
2) If you're using ANY overclocking on the GPU, go back to stock speeds. This includes factory overclocking. Any overclocking, even if set by the factory, has the potential to cause the calculation to fail.
3) If it's not overclocked, try underclocking the GPU, especially the memory clock. Sometimes even stock clock speeds cause problems. Try reducing the memory clock first as this seems to do the most good and doesn't affect the calculation speed very much.
could i reduce the GPU load with an app.xml?
Unfortunately, no. The way to reduce the stress on the GPU, as mentioned above, is to lower its clock speeds and improve its cooling.
ps:
at the weekend i installed os x 10.9 GM and i could tell you that OCL for Intel GPU´s above HD4000 are now supported. i test it with Luxmark v2.1 beta 2 and boon report is:
OpenCL: Intel GPU 0: HD Graphics 4000 (driver version 1.2(Sep 19 2013 22:31:23), device version OpenCL 1.2, 1024MB, 1024MB available, 154 GFLOPS peak)
While Intel GPUs can be used for some other GPU tasks, they can't be used for Genefer. Genefer requires double precision floating point, which the Intel GPUs do not have.
____________
My lucky number is 75898524288+1 |
|
|
TylerProject administrator
Volunteer tester
Send message
Joined: 4 Dec 12
Posts: 1081
ID: 183129
Credit: 1,384,625,026
RAC: 7,097
            
|
Just this one task I downloaded as soon I saw the error, I cancelled the rest. It doesn't go to sleep as per user defined preferences. So maybe a bug in the program?
Is your screen set to turn off? Because this utilizes the GPU perhaps the gpu is put to sleep when the monitor goes off, which people don't usually make always stay on.
____________
275*2^3585539+1 is prime!!! (1079358 digits)
 Proud member of Aggie the Pew
 |
|
|
|
|
|
Hi,I run with 2 5850 cards and now 1 task stop at 3.592%.I stop and start again and the same.Is the task failed or is it a Primefound? |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
Hi,I run with 2 5850 cards and now 1 task stop at 3.592%.I stop and start again and the same.Is the task failed or is it a Primefound?
Look in stderr.txt in the appropriate slot directory and see what it says.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Hey all,
I'm trying to get some GPU tasks on my puter but no luck.
I have i7 2600k and a 7850 (ASUS HD 7850 Direct CU II).
have the updated driver from AMD but didn't specifically download
anything else.
It wont send me one to try,
Thanks. |
|
|
|
|
Hey all,
I'm trying to get some GPU tasks on my puter but no luck.
I have i7 2600k and a 7850 (ASUS HD 7850 Direct CU II).
have the updated driver from AMD but didn't specifically download
anything else.
It wont send me one to try,
Thanks.
1. BOINC-Manager has GPU enabled? (Activity / Use GPU always) and (Tools / Computing preferences)
2. In the PrimeGrid preferences is ATI GPU enabled?
3. in the PrimeGrid preferences is "Generalized Fermat Prime Search" - option "AMD/ATI (OpenCL)" enabled?
Edit: you have an ATI 7850, this card will need about 160000 seconds per WU, so your CPU with 4 cores (AVX) is much faster.....
____________
DeleteNull |
|
|
|
|
Hey all,
I'm trying to get some GPU tasks on my puter but no luck.
I have i7 2600k and a 7850 (ASUS HD 7850 Direct CU II).
have the updated driver from AMD but didn't specifically download
anything else.
It wont send me one to try,
Thanks.
1. BOINC-Manager has GPU enabled? (Activity / Use GPU always) and (Tools / Computing preferences)
2. In the PrimeGrid preferences is ATI GPU enabled?
3. in the PrimeGrid preferences is "Generalized Fermat Prime Search" - option "AMD/ATI (OpenCL)" enabled?
Edit: you have an ATI 7850, this card will need about 160000 seconds per WU, so your CPU with 4 cores (AVX) is much faster.....
Hmmm on number 1, i didn't have activity on always (had it on based on preferences) fixed that. preferences was fine.
2 yes and 3 yes
avx is faster on the short search perhaps, i want a long search (I have all 3 GPU
projects selected though and get 0)
still getting zero on updage. i see there are some left.
EDIT: at least i read about force AVX ran the first 8 or so with SSE3
and another 8 on the IVY bridge...
will be fun to see the speed increase.
Once I get this working on the 7850, I will plug in another I7 3770K
that has a 7950 in it. I just dont now, because it uses 70 extra watts
just to be plugged in. My two computers are already 180 KW a month,
want it to be worthwhile to run 3. |
|
|
|
|
|
Hi John,
your BOINC-Version is 6.10.58. You need a newer one. Try http://boinc.berkeley.edu/dl/boinc_7.2.23_windows_x86_64.exe from http://boinc.berkeley.edu/dl/?C=M;O=D
To get AVX-CPU workunits in Windows you have to patch the boinc-client (or use an app_info.xml, or use Linux)
____________
DeleteNull |
|
|
|
|
Hi John,
your BOINC-Version is 6.10.58. You need a newer one. Try http://boinc.berkeley.edu/dl/boinc_7.2.23_windows_x86_64.exe from http://boinc.berkeley.edu/dl/?C=M;O=D
To get AVX-CPU workunits in Windows you have to patch the boinc-client (or use an app_info.xml, or use Linux)
I believe there is a "force AVX" check-box on the preferences page to allow Windows hosts to receive GFN CPU units utilizing AVX, without resorting to the drastic measures suggested.
Good advice to update from 6.10.xx... that's old and rusty. Pre-AVX-aware. |
|
|
JimBHonorary cruncher
Send message
Joined: 4 Aug 11
Posts: 920
ID: 107307
Credit: 989,312,479
RAC: 2,016
        
|
Hey all,
I'm trying to get some GPU tasks on my puter but no luck.
I have i7 2600k and a 7850 (ASUS HD 7850 Direct CU II).
have the updated driver from AMD but didn't specifically download
anything else.
It wont send me one to try,
Thanks.
John, looking at your computers BOINC knows there's a GPU there but doesn't display anything about the model. When I look at the computers of others who've posted in this thread there is an actual model number (or range of them). As of when I'm writing this, that computer is still on 6.10.58. I'd update to a newer BOINC client and then if you still have the problem you may want to reinstall the driver. I know I had this problem during testing and it was a software issue of some kind. |
|
|
|
|
|
hello, so my first genefer version 208 (OCLcudaGFN) is sussesfull finished. 491733742
before i run sussesfull a genefer CUDA task (can´t find the number) to check the difference in speed an what i could do beside the task. for the cuda task i need round about 55 h and for the ocl task 35 h
with the ocl app my gfx card is about 4-5 C (52/57) hotter both with 100% usage with atmonitor and you can´t use a browser with the ocl app beside.
my gtx 650 is not over clocked, and i never got problems with the screensaver with this or other projects but it is now disabled. rack and fans are cleaned at the last weekend |
|
|
|
|
Hi John,
your BOINC-Version is 6.10.58. You need a newer one. Try http://boinc.berkeley.edu/dl/boinc_7.2.23_windows_x86_64.exe from http://boinc.berkeley.edu/dl/?C=M;O=D
To get AVX-CPU workunits in Windows you have to patch the boinc-client (or use an app_info.xml, or use Linux)
I believe there is a "force AVX" check-box on the preferences page to allow Windows hosts to receive GFN CPU units utilizing AVX, without resorting to the drastic measures suggested.
Good advice to update from 6.10.xx... that's old and rusty. Pre-AVX-aware.
OK, I have the force avx box going and have gotten some,
when I finish my current work units I will update BOINC to newer... I just dont like the way they moved my tabs around....
Change sux (I hve newer versions on other comptuers) |
|
|
|
|
I just dont like the way they moved my tabs around....
Nobody likes... )))
____________
|
|
|
|
|
|
And BINGO,
Whoever suggested new version, that did it for me
Funny thing is it says it's going to take 31 days to finish,
and gave me like 20 to do it :)
Hope the estimate is wrong (It often is I know)
Great haven't done a world record search since it was probably less than 5 million digits. And of course without GPU.
Looking forward to this!
should get the HD 7950 going :)
Maybe i will, but tonight is my Friday so will not mix liquid refreshment,
with open computer cases... TALLYHO |
|
|
|
|
And BINGO,
Whoever suggested new version, that did it for me
Funny thing is it says it's going to take 31 days to finish,
and gave me like 20 to do it :)
Hope the estimate is wrong (It often is I know)
Great haven't done a world record search since it was probably less than 5 million digits. And of course without GPU.
Looking forward to this!
should get the HD 7950 going :)
Maybe i will, but tonight is my Friday so will not mix liquid refreshment,
with open computer cases... TALLYHO
Well it errorred out,
oh well i got svx goin i guess that will do |
|
|
288larssonVolunteer tester
Send message
Joined: 17 Apr 10
Posts: 136
ID: 58815
Credit: 5,967,247,909
RAC: 3,284,288
                           
|
|
hi can not get any work to my new AMD Radeon R9 290X (Hawaii) |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
hi can not get any work to my new AMD Radeon R9 290X (Hawaii)
Wasn't this task http://www.primegrid.com/result.php?resultid=495551493 sent to your computer with the R9 290X?
____________
My lucky number is 75898524288+1 |
|
|
288larssonVolunteer tester
Send message
Joined: 17 Apr 10
Posts: 136
ID: 58815
Credit: 5,967,247,909
RAC: 3,284,288
|
hi can not get any work to my new AMD Radeon R9 290X (Hawaii)
Wasn't this task http://www.primegrid.com/result.php?resultid=495551493 sent to your computer with the R9 290X?
Hi Yes wrong question how can I get it to run? |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,424,319,916
RAC: 3,939,601
           
|
|
That new card is going to be interesting... |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
hi can not get any work to my new AMD Radeon R9 290X (Hawaii)
Wasn't this task http://www.primegrid.com/result.php?resultid=495551493 sent to your computer with the R9 290X?
Hi Yes wrong question how can I get it to run?
The task should run automatically, same as any other BOINC task.
____________
My lucky number is 75898524288+1 |
|
|
288larssonVolunteer tester
Send message
Joined: 17 Apr 10
Posts: 136
ID: 58815
Credit: 5,967,247,909
RAC: 3,284,288
|
|
hi you're right, it seems that boinc does not detect the card correctly
2013-10-25 22:50:16 | | Starting BOINC client version 7.2.23 for windows_x86_64
2013-10-25 22:50:16 | | log flags: file_xfer, sched_ops, task
2013-10-25 22:50:16 | | Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6
2013-10-25 22:50:16 | | Data directory: C:\ProgramData\BOINC
2013-10-25 22:50:16 | | OpenCL: AMD/ATI GPU 0: Hawaii (driver version 1348.4 (VM), device version OpenCL 1.2 AMD-APP (1348.4), 2048MB, 2048MB available, 3520 GFLOPS peak)
2013-10-25 22:50:16 | | OpenCL CPU: Intel(R) Core(TM) i7-4930K CPU @ 3.40GHz (OpenCL driver vendor: Advanced Micro Devices, Inc., driver version 1348.4 (sse2,avx), device version OpenCL 1.2 AMD-APP (1348.4))
2013-10-25 22:50:16 | | App version needs CAL but GPU doesn't support it
2013-10-25 22:50:16 | PrimeGrid | Application uses missing ATI GPU
2013-10-25 22:50:16 | | Processor: 7 GenuineIntel Intel(R) Core(TM) i7-4930K CPU @ 3.40GHz [Family 6 Model 62 Stepping 4]
2013-10-25 22:50:16 | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 popcnt aes nx lm vmx tm2 dca pbe
2013-10-25 22:50:16 | | OS: Microsoft Windows 8.1: Core x64 Edition, (06.03.9600.00)
2013-10-25 22:50:16 | | Memory: 15.94 GB physical, 31.94 GB virtual
2013-10-25 22:50:16 | | Disk: 111.45 GB total, 25.81 GB free
2013-10-25 22:50:16 | | Local time is UTC +2 hours
|
|
|
|
|
hi you're right, it seems that boinc does not detect the card correctly
2013-10-25 22:50:16 | | Starting BOINC client version 7.2.23 for windows_x86_64
2013-10-25 22:50:16 | | log flags: file_xfer, sched_ops, task
2013-10-25 22:50:16 | | Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6
2013-10-25 22:50:16 | | Data directory: C:\ProgramData\BOINC
2013-10-25 22:50:16 | | OpenCL: AMD/ATI GPU 0: Hawaii (driver version 1348.4 (VM), device version OpenCL 1.2 AMD-APP (1348.4), 2048MB, 2048MB available, 3520 GFLOPS peak)
2013-10-25 22:50:16 | | OpenCL CPU: Intel(R) Core(TM) i7-4930K CPU @ 3.40GHz (OpenCL driver vendor: Advanced Micro Devices, Inc., driver version 1348.4 (sse2,avx), device version OpenCL 1.2 AMD-APP (1348.4))
2013-10-25 22:50:16 | | App version needs CAL but GPU doesn't support it
2013-10-25 22:50:16 | PrimeGrid | Application uses missing ATI GPU
2013-10-25 22:50:16 | | Processor: 7 GenuineIntel Intel(R) Core(TM) i7-4930K CPU @ 3.40GHz [Family 6 Model 62 Stepping 4]
2013-10-25 22:50:16 | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 popcnt aes nx lm vmx tm2 dca pbe
2013-10-25 22:50:16 | | OS: Microsoft Windows 8.1: Core x64 Edition, (06.03.9600.00)
2013-10-25 22:50:16 | | Memory: 15.94 GB physical, 31.94 GB virtual
2013-10-25 22:50:16 | | Disk: 111.45 GB total, 25.81 GB free
2013-10-25 22:50:16 | | Local time is UTC +2 hours
Do you have tried BOINC 7.2.26? There is an update of AMD detection. |
|
|
288larssonVolunteer tester
Send message
Joined: 17 Apr 10
Posts: 136
ID: 58815
Credit: 5,967,247,909
RAC: 3,284,288
|
Do you have tried BOINC 7.2.26? There is an update of AMD detection.
Yes same problem |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
hi you're right, it seems that boinc does not detect the card correctly
2013-10-25 22:50:16 | | OpenCL: AMD/ATI GPU 0: Hawaii (driver version 1348.4 (VM), device version OpenCL 1.2 AMD-APP (1348.4), 2048MB, 2048MB available, 3520 GFLOPS peak)
2013-10-25 22:50:16 | | OpenCL CPU: Intel(R) Core(TM) i7-4930K CPU @ 3.40GHz (OpenCL driver vendor: Advanced Micro Devices, Inc., driver version 1348.4 (sse2,avx), device version OpenCL 1.2 AMD-APP (1348.4))
2013-10-25 22:50:16 | | App version needs CAL but GPU doesn't support it
2013-10-25 22:50:16 | PrimeGrid | Application uses missing ATI GPU
Got it, thanks.
The problem is that the server is actually asking for CAL support (instead of OpenCL) and I guess CAL isn't supported on the new GPUs. I suspect that PPS sieve also won't work on the new GPU, since it's configured the same way. It's possible this affects Linux as well as Windows. I suspect Mac works.
This is, in theory, easy to fix. I need to change a configuration file on the server and the fix should be effective immediately for new tasks that are sent out. I could make that change right now, but my availability over the next several days is going to be extremely limited, and if something breaks because of this change, it may stay broken for a while. Therefore, I don't want to simply fix the configuration; I'm afraid the "fix" might have unintended consequences and break ATI support for older GPUs on Windows and/or Linux. Therefore, I'm not going to make the change just yet. I'd like to set this up on the test server first, and I probably won't be able to do that for several days, at least.
You can help by doing a few simple tests:
1) Try running a BOINC PPS sieve task on your new ATI GPU. It should fail to work in the same way as the GFN task fails.
2) Try running the ATI GFN app manually. It's not necessary to run a real test. Running the benchmarks is sufficient. You can run the app directly in the C:\ProgramData\BOINC\projects\www.primegrid.com directory. The command line would be:
primegrid_genefer_3_1_2_7_2.10_windows_intelx86__atiGFN.exe -b
If test 2 above works, you should be able to run using app_info until I get the server configuration fixed.
____________
My lucky number is 75898524288+1 |
|
|
288larssonVolunteer tester
Send message
Joined: 17 Apr 10
Posts: 136
ID: 58815
Credit: 5,967,247,909
RAC: 3,284,288
|
Got it, thanks.
You can help by doing a few simple tests:
1) Try running a BOINC PPS sieve task on your new ATI GPU. It should fail to work in the same way as the GFN task fails.
2) Try running the ATI GFN app manually. It's not necessary to run a real test. Running the benchmarks is sufficient. You can run the app directly in the C:\ProgramData\BOINC\projects\www.primegrid.com directory. The command line would be:
primegrid_genefer_3_1_2_7_2.10_windows_intelx86__atiGFN.exe -b
If test 2 above works, you should be able to run using app_info until I get the server configuration fixed.
1) Yes
2) Works
Generalized Fermat Number Bench
2199064^8192+1 Time: 73.2 us/mul. Err: 0.2344 51956 digits
1798620^16384+1 Time: 76.9 us/mul. Err: 0.2266 102481 digits
1471094^32768+1 Time: 76.9 us/mul. Err: 0.2344 202102 digits
1203210^65536+1 Time: 83 us/mul. Err: 0.2656 398482 digits
984108^131072+1 Time: 132 us/mul. Err: 0.2188 785521 digits
804904^262144+1 Time: 273 us/mul. Err: 0.2188 1548156 digits
658332^524288+1 Time: 859 us/mul. Err: 0.2188 3050541 digits
538452^1048576+1 Time: 1.25 ms/mul. Err: 0.2266 6009544 digits
440400^2097152+1 Time: 2.71 ms/mul. Err: 0.2266 11836006 digits
360204^4194304+1 Time: 5.72 ms/mul. Err: 0.2031 23305854 digits
294612^8388608+1 Time: 11.3 ms/mul. Err: 0.1895 45879398 digits
Genefer Mark = 76.
runing using app_info. Works:)) Thanks.
|
|
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 842
ID: 164101
Credit: 306,514,171
RAC: 5,382
|
AMD Radeon R9 290X (Hawaii)
Genefer Mark = 76.
NVIDIA GeForce GTX TITAN: Genefer Mark = 120.
AMD Radeon HD 7970 GHz: Genefer Mark = 94.
Slower than the HD 7970... as expected :-( FP64 unit clock ratio is 1/8, on a Tahiti it is 1/4.
The AMD Radeon R9 280X is a Tahiti then it should be faster than the 290X for FP64 GPU computation.
Yves |
|
|
|
|
|
Has anyone tried the AMD r9 280X?
Is it on pair with 7970Ghz, with a lower price, as it looks on this review?
http://www.anandtech.com/show/7400/the-radeon-r9-280x-review-feat-asus-xfx/19
____________
676754^262144+1 is prime |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
|
Is the 290x still not getting tasks from the server? Sorry about dropping the ball on that.
____________
My lucky number is 75898524288+1 |
|
|
288larssonVolunteer tester
Send message
Joined: 17 Apr 10
Posts: 136
ID: 58815
Credit: 5,967,247,909
RAC: 3,284,288
|
Is the 290x still not getting tasks from the server? Sorry about dropping the ball on that.
Yes
2013-12-25 14:47:36 | PrimeGrid | Requesting new tasks for ATI
2013-12-25 14:47:37 | PrimeGrid | Scheduler request completed: got 0 new tasks
runing using app_info. Works |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
Is the 290x still not getting tasks from the server? Sorry about dropping the ball on that.
Yes
2013-12-25 14:47:36 | PrimeGrid | Requesting new tasks for ATI
2013-12-25 14:47:37 | PrimeGrid | Scheduler request completed: got 0 new tasks
runing using app_info. Works
Ok, someone will have to catch me online sometime in the chatroom so we can do some live tests to see what the problem is.
____________
My lucky number is 75898524288+1 |
|
|
Ross*Send message
Joined: 22 Apr 10
Posts: 66
ID: 59181
Credit: 1,185,092,437
RAC: 27,456
     
|
|
Hi
re the ATI 290x
when I first got one they just trashed then , now will not get tasks.
Boinc recognised the GPU [Hawaii] but still no tasks. I have the latest drivers as well.
Perhaps it needs a app_info , but I have know idea what to put in it.
Cheers
Ross*
____________
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
You can help by doing a few simple tests:
1) Try running a BOINC PPS sieve task on your new ATI GPU. It should fail to work in the same way as the GFN task fails.
1) Yes
Did that "yes" mean it worked or that, as expected, it didn't work?
____________
My lucky number is 75898524288+1 |
|
|
288larssonVolunteer tester
Send message
Joined: 17 Apr 10
Posts: 136
ID: 58815
Credit: 5,967,247,909
RAC: 3,284,288
|
You can help by doing a few simple tests:
1) Try running a BOINC PPS sieve task on your new ATI GPU. It should fail to work in the same way as the GFN task fails.
1) Yes
Did that "yes" mean it worked or that, as expected, it didn't work?
it didn't work
|
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
You can help by doing a few simple tests:
1) Try running a BOINC PPS sieve task on your new ATI GPU. It should fail to work in the same way as the GFN task fails.
1) Yes
Did that "yes" mean it worked or that, as expected, it didn't work?
it didn't work
Ok, next I need somebody with a 290x to contact me in the Chatzy chat room so we can do some live testing so I can see what's happening on the server when we're both online.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
|
At the present time, you'll need to use app_info to use Hawaii (290) GPUs. There doesn't appear to be an easy fix.
____________
My lucky number is 75898524288+1 |
|
|
Ross*Send message
Joined: 22 Apr 10
Posts: 66
ID: 59181
Credit: 1,185,092,437
RAC: 27,456
|
At the present time, you'll need to use app_info to use Hawaii (290) GPUs. There doesn't appear to be an easy fix.
Hi
app_info please
Has anyone run a OCL task on a r9 290x?
Cheers
Ross*
____________
|
|
|
288larssonVolunteer tester
Send message
Joined: 17 Apr 10
Posts: 136
ID: 58815
Credit: 5,967,247,909
RAC: 3,284,288
|
Hi
app_info please
Has anyone run a OCL task on a r9 290x?
Cheers
Ross*
hi my app_info.xml file is here:
<app_info>
<app>
<name>genefer</name>
<user_friendly_name>Genefer</user_friendly_name>
</app>
<file_info>
<name>primegrid_genefer_3_1_2_7_2.10_windows_intelx86__atiGFN.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>genefer</app_name>
<version_num>210</version_num>
<api_version>7.0.64</api_version>
<avg_ncpus>1.000000</avg_ncpus>
<max_ncpus>1.000000</max_ncpus>
<file_ref>
<file_name>primegrid_genefer_3_1_2_7_2.10_windows_intelx86__atiGFN.exe</file_name>
<main_program/>
</file_ref>
<platform>windows_intelx86</platform>
<coproc>
<type>ATI</type>
<count>1.000000</count>
</coproc>
</app_version>
<app>
<name>genefer_wr</name>
<user_friendly_name>Genefer (World Record)</user_friendly_name>
</app>
<file_info>
<name>primegrid_genefer_3_1_2_7_2.10_windows_intelx86__atiGFNWR.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>genefer_wr</app_name>
<version_num>210</version_num>
<api_version>7.0.64</api_version>
<avg_ncpus>1.000000</avg_ncpus>
<max_ncpus>1.000000</max_ncpus>
<file_ref>
<file_name>primegrid_genefer_3_1_2_7_2.10_windows_intelx86__atiGFNWR.exe</file_name>
<main_program/>
</file_ref>
<platform>windows_intelx86</platform>
<coproc>
<type>ATI</type>
<count>1.000000</count>
</coproc>
</app_version>
<app>
<name>pps_sr2sieve</name>
<user_friendly_name>Proth Prime Search (Sieve)</user_friendly_name>
</app>
<file_info>
<name>primegrid_tpsieve_1.39_windows_intelx86__atiPPSsieve.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>pps_sr2sieve</app_name>
<version_num>124</version_num>
<platform>windows_x86_64</platform>
<avg_ncpus>1</avg_ncpus>
<max_ncpus>1</max_ncpus>
<flops>1.0e11</flops>
<coproc>
<type>ATI</type>
<count>1</count>
</coproc>
<cmdline>-m 64</cmdline>
<file_ref>
<file_name>primegrid_tpsieve_1.39_windows_intelx86__atiPPSsieve.exe</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info> |
|
|
Ross*Send message
Joined: 22 Apr 10
Posts: 66
ID: 59181
Credit: 1,185,092,437
RAC: 27,456
|
|
Hi
app_info please
Has anyone run a OCL task on a r9 290x?
Cheers
Ross*
hi my app_info.xml file is here:
<app_info>
<app>
<name>genefer</name>
<user_friendly_name>Genefer</user_friendly_name>
</app>
<file_info>
<name>primegrid_genefer_3_1_2_7_2.10_windows_intelx86__atiGFN.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>genefer</app_name>
<version_num>210</version_num>
<api_version>7.0.64</api_version>
<avg_ncpus>1.000000</avg_ncpus>
<max_ncpus>1.000000</max_ncpus>
<file_ref>
<file_name>primegrid_genefer_3_1_2_7_2.10_windows_intelx86__atiGFN.exe</file_name>
<main_program/>
</file_ref>
<platform>windows_intelx86</platform>
<coproc>
<type>ATI</type>
<count>1.000000</count>
</coproc>
</app_version>
<app>
<name>genefer_wr</name>
<user_friendly_name>Genefer (World Record)</user_friendly_name>
</app>
<file_info>
<name>primegrid_genefer_3_1_2_7_2.10_windows_intelx86__atiGFNWR.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>genefer_wr</app_name>
<version_num>210</version_num>
<api_version>7.0.64</api_version>
<avg_ncpus>1.000000</avg_ncpus>
<max_ncpus>1.000000</max_ncpus>
<file_ref>
<file_name>primegrid_genefer_3_1_2_7_2.10_windows_intelx86__atiGFNWR.exe</file_name>
<main_program/>
</file_ref>
<platform>windows_intelx86</platform>
<coproc>
<type>ATI</type>
<count>1.000000</count>
</coproc>
</app_version>
<app>
<name>pps_sr2sieve</name>
<user_friendly_name>Proth Prime Search (Sieve)</user_friendly_name>
</app>
<file_info>
<name>primegrid_tpsieve_1.39_windows_intelx86__atiPPSsieve.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>pps_sr2sieve</app_name>
<version_num>124</version_num>
<platform>windows_x86_64</platform>
<avg_ncpus>1</avg_ncpus>
<max_ncpus>1</max_ncpus>
<flops>1.0e11</flops>
<coproc>
<type>ATI</type>
<count>1</count>
</coproc>
<cmdline>-m 64</cmdline>
<file_ref>
<file_name>primegrid_tpsieve_1.39_windows_intelx86__atiPPSsieve.exe</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>
HI
Thanks for the app_info. works a treat , will take 7hrs on the short ones , so not the fastest
Cheers
Ross*
____________
|
|
|
axnVolunteer developer
Send message
Joined: 29 Dec 07
Posts: 285
ID: 16874
Credit: 28,027,106
RAC: 0
       
|
|
can i download an standalone ocl app from somewhere in order to benchmark my Radeon? |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
can i download an standalone ocl app from somewhere in order to benchmark my Radeon?
The first post in the thread New genefer (3.x.x) apps now available for testing contains the links to download all of the latest Genefer apps.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
I have an iMac with GTX 775M (Kepler) graphics (has double precision), but the server won't send GPU tasks.
I have BOINC GPU "Run always" on, PrimeGrid Use GPU = yes, Genefer GFN-Short "CPU" and "Nvidia OpenCL" checked, Genefer GFN-WR "Nvidia OpenCL" checked.
Collatz runs on the GPU without problem.
primegrid_genefer_3_2_4_0_3.04_x86_64-apple-darwin__cpuGFN -b -x fma runs fine.
But the server only sends Genefer 3.04 (cpuGFN) tasks.
What am I missing? |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14036
ID: 53948
Credit: 475,998,733
RAC: 222,579
|
I have an iMac with GTX 775M (Kepler) graphics (has double precision), but the server won't send GPU tasks.
I have BOINC GPU "Run always" on, PrimeGrid Use GPU = yes, Genefer GFN-Short "CPU" and "Nvidia OpenCL" checked, Genefer GFN-WR "Nvidia OpenCL" checked.
Collatz runs on the GPU without problem.
primegrid_genefer_3_2_4_0_3.04_x86_64-apple-darwin__cpuGFN -b -x fma runs fine.
But the server only sends Genefer 3.04 (cpuGFN) tasks.
What am I missing?
I'm not sure. From my perspective, everything looks okay. If you happen to catch me online in our Chatzy chatroom (the link's over there on the left), I can turn on detailed logging while I'm watching and I can see exactly what the server is doing.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
I'll wait a couple of days until the Genefer CPU tasks are complete, so that the request will start out fresh. |
|
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 842
ID: 164101
Credit: 306,514,171
RAC: 5,382
|
|
Here is a trick to speed up genefer on AMD GPUs: set the environment variable AMD_OCL_BUILD_OPTIONS to -cl-opt-disable !
On a R9 270 (Pitcairn) with Catalyst 14.4, genefer is 10% faster:
geneferocl-windows.exe
658332^524288+1 Time: 1.01 ms/mul. Err: 0.1875 3050541 digits
538452^1048576+1 Time: 2.11 ms/mul. Err: 0.1953 6009544 digits
440400^2097152+1 Time: 4.18 ms/mul. Err: 0.1719 11836006 digits
360204^4194304+1 Time: 8.53 ms/mul. Err: 0.1719 23305854 digits
294612^8388608+1 Time: 18.3 ms/mul. Err: 0.1641 45879398 digits
Genefer Mark = 48.
set AMD_OCL_BUILD_OPTIONS=-cl-opt-disable
geneferocl-windows.exe
658332^524288+1 Time: 879 us/mul. Err: 0.1875 3050541 digits
538452^1048576+1 Time: 1.85 ms/mul. Err: 0.1953 6009544 digits
440400^2097152+1 Time: 3.52 ms/mul. Err: 0.1719 11836006 digits
360204^4194304+1 Time: 8.14 ms/mul. Err: 0.1719 23305854 digits
294612^8388608+1 Time: 17.2 ms/mul. Err: 0.1641 45879398 digits
Genefer Mark = 53.
That may be different on other GPU/driver but according to the generated code it should be faster on many architectures.
AMD must work on the OpenCL "optimizer"!!!
|
|
|
pschoeferVolunteer developer
Volunteer tester
Send message
Joined: 20 Sep 05
Posts: 686
ID: 845
Credit: 3,008,811,265
RAC: 591,044
           
|
|
After upgrading to Catalyst 14.9, GeneferOCL doesn't find my R9-280X anymore. Both v3.2.3 and v3.2.5-dev report
No AMD compatible OpenCL device found.
It's not a problem with OpenCL in general, as I'm able to run wwwwcl (which does not work with earlier drivers). wwwwcl64 -l reports
List of available platforms and devices
Platform 0 is a Advanced Micro Devices, Inc. AMD Accelerated Parallel Processing, version OpenCL 1.2 AMD-APP (1573.4)
Device 0 is a Advanced Micro Devices, Inc. Tahiti
____________
  |
|
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 842
ID: 164101
Credit: 306,514,171
RAC: 5,382
|
After upgrading to Catalyst 14.9, GeneferOCL doesn't find my R9-280X anymore.
It works on a R9-270 and a task was validated:
http://www.primegrid.com/result.php?resultid=574394388
It's not a problem with OpenCL in general, as I'm able to run wwwwcl (which does not work with earlier drivers). wwwwcl64 -l reports
List of available platforms and devices
Platform 0 is a Advanced Micro Devices, Inc. AMD Accelerated Parallel Processing, version OpenCL 1.2 AMD-APP (1573.4)
Device 0 is a Advanced Micro Devices, Inc. Tahiti
On my computers, wwwwcl finds two platforms: the CPU
Platform 0 is a Intel(R) Corporation Intel(R) OpenCL, version OpenCL 1.2
with no device or a HD Graphics one
and the GPU
Platform 1 is a Advanced Micro Devices, Inc. AMD Accelerated Parallel Processing, version OpenCL 1.2 AMD-APP (1573.4)
Device 0 is a Advanced Micro Devices, Inc. Pitcairn
I'm going to compare the source codes to find why the list of devices may be different.
|
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1963
ID: 352
Credit: 6,398,257,295
RAC: 2,430,869
                      
|
After upgrading to Catalyst 14.9, GeneferOCL doesn't find my R9-280X anymore. Both v3.2.3 and v3.2.5-dev report
No AMD compatible OpenCL device found.
Works for me on R280, both PRPNet and BOINC Genefer.
IIRC, I haven't uninstalled previous Catalyst prior to upgrading to 14.9 if it makes any difference.
>geneferocl -l
geneferocl 3.2.3 (Windows/OpenCL/32-bit)
Running on platform 'AMD Accelerated Parallel Processing', device 'Tahiti', version 'OpenCL 1.2 AMD-APP (1573.4)' and driver '1573.4 (VM)'.
primegrid_genefer_3_2_3_0_3.03_windows_intelx86__atiGFNWR.exe -l
geneferocl 3.2.3 (Windows/OpenCL/32-bit)
Running on platform 'AMD Accelerated Parallel Processing', device 'Tahiti', version 'OpenCL 1.2 AMD-APP (1573.4)' and driver '1573.4 (VM)'.
____________
My stats |
|
|
pschoeferVolunteer developer
Volunteer tester
Send message
Joined: 20 Sep 05
Posts: 686
ID: 845
Credit: 3,008,811,265
RAC: 591,044
|
|
Problem solved. Removing the driver in safe mode and reinstalling fixed it.
____________
|
|
|
