The work is available but according to the last contact log of your Mac you won't get any because your GPU driver or rather CUDA version is to new.
2016-10-07 15:35:46.7508 [PID=5908 ] [version] Checking plan class 'BRP4G-cuda32-OSX'
2016-10-07 15:35:46.7508 [PID=5908 ] [version] parsed project prefs setting 'gpu_util_brp': 0.500000
2016-10-07 15:35:46.7508 [PID=5908 ] [version] CUDA version required max: 6049, supplied: 8000
2016-10-07 15:35:46.7508 [PID=5908 ] [version] Checking plan class 'BRP4G-opencl-ati-lion'
2016-10-07 15:35:46.7508 [PID=5908 ] [version] OS version required max: 140000, supplied: 150400
2016-10-07 15:35:46.7508 [PID=5908 ] [version] Checking plan class 'BRP4G-Beta-cuda32-OSX'
2016-10-07 15:35:46.7508 [PID=5908 ] [version] parsed project prefs setting 'gpu_util_brp': 0.500000
2016-10-07 15:35:46.7508 [PID=5908 ] [version] CUDA version required max: 6049, supplied: 8000
Thanks for the information. I actually started my system with CUDA Version 7.xx.xx Driver on El Capitan 10.11.3. Was told by TBar at SETI to keep it Updated. (Especially with his newly developed CUDA75 App for MACs at SETI due to OpenCL issues on MAC.)
I was wondering why my MAC was only getting the BRP6 CUDA55 Units here... Oh well... Any idea when Einstein will "catch up" to my Driver Version??? I can't step it back down past the 7.xx.xx Driver's anyway...
Bad news - on Windows x64 systems all these tasks error right at the beginning.
I think the problem is in the x86_64 version of the app only. 32-bit one should be OK. Can you please issue the x86 version to x86_64 machines too ? This would be similar to what was done on BRP6.
So far so good, Scotty! Seems to work well on x64 systems too.
First performance results (BRP4G cuda32 -> cuda55):
750 Ti (XP32, Core i7 860): 2600 s -> 2000 s
750 Ti (W7 x64, Xeon E3-1275 v5): 2400 s -> 1900 s
Tesla K20c (W10, Xeon E5-2650 v3): 1300 s -> 1000 s
I'm running 2 at a time as I found that was the best for my machines.
I'm only seeing about a 12% improvement on my 980Tis of the cuda 55 over the cuda 32 (22 mins vs 25 minutes)
Interestingly if you pair a Arecibo cuda 55 with a Parkes cuda 55 work unit, the Arecibo work unit finishes 25% faster than 2 Arecibo cuda 32 (18 minutes vs 25 mins)
edit...
These are all version 1.57 I have several v1.56 on another machine that I will test right now
Holmis wrote:The work is
)
Thanks for the information. I actually started my system with CUDA Version 7.xx.xx Driver on El Capitan 10.11.3. Was told by TBar at SETI to keep it Updated. (Especially with his newly developed CUDA75 App for MACs at SETI due to OpenCL issues on MAC.)
I was wondering why my MAC was only getting the BRP6 CUDA55 Units here... Oh well... Any idea when Einstein will "catch up" to my Driver Version??? I can't step it back down past the 7.xx.xx Driver's anyway...
TL
TimeLord04
Have TARDIS, will travel...
Come along K-9!
Join SETI Refugees
For now I issued the BRP4G
)
For now I issued the BRP4G Cuda 5.5 app versions as "Beta", as these were on BRP6, so you can decide yourself whether to run these or not.
BM
Thanks Bernd, Will give
)
Thanks Bernd,
Will give those a try here soon.
Zalster
Bernd Machenschalk wrote:For
)
Great, thanks! Will check it on a few systems are report back.
Why don't I see this app for Windows on the Applications page?
-----
Bad news - on a Windows x64
)
Bad news - on Windows x64 systems all these tasks error right at the beginning.
I think the problem is in the x86_64 version of the app only. 32-bit one should be OK. Can you please issue the x86 version to x86_64 machines too ? This would be similar to what was done on BRP6.
-----
I deprecated the
)
I deprecated the windows_x86_64 cuda55 version. All Windows clients should get the 32Bir version.
BM
So far so good! Seems to work
)
So far so good, Scotty! Seems to work well on x64 systems too.
First performance results (BRP4G cuda32 -> cuda55):
750 Ti (XP32, Core i7 860): 2600 s -> 2000 s
750 Ti (W7 x64, Xeon E3-1275 v5): 2400 s -> 1900 s
Tesla K20c (W10, Xeon E5-2650 v3): 1300 s -> 1000 s
That's a ~30% improvement!
-----
I'm running 2 at a time as I
)
I'm running 2 at a time as I found that was the best for my machines.
I'm only seeing about a 12% improvement on my 980Tis of the cuda 55 over the cuda 32 (22 mins vs 25 minutes)
Interestingly if you pair a Arecibo cuda 55 with a Parkes cuda 55 work unit, the Arecibo work unit finishes 25% faster than 2 Arecibo cuda 32 (18 minutes vs 25 mins)
edit...
These are all version 1.57 I have several v1.56 on another machine that I will test right now
So version 1.56 is throwing
)
So version 1.56 is throwing out these messages on my computer.
Files that help describe the problem:
C:\Users\Zalster\AppData\Local\Temp\WER355D.tmp.WERInternalMetadata.xml
C:\Users\Zalster\AppData\Local\Temp\WER573F.tmp.appcompat.txt
C:\Users\Zalster\AppData\Local\Temp\WER576F.tmp.mdmp
Read our privacy statement online:
http://go.microsoft.com/fwlink/?linkid=104288&clcid=0x0409
If the online privacy statement is not available, please read our privacy statement offline:
C:\Windows\system32\en-US\erofflps.txt
I'm also not seeing any CPU usage on BOINCtasks. Hope this helps
Edit..
Also get
The instruction at 0x00784548 referenced memory at 0x010b42de. The memory could not be read. Click ok to terminate the program
v1.56 has been deprecated,
)
v1.56 has been deprecated, v1.57 (x86) is sent now and that seems to work on both x86 and x64 systems.
-----