My once a leap year posting seeing as its the 29th of Feb :-)
Farm status
Intel GPUs
All off, although I did have one running overnight
Nvidia GPUs
All off
Parallellas and Pis
All running Seti Beta work
BOINC testing
I mentioned previously that the Pi's haven't been updating their BOINC log files since about October 2015. This seems to have crept into the Debian AMD64 builds as well so I raised Debian bug 815214
Apparently this is caused by Debian changing over to using systemd to start services. There is a work around that is in the Sid and Stretch releases as BOINC 7.6.28. I am waiting for the Raspberry Pi repository to catch up before testing it. I suspect its only a partial fix.
Seti beta testing
As mentioned above the entire Bramble (that's what the Raspberry Pi folks call it when you have more than one) is running Seti Beta work. We have an app that works but we're trying to make it faster. My Pi2's have been taking an average of 27 hours a work unit. The B+ is still working on its one, but will probably come in close to 60 hours and its got a medium overclock (900Mhz).
I also have one of the Parallella's running Seti beta work. They seem to be slower than the Pi2 but only slightly. They all run the same app.
29 February 2016
20 February 2016
20th of February
Farm status
Intel GPUs
All off
Nvidia GPUs
All off
Parallellas and Pis
Crunching Einstein BRP4 work and Seti Beta app testing
We did get one cool day during the week where I had the Intel GPU machines running, but its back to hot and humid again.
Electrical issues
I had an electrician out to replace a failed smoke detector and a broken light switch. When I went to turn the power off I noticed that one of the two earth leakage circuit breakers was in the off position but everything was still running. Resetting it didn't help - it kept tripping. Its been replaced with a temporary 15 amp one as the electrician didn't have a large enough one (20 amp) at the time. He'll be back next week with the necessary parts to finish the job.
When I powered up the farm I managed to trip the new earth leakage circuit breaker, so I had to slowly go through each power board and try and work out which one was causing the fault. It was a brand new 8 socket one I put in the day before. A bit more checking (plugging things in one at a time) revealed one of the new Intel GPU machines. I replaced the power cable and it seems to be working now, When I replaced the PC I didn't replace all the power cables, I just reused the old power cable that was threaded through the shelving unit.
ARM testing
As mentioned previously we're testing a Seti app for the ARM based devices. I have two Pi2's allocated to running the Seti v8.01 Multi-beam app at the moment with run times around 27 hours a work unit. The Pi2's are running at stock speed. The new Pi2 that I got during the week is running Einstein BRP4 work. That brings the "Bramble" as the Raspberry Pi people call them, up to 4 x Pi2's and one B+. I also have one of the two Parallella's running that same app with similar run times to the Pi2's. The B+ looks like it going for 48 hours to complete its first v8.01 work unit. Hopefully the guys looking through the code will be able to improve on the run times.
Other testing
Apart from the ARM testing I also have some issues with Debian Stretch in that the BOINC log files seem to have disappeared. This happened on the Pi's around October last year (also running Stretch). I will be wiping my Intel GPU Linux test machine and reinstall it in an attempt to work out why the log files that used to be in /var/lib/boinc-client are no longer there. BOINC still works fine, I can see the logs via BOINCtasks running on another machine but where they are going I have no idea - neither does the Linux maintainer for BOINC. It may be related to GCC version 5 (which also broke the Einstein BRP4 v1.06 app and the Intel GPU Seti/Linux app).
Intel GPUs
All off
Nvidia GPUs
All off
Parallellas and Pis
Crunching Einstein BRP4 work and Seti Beta app testing
We did get one cool day during the week where I had the Intel GPU machines running, but its back to hot and humid again.
Electrical issues
I had an electrician out to replace a failed smoke detector and a broken light switch. When I went to turn the power off I noticed that one of the two earth leakage circuit breakers was in the off position but everything was still running. Resetting it didn't help - it kept tripping. Its been replaced with a temporary 15 amp one as the electrician didn't have a large enough one (20 amp) at the time. He'll be back next week with the necessary parts to finish the job.
When I powered up the farm I managed to trip the new earth leakage circuit breaker, so I had to slowly go through each power board and try and work out which one was causing the fault. It was a brand new 8 socket one I put in the day before. A bit more checking (plugging things in one at a time) revealed one of the new Intel GPU machines. I replaced the power cable and it seems to be working now, When I replaced the PC I didn't replace all the power cables, I just reused the old power cable that was threaded through the shelving unit.
ARM testing
As mentioned previously we're testing a Seti app for the ARM based devices. I have two Pi2's allocated to running the Seti v8.01 Multi-beam app at the moment with run times around 27 hours a work unit. The Pi2's are running at stock speed. The new Pi2 that I got during the week is running Einstein BRP4 work. That brings the "Bramble" as the Raspberry Pi people call them, up to 4 x Pi2's and one B+. I also have one of the two Parallella's running that same app with similar run times to the Pi2's. The B+ looks like it going for 48 hours to complete its first v8.01 work unit. Hopefully the guys looking through the code will be able to improve on the run times.
Other testing
Apart from the ARM testing I also have some issues with Debian Stretch in that the BOINC log files seem to have disappeared. This happened on the Pi's around October last year (also running Stretch). I will be wiping my Intel GPU Linux test machine and reinstall it in an attempt to work out why the log files that used to be in /var/lib/boinc-client are no longer there. BOINC still works fine, I can see the logs via BOINCtasks running on another machine but where they are going I have no idea - neither does the Linux maintainer for BOINC. It may be related to GCC version 5 (which also broke the Einstein BRP4 v1.06 app and the Intel GPU Seti/Linux app).
14 February 2016
Valentines Day 2016
Farm status
Intel GPU
Off
Nvidia GPUs
Off
Parallellas and Pis
Running Einstein and Seti Beta
Seti Beta testing
We're still trying to get an app that completes work in a reasonable time. My Pi2's are taking on average 28 hours a work unit. The Parallella that is also running Seti Beta is slightly worse. Given the hot weather (today was 37 degrees C) the only computers running are the Parallellas and Pi's.
A couple of the guys are looking at the code and a couple of others are looking at compile/build options. I think we've exhausted the compile/build options as that's where they started so the only other likely way to get speed improvements are through optimising the code. I suggested the use of fftw wisdom files, however the people looking at the code haven't confirmed that it even uses fftw wisdom files.
I ordered another Pi2, 4 official cases and 5 official power adapters. One of the Pi2's red LED was flashing earlier in the week indicating power issues. The cases are out of stock and won't arrive until April. I was planning on re-casing the 3 existing Pi2's and having one for the new Pi2.
Intel GPU
Off
Nvidia GPUs
Off
Parallellas and Pis
Running Einstein and Seti Beta
Seti Beta testing
We're still trying to get an app that completes work in a reasonable time. My Pi2's are taking on average 28 hours a work unit. The Parallella that is also running Seti Beta is slightly worse. Given the hot weather (today was 37 degrees C) the only computers running are the Parallellas and Pi's.
A couple of the guys are looking at the code and a couple of others are looking at compile/build options. I think we've exhausted the compile/build options as that's where they started so the only other likely way to get speed improvements are through optimising the code. I suggested the use of fftw wisdom files, however the people looking at the code haven't confirmed that it even uses fftw wisdom files.
I ordered another Pi2, 4 official cases and 5 official power adapters. One of the Pi2's red LED was flashing earlier in the week indicating power issues. The cases are out of stock and won't arrive until April. I was planning on re-casing the 3 existing Pi2's and having one for the new Pi2.
06 February 2016
6th of February
Farm status
Intel GPUs
Running down their cache of Seti work
Nvidia GPUs
Off
Parallellas and Pis
One Pi2 doing Seti Beta, the rest doing Einstein BRP4 work
Project news - Einstein
It seems Debian released an updated ca-certificates that broke connectivity to the project for people running the Jessie release (the Pi's included). They removed a certificate that the project need to validate against.
Fortunately I haven't updated my B+ that's still running Jessie so it can still connect to the project. The Stretch release doesn't have this issue as it has an updated libcurl.
Project news - Seti
They have now got a Multi Beam v8 app that will run on the Pi's and Pi2. Well I haven't tried it on the B+ but it says its for armv6l architecture. This is over on the beta project. I've run a few work units through without problems so far. Times have ranged from 1000 seconds (task overflowed results) to 100,000 seconds elapsed. The average depending on the angle range seems to be around 92,000 to 100,000 seconds.
There is an optimised version of the app available that has been compiled to make use of the Pi2's armv7l CPU features but haven't tried it yet.
Other stuff
I have been using one of the older i7-3770 machines to test Linux functionality and getting BOINC running on it. That's fine for CPU tasks and all is well but I want to also be able to use the Intel HD graphics. I have installed Beignet on it however when testing the Seti beta app it crashed. We think this maybe due to the fact its using GCC v5.0 and the apps are compiled using an older GCC that isn't compatible. Further testing to be carried out. I couldn't get any OpenCL work from Einstein (but plenty of CPU tasks available).
After getting this working I will try and get the Nvidia drivers running under Linux. If it works out then I could get the Intel GPU and Nvidia GPU machines running Linux rather than Windows.
I have also trialled the Intel HD Graphics 530 drivers under Windows and they produce invalid results. I've tried 3 different driver versions and logged a bug with Intel. They told me to deal with the OEM. So much for Intel support.
Intel GPUs
Running down their cache of Seti work
Nvidia GPUs
Off
Parallellas and Pis
One Pi2 doing Seti Beta, the rest doing Einstein BRP4 work
Project news - Einstein
It seems Debian released an updated ca-certificates that broke connectivity to the project for people running the Jessie release (the Pi's included). They removed a certificate that the project need to validate against.
Fortunately I haven't updated my B+ that's still running Jessie so it can still connect to the project. The Stretch release doesn't have this issue as it has an updated libcurl.
Project news - Seti
They have now got a Multi Beam v8 app that will run on the Pi's and Pi2. Well I haven't tried it on the B+ but it says its for armv6l architecture. This is over on the beta project. I've run a few work units through without problems so far. Times have ranged from 1000 seconds (task overflowed results) to 100,000 seconds elapsed. The average depending on the angle range seems to be around 92,000 to 100,000 seconds.
There is an optimised version of the app available that has been compiled to make use of the Pi2's armv7l CPU features but haven't tried it yet.
Other stuff
I have been using one of the older i7-3770 machines to test Linux functionality and getting BOINC running on it. That's fine for CPU tasks and all is well but I want to also be able to use the Intel HD graphics. I have installed Beignet on it however when testing the Seti beta app it crashed. We think this maybe due to the fact its using GCC v5.0 and the apps are compiled using an older GCC that isn't compatible. Further testing to be carried out. I couldn't get any OpenCL work from Einstein (but plenty of CPU tasks available).
After getting this working I will try and get the Nvidia drivers running under Linux. If it works out then I could get the Intel GPU and Nvidia GPU machines running Linux rather than Windows.
I have also trialled the Intel HD Graphics 530 drivers under Windows and they produce invalid results. I've tried 3 different driver versions and logged a bug with Intel. They told me to deal with the OEM. So much for Intel support.
Subscribe to:
Posts (Atom)