Posts by mmonnin

1) Message boards : GPUs : 8.22 seti@home_v8 (opencl_nvidia_SoG) "Waiting to Run" (Message 92804)
Posted 2 days ago by mmonnin
Post:
Windows 10 probably updated your driver to a version that does not include OpenCL support.
2) Message boards : Questions and problems : Running boinc on Vast.AI using boinc client docker. (Message 92724)
Posted 10 days ago by mmonnin
Post:
BOINC should be in a Ubuntu repository. The BOINC manager should also install as well giving you a GUI. The last time installed BOINC from the default repository it came installed BOINC v7.9.3.

This should start the client from terminal. "/etc/init.d/boinc-client start" I start/restart clients this way.

The data directory for Ubuntu should be /var/lib/boinc-client unless changed. remote_hosts.cfg should be saved there.

I believe the docker is an interface between science exe files and BOINC giving it check pointing and other BOINC functions vs integrating science and BOINC functions into one exe.
3) Message boards : Questions and problems : Windows install issues (Message 92628)
Posted 19 days ago by mmonnin
Post:
It needs to be installed as a service?
4) Message boards : Questions and problems : need help debugging a problem: Linux 7.16.1 (Message 92627)
Posted 19 days ago by mmonnin
Post:
I didn't mean with the bit of code that you could change the output of BOINC. BOINC will always follow the compute capability, driver version, RAM size and estimated FLOPs sequence to designate the device IDs, with the best GPU always being device 0, the second best device 1 etc. Of course if all devices are the same for all four comparison points, then it gets tricky.

I got the code entry from this Stack Overflow thread, while this thread in the Nvidia Dev forums suggests that the device ID numbering isn't always following the PCIe slot order, so don't just think that because a GPU is in the slot nearest the CPU, that that device gets DeviceID 0.


BOINC always shows my cards as 2x 1070s even though it is a 1070 and a 1070Ti. It is not ordered by best/fastest/FLOPs. 1070 is in the 1st slot and 1070Ti is in the 2nd slot.
5) Message boards : BOINC Manager : Unable to connect to client upon installation no matter what I do. (Message 92547)
Posted 26 days ago by mmonnin
Post:
You're client is not running if boinc.exe is not running. Thus BOINC Mgr has nothing to connect to.

Was it installed as a service?

I'm not at a PC with BOINC atm but there should only be a few text files in the BOINC data directory (not the install directory) that aren't project specific. Something like stderr I think is the log file.
6) Message boards : Questions and problems : What happened to "requested" and "granted" credits? (Message 92546)
Posted 26 days ago by mmonnin
Post:
WCG does as well.
7) Message boards : BOINC Manager : Boinc Manager won't run on Ubuntu 18.04 after update (Message 92509)
Posted 16 Aug 2019 by mmonnin
Post:
Linux can take an age to boot after updates and filed boots....
In the past I've found the reciting the incantation "If you don't behave yourself you stupid lump of silicone I'll blow Linux away and install Windows CE" loudly and repeated a dozen times has worked ;-)


If CE isn't enough add 2 more Windows OS versions. CE-ME-NT
8) Message boards : GPUs : Seems SLI is causing GPU problems (Message 92440)
Posted 9 Aug 2019 by mmonnin
Post:
I put the SLI connector back on a pair of gtx1070 and have about 50 error'ed out GPUGrid tasks since. I assume the problem is SLI related but other people are reporting bad workunits at the same time over there.

Is this a known problem? Maybe it is particular to the CUDA implementation by GPUGrid. Anyway it is coming off.

[EDIT] Well what do you know! There is a suggestion they failed to update some security certificate. Murphys law: things left to themselves tend to get worst.

Leaving my SLI connector off just to be sure.


The GPUGrid Windows licence has expired as well.

Project specific errors should be posted on project sites. There were at least 4 separate threads about tasks with errors as of 2 days ago.
9) Message boards : Questions and problems : resource "share' not working as expected when set to zero (Message 92371)
Posted 5 Aug 2019 by mmonnin
Post:
Not every project updates their server code in sync with BOINC releases - many of them leave it untouched for months or years. For a long time, there were reports that specific projects wouldn't allow the special 0 value to be entered via their websites.


YoYo for some reason doesn't allow 0 so its never a backup project for me. Admins have confirmed 0 is not allowed. Thats the only project I've come across that doesn't allow 0.
10) Message boards : Questions and problems : Completing tasks but project information not updating (Message 92348)
Posted 31 Jul 2019 by mmonnin
Post:
There has been some recent credit for QCN Taiwan but its been a few weeks. I think the QCN projects also need the location of the device set in the profile for tasks to receive credit.
11) Message boards : Questions and problems : Computing suspended even though laptop is plugged in (Message 92301)
Posted 25 Jul 2019 by mmonnin
Post:
I'm guessing BOINC reads the status of battery or A/C power like it does if the PC is idle or not. BOINC doesn't know if the PC is idle, the OS does and BOINC reads it. So I'm guessing A/C vs battery might be the same thing.
12) Message boards : Projects : How to link accounts? (Message 92287)
Posted 24 Jul 2019 by mmonnin
Post:
They are all separate.

The only link is via CPID, Cross Project IDs. Mine is 9deac61bf5b5680cd09f705783700b45. You'll have to use the same email between projects and attach to the same host and they will end up with the same CPID. The forums here isn't a project so there's nothing to link. External stats like Free-DC can track by CPID.

I don't understand this part " BOINC universe @ home account, and link those 2 with the universeathome.pl account". The Universe@Home Project site is at universeathome.pl
13) Message boards : BOINC client : Wish List: Allow {app d=N} to assign different GPU WUs to same GPU (Message 92262)
Posted 22 Jul 2019 by mmonnin
Post:
that already works by default but it can be messy (inefficieint)

with MW set to 0.19 and Einstein set to 0.33 I ended up after while with one MV taking 0nly 1/5 of gpu and one Einstein taking only 1/3 of the same one and wasted gpu time.

Normally 3 Einstein on one gpu or 5 MW but if both enabled then things get haphazard quickly.


That doesn't control which task runs on which GPU, which is the desire.
14) Message boards : Projects : QuChemPedIA (Message 92261)
Posted 22 Jul 2019 by mmonnin
Post:
The project title at WUProp now includes 'Private'.
15) Message boards : Projects : Combining BOINC with cryptocurrency (Message 92260)
Posted 22 Jul 2019 by mmonnin
Post:
if researchers had to pay for users electricity they'd take the crunching private and there'd be no more project for users. They turn to BOINC because they often have no extra funding for computing. Some projects are funded by project admins w/o grants or outside funding.
16) Message boards : Questions and problems : CPU Temperature vs Performance (Message 92259)
Posted 22 Jul 2019 by mmonnin
Post:
Drop the CPU core voltage a bit to reduce frequency. You should be able to reduce wattage a decent amount for little performance loss.
17) Message boards : Questions and problems : Option for downloading and computing only one task at all (Message 92210)
Posted 17 Jul 2019 by mmonnin
Post:
I haven't tried this, but I suppose you could set the resource share for all projects to 0. Additionally, you could set all projects to No New Tasks. Then, whenever you want to run Boinc, you could select the project you want to run and Allow New Tasks. With the resource share at 0 it should only download enough tasks for the processors you have active. When you don't want that project to download any further tasks you can set it back to No New Tasks. If you don't want all CPU cores to crunch tasks, follow Les' advice and reduce the percentage.

I know this seems to make running Boinc quite a bit more manual than automatic, but I'm guessing you are okay with that.


This works. I do this to have a backup project in case the project I am running runs out of work. With 0 resource share the task count = CPU count. No tasks will be waiting to run.
18) Message boards : Questions and problems : boinc on debian server does not work (Message 92208)
Posted 17 Jul 2019 by mmonnin
Post:
There hasn't been any Linux work at climateprediction for ages.
And you don't really have enough memory for that project anyway.


Ages? I am still running some now. They started like 22 days ago but thats the length of the task.
19) Message boards : Projects : LHC@home moves SixTrack 5.02 to production (Message 92126)
Posted 10 Jul 2019 by mmonnin
Post:
Recent announcement from Alessio Mereghetti... https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=5011

This is a major rebuild of LHC's SixTrack app moved from testing to production. WinXP is not supported but they're working on fixing that.
ATM there are 750K tasks available. Get 'em while they're hot!!


Thanks, was a fan of CERN hoping they could find a small black hole to prove the critics were correct and they could inadvertently destroy the world.

..I quit reading when I got to the "huge source files and Fortran90" statement.

They need to get on the GPU band wagon.
That being said, I am crunching on their stuff.


GPU computing is not the solution for every compute problem. Otherwise why don't we just get rid of CPUs all together and just use GPUs as the main processor in computers? They are better than each other at different types of calculations. GPUs excel at specific types of calculations. If there are branching in the code then GPUs suffer and the math needs to be highly parallel to make use of small 'cores' on GPUs.
20) Message boards : Questions and problems : Had to limit # concurrent tasks due to "hi priority" WCG downloads (Message 92125)
Posted 10 Jul 2019 by mmonnin
Post:
Use an app_config to set cpu usage on the GPU tasks to 0.1 CPUs so they will still run even if CPU tasks become high priority. This does not change how much CPU the GPU tasks will actually use. You can reduce the # of CPU tasks via BOINC Mgr setting or Process Lasso to allow GPUGrid tasks to get all the CPU cycles it needs.


Next 20

Copyright © 2019 University of California. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.