Welcome to MilkyWay@home

Posts by Jim1348

1) Message boards : MilkyWay@home Science : Astronomers Detected a Huge New Structure in The Milky Way, And Don't Know What It Is (Message 71057)
Posted 7 Aug 2021 by Jim1348
In August 2019, the researchers used FAST to look for HI radio emissions, and the data revealed what appeared to be a large structure. When they calculated how fast the structure was moving, they got a surprise: its velocity was consistent with a distance of around 71,750 light-years from the galactic center - the outer regions of the galaxy.

That distance - farther than any known spiral arms in that region of the galaxy - would mean the thing is absolutely huge, with a size of around 3,590 light-years in length and 675 light-years in width, based on the FAST data. But then, when the researchers combined their findings with data from the HI4PI all-sky HI survey, they found that it could be even larger still - as much as around 16,300 light-years in length.

That would make it even more colossal than the gas structure known as Gould's Belt, which was recently found to be 9,000 light-years long.

This was cited by QuantumHelos on the Einstein forum. It seems applicable here too.
2) Message boards : Number crunching : N-Body Simulation MT (Message 71041)
Posted 30 Jul 2021 by Jim1348
Only the project can tell you why the limit. But you can run on any number of virtual cores (threads) up to that limit using an "app_config.xml" file.
This is for running on 12 cores, with one work unit per core; change as desired.


<cmdline>--nthreads 1</cmdline>

You create it in a text processor (Notepad), use the "save as" function to save it as an ".xml" file, and place it in the project data folder.
Then restart BOINC or read in the file to activate it.
I usually just reboot. It works on Windows or Linux.
3) Message boards : News : New Poll Regarding GPU Application of N-Body (Message 71036)
Posted 28 Jul 2021 by Jim1348
Nvidia provides a CUDA sample that looks closely related to the N-body problem.

On my computer, it is in directory:

C:\ProgramData\NVIDIA Corporation\CUDA Samples\v11.4\5_Simulations

Its name is nbody.

That must be in the toolkit. They have some info on it:

If this could make it go faster on CUDA, I am all for it. There are plenty of Nvidia cards around.
They could use the AMD cards for the separation work units.
4) Message boards : Number crunching : Feeder not running (Message 71027)
Posted 27 Jul 2021 by Jim1348
I have not been able to get any CPU work or GPU work for the past few hours.
5) Message boards : Number crunching : Can you add option to not run CPU versions of applications for which GPU versions are available? (Message 71017)
Posted 23 Jul 2021 by Jim1348
Yes Boinc does have problems cpu and gpu scheduling and yes you can run both on the same pc with fewer problems...load up a 2nd copy of Boinc in a different directory and run one for cpu tasks and one for gpu tasks. It's not as easy as it sounds but not that difficult either.

I finally got around to doing it. I have done it before when needed, and it does work fine, at least on Linux.

I think that is the hard way though. The project could provide that option I would think.
I doubt that very many people will do the double instances of BOINC trick. It is a bit involved the first time.
6) Message boards : News : Separation Validator Updates/Brief Server Outage(s) (Message 71012)
Posted 23 Jul 2021 by Jim1348
Having not done MW GPU for a while, I thought I would jump in and see.

It is working OK for the moment on my RX 570 under Ubuntu 20.04.2 (latest 21.20 drivers) on a Ryzen 3600.
7) Message boards : Number crunching : Do Windows machines prefer other Windows for validation? (Message 70842)
Posted 31 May 2021 by Jim1348
Yes, that is so. I was just a little surprised that it did not occur the other way, but now I see that it has.
That is, when I get an invalid on the Windows machine, they are validated on two other Linux machines.

So it works both ways. But given the preponderance of Windows PCs, I am a little better off running it that way and getting fewer invalids.
The speed seems to be about the same on either, though I have not done really long-term tests to find out.

Thanks for your input.
8) Message boards : Number crunching : Do Windows machines prefer other Windows for validation? (Message 70840)
Posted 29 May 2021 by Jim1348
I should note that I am running the 1.80 N-Body using one core per work unit, as limited by an app_config.xml file.
All available cores are running, reserving one or two cores to support a GPU.

On the Linux machine I now have 19 invalids, and all have been validated by two Windows machines.
So I think it is clear by now that Windows machines prefer to validate against each other.

Also, I can compare Windows v. Linux on two Ryzen 3600s.
On the Linux (Ubuntu 20.04.2) machine, I see a 9% invalid rate, and a 21% error rate.
On the Win10 machine, I see a 0% invalid rate and also a 21% error rate.

So I will run them on Windows.
9) Message boards : Number crunching : Do Windows machines prefer other Windows for validation? (Message 70828)
Posted 22 May 2021 by Jim1348
I have four invalids on my Ryzen 3600 running Ubuntu 20.04.2.

That is not a large number, and I am not concerned about the amount. But they all validated against other Windows machines.

I also have a number on my Ubuntu machine that validated against Windows machines, so it is not exactly a bug. Maybe it is just a preference?
10) Message boards : Number crunching : problem with de_nbody tasks never finishing (Message 70728)
Posted 10 Apr 2021 by Jim1348
After switching to MS Defender it didn't take long for a Milkyway@home N-Body Simulation 1.76 (3 CPUs) hangup to occur. But I had forgotten to exclude the BOINC folders.

I haven't excluded them either. But you seem to be running mobile CPUs. Are the work units being suspended? I run my machines 24/7, since they are dedicated.
It could be one of the power-down tricks that Intel or Microsoft uses that causes the problem. I set my power options to "high performance" mode.
Good luck.
11) Message boards : Number crunching : problem with de_nbody tasks never finishing (Message 70725)
Posted 10 Apr 2021 by Jim1348
I have never seen the problem, and though I have not run a very large number of N-body, I have probably done enough to expect to see it.

I always suspect AV software when something is hanging up. It may be the "real time protection", which often still monitors processes even if you exclude the BOINC Data folder.
I use Microsoft Defender for Win10 and don't have problems with it.
12) Message boards : Number crunching : I gotta remember... (Message 70616)
Posted 19 Feb 2021 by Jim1348
My laptop is the only PC running on Winders and it's Win 7 Pro, it will not run Winders 10. This PC does have Winders 10 installed on a separate M.2 SSD, but is only for gaming and I haven't been gaming for over a month now. Currently I'm running Linux Mint exclusively for now.

I used to assume that Windows 7 did the same as Windows 10. But I found otherwise on at least one project, MLC, on their GPU work units strangely enough. Win10 had a big advantage.
And I suspect Win 10 might have some advantage here on the nBody CPU work also. So I upgraded to Win 10, a bit reluctantly, and am using it fine now.
13) Message boards : Number crunching : I gotta remember... (Message 70612)
Posted 19 Feb 2021 by Jim1348
My current RAC is slightly higher than it was before we had that outage a few weeks ago. It took less time to get there this time than it took to get there before the outage. So my app_config.xml has done a bit of good, the way I have it set up.

Thanks. That is the way I would like to run it anyway.

I am still puzzling a bit over Windows 10 vs. Linux. I think Win10 does a little better with all cores running on multi-core than Linux does on all cores.
I am not so sure on single cores; they are close. I want to use my Linux machines in that case. It is probably not worth investigating further.
14) Message boards : Number crunching : I gotta remember... (Message 70607)
Posted 19 Feb 2021 by Jim1348
I have one more question: Will allowing more cores for the MT tasks and less MTs running be more efficient, or does it really matter?

I have spent some considerable time looking at that, using an app_config.xml to limit my machines to a single core, and then allowing them to run on all of them (or at least more than one).

It is a tricky answer. For CPU efficiency per se, single core is best, with an efficiency on my Ryzen 3000 series being around 99% on either Win10 or Ubuntu. But you of course get them done faster with multi-core. However, the CPU efficiency is less with multi-core, around 70 to 80%. And the curious thing is that the total CPU run time is less on multi-core. I would expect it to be more, since the efficiency is less. But that is the tricky part: multi-core ties up all the CPU threads, so that they are not available to other projects. You are really wasting CPU capability. But if MW is the only project you want to run, use multi-core. If you want to run other projects also, use single core.

EDIT: The better way to compare them would really be to run a machine for a long time on single core (all of them), and then compare the RAC to the same machine running multi-core on all the cores. It could turn out that the total output was still greater on single-core.
But that is a long test.
15) Message boards : Number crunching : Can you add option to not run CPU versions of applications for which GPU versions are available? (Message 70580)
Posted 13 Feb 2021 by Jim1348
That can be achieved by setting up a second location preference.

But that requires two PCs. I don't think you can exclude them on a single PC with a single location.
That would be useful, though I think BOINC has scheduling/priority problems when you try to run both CPU and GPU work units on the same PC anyway.
16) Message boards : Number crunching : Run Multiple WU's on Your GPU (Message 70342)
Posted 13 Jan 2021 by Jim1348
I would say Nvidia cards would be better used for GPUGrid, but they have no work available....
Nvidia is cerainly not efficient with MW. Folding has a CUDA app now, and enough work. They are past their shortages.
You can run the GPU on folding, and by eliminating the CPU slot (enabled by default) run BOINC on the CPU if you want to.
17) Questions and Answers : Unix/Linux : GPU tasks error out on Linux (Message 70191)
Posted 24 Nov 2020 by Jim1348
I did not have many errors on my RX 570 until I upgraded to Ubuntu 20.04.1, which requires the latest (20.45) drivers.
Now I get many zero-second fails, while the others complete normally.

The same thing happens on Einstein, so it is not just MW. It seems to be a driver problem, but maybe MW can work around it?
18) Message boards : Number crunching : GPU upgrade time (Message 70184)
Posted 22 Nov 2020 by Jim1348
They are now using the ATLAS system to detect the potential impacts.

If you want something to crunch, the Asteroids project is using the ATLAS data to determine various characteristics of asteroids.
19) Message boards : Number crunching : GPU upgrade time (Message 70178)
Posted 20 Nov 2020 by Jim1348
By the way, I tried a GTX 970 too, and quickly gave up on it.
I am running a few on my RX 570 under Ubuntu 20.04.1, just to check out the drivers, which are not as easy to install as Nvidia.
20) Message boards : Number crunching : GPU upgrade time (Message 70177)
Posted 20 Nov 2020 by Jim1348
Any advise would be appreciated.

Never use Nvidia here. The AMD cards are best. It is in part due to their better (or more compatible) OpenCl, and also the dual-precision performance.

The last time I checked, you were still being limited by the timeouts after downloading a group of work units. Buying an expensive card just meant you got a more expensive wait time. Maybe that has been fixed. Otherwise, I don't think you will do much better than an RX 570, certainly not for the cost.

But I am waiting for the nBody work to be put on GPUs before jumping in again, if it ever happens. Good luck.

Next 20

©2021 Astroinformatics Group