Posts by ritterm
log in
1) Message boards : Number crunching : Reserving a core for Collatz (Message 21503)
Posted 831 days ago by Profile ritterm
mikey wrote:
The key is my system is different from your system so try it, if you like it better then reserve a core, if not then don't. Each pc is different in SOOO many ways that there is no hard and fast answer.

Yes, indeed. My host (Q6600, 4-ish GB RAM, GTX 570, Win7-64) seems to do fine with no dedicated CPU. Thanks for the feedback, guys!
2) Message boards : News : Sieve in Production; Large and Solo deprecated (Message 21498)
Posted 832 days ago by Profile ritterm
JohnMD wrote:
Please have a little patience.
The admins' sieve is so efficient - and rewarding - that they can't cope with our support. Give them a chance !

I have worked this project many times through the years, am familiar with its ups and downs, and have no problem rolling with the changes. I was just asking the question since Slicker brought it up. Jon is a model admin who does his job at least as well as those at other projects. No impatience here... :-)
3) Message boards : News : Sieve in Production; Large and Solo deprecated (Message 21493)
Posted 832 days ago by Profile ritterm
Slicker wrote:
Well, when there are several hundred people who burn through hundreds of WUs at a time with errors and keep doing it, the server gets bogged down a little.

Isn't there a way for the server to blacklist hosts that "behave badly"? I'm sure I've seen this at other projects. Maybe it's not practical here or there's another good reason it's not implemented.
4) Message boards : Number crunching : Reserving a core for Collatz (Message 21467)
Posted 834 days ago by Profile ritterm
...reserve a core for Collatz GPU work...

Generally speaking, is it necessary to this? I've monitored WUs several times throughout their runs, particularly at the beginning and end, but I don't see a CPU being used very much (certainly not like GPUGrid). Am I just not looking at the right time(s)?
5) Message boards : Windows : Getting imediate error as soon as the task start (Message 21452)
Posted 836 days ago by Profile ritterm
Since we start getting sieve the cpu task work but the Gpu tasks give a error...

Hi Johan,

Have you seen this post where it says:

If the sieve application fails immediately after it starts, you probably need to install the latest Microsoft Visual C++ runtime and, since BOINC will send your computer both 32 and 64 bit applications, you will need to install both the 32 and 64 but C++ runtime (see main page for a link or just Google it).

Good luck.

MarkR
6) Message boards : News : CUDA vs nVidia OpenCL User Poll (Message 18567)
Posted 1435 days ago by Profile ritterm
Manual select in preferences, with CUDA being default, or a) with a manual override...

Me, too.
7) Message boards : Number crunching : Solo validator has been down 22 hrs .Is anybody at home? (Message 17964)
Posted 1510 days ago by Profile ritterm
...he is probably out thinning out the local deer population.

Perhaps not anymore... :D Solo validator is back up and running. One of my recent WU's has validated.
8) Message boards : News : Solo Validator Back Online (11/28/2013) (Message 17963)
Posted 1510 days ago by Profile ritterm
Looks like the solo validator is offline again ...

And it's back up, now. Quite a backlog, but one of my recent solo's has validated. :-)
9) Message boards : Number crunching : Still have Computation Errors on ATI 7xxx series (Message 16366)
Posted 1731 days ago by Profile ritterm
I've switched to the solo app and will see how it goes...

Hmmm... Solos not working so well for me. 4 of 5 so far have ended in compute error "GPU steps do not match CPU steps".

[There was discussion of a similar error a couple of weeks ago here. Sorry...I don't mean to hijack the thread.]
10) Message boards : Number crunching : Still have Computation Errors on ATI 7xxx series (Message 16358)
Posted 1732 days ago by Profile ritterm
It is sending the 2.09 app to you when it should not be. The HD 7xxx GPUs use different instructions than the HD 6xxx and earlier GPUs which causes them to only complete half the workunit but still report as if it had done 100%. So, the scheduler is supposed to only send OpenCL apps to HD 7xxx and later AMD GPUs. Obviously, that isn't working with the new scheduler.

Until I get it fixed, the work around is to unselect the mini_collatz app. Select solo_collatz AND set "If no work for selected applications is available, accept work from other applications?" to Yes. That way, you will get mini_collatz apps for your CPU and solo collatz for your GPU.

Thanks for the feedback, Slicker. I had a feeling I wasn't getting the right app, but wasn't sure how all that worked. I've switched to the solo app and will see how it goes.

Laptops can get pretty warm when doing GPU crunching since they often don't have a dedicated GPU fan, so I would suggest monitoring the temperature...

Right and I've been doing that. The GPU approaches 70-deg C and CPU is in the upper-70s, which is probably pushing it. I'd like to use a laptop cooler, but, unfortunately, I need to keep it in a docking station and there's no good way to make all that work. I may not be able to use this host as a long-term GPU cruncher.
11) Message boards : Number crunching : Still have Computation Errors on ATI 7xxx series (Message 16354)
Posted 1732 days ago by Profile ritterm
I'm running a laptop (Computer 125973) that has a 7570M GPU. I've been running mini-collatz WU's for several days and have an invalid rate of about 33%.

Is it possible/likely that I'm having the same issue as is being discussed here? Is the GPU not robust enough to be reliable? Am I using the wrong driver? I'm not sure how to decipher AMD's drivers, but I doubt that what's installed is the latest and greatest.

Thanks for your thoughts and advice.

MarkR
12) Message boards : News : Server Upgrade on Thursday, April 18th (Message 16291)
Posted 1737 days ago by Profile ritterm
You need an Ivy Bridge CPU for the GPU to be used a an OpenCL device, the Sandy Bridge CPU's GPU is not supported...

Thanks for the info, Claggy...<sigh> :-(
13) Message boards : News : Server Upgrade on Thursday, April 18th (Message 16287)
Posted 1737 days ago by Profile ritterm
This upgrade should enable Collatz to support OpenCL applications on Intel embedded GPUs.

Forgive my ignorance, but how do I know or check if my i5 lappie is a candidate for this. At start-up, the BOINC manager log says "No usable GPU's found". Is it a matter of simply updating the driver? The Windows Device Mangler indicates it's using version 8.15.10.2253 from 11/28/10. And, off course, after running the driver update search, it tells me that the "driver is up to date".

Cheers!

MarkR
14) Message boards : Number crunching : Lot's of GPU, not many points? (Message 10974)
Posted 2564 days ago by Profile ritterm
Yes the WU are four times longer but the credits just dont make up for the longer processing time. I ran the new work units for a day or two on my ATI 5970 GPU and my average credit droped like a stone.

I think the new longer WUs pay 4 times as much, so, on a cr/time basis, they pay the same. I have noticed, though, that my pendings are building up a lot faster. Could it be that your RAC is dropping because it's taking a while for your pendings to catch up?
15) Message boards : Number crunching : Pending Credit Mismatch? (Message 10961)
Posted 2565 days ago by Profile ritterm
I made a change to the "Pending credit" page accessed from the Account page so it now shows a better estimate of the claimed credit. It uses an average of 750 credits for the smaller WUs and 3000 credits for the new larger WUs. That way, the total is at least close to the actual pending credits.

Well done, Slicker...Thanks. <insert thumbs-up smiley here>
16) Message boards : Number crunching : Pending Credit Mismatch? (Message 10919)
Posted 2566 days ago by Profile ritterm
There have been oodles (a.k.a. a very large number) of permission issues and rebuilding and restoring everything. I'm guessing this is another one of them.

Thanks for the feedback, Slicker. I wasn't too worried about that particular table as the "Show:Pending" view looked to be accurate, but wanted to point it out in case it needed some attention.
17) Message boards : Number crunching : Pending Credit Mismatch? (Message 10908)
Posted 2566 days ago by Profile ritterm
This is usually explained as a database issue, not a problem just different parts of the database updating at different times. So the 'Tasks" -> "Show:Pending' page is updated once every 8 hours while the 'Pending Credit' page is updated every 10 minutes, as an example. The numbers above are mine and TOTALLY made up and are used as an example only!

Thanks, mikey...Yeah, I figured there's a d/b problem somewhere and was just curious if anybody else was having the same issue. In my case, the pending credit list hasn't changed for a couple of days.
18) Message boards : Number crunching : Pending Credit Mismatch? (Message 10886)
Posted 2567 days ago by Profile ritterm
Anybody else have a mismatch in their pending credit stats? If I follow the "Pending Credit" link from my "Your account" page, only four pending WUs are listed. However, if I follow "Tasks" -> "Show:Pending", there are a total of 23 (at the time of this post), which include the aforementioned WUs.
19) Message boards : Number crunching : Is this long running WU bad? (Message 9861)
Posted 2668 days ago by Profile ritterm
If you exit and then restart Boinc it should get back to normal, some people have found that suspending it and the restarting the unit also works.

Thanks, mikey. As you suggested, a simple suspend/resume did the trick. As soon as it resumed, the elapsed time changed to 30 min and it was done in another minute. I reported the task and got the usual credit.

I'd like to think I've been doing this BOINC thing long enough to have tried that first, but I don't think I'd ever seen a CC WU lock up like that on either of my GPU-crunching hosts. Sometimes it takes someone else to remind you to try the simple solution! :D
20) Message boards : Number crunching : Is this long running WU bad? (Message 9853)
Posted 2669 days ago by Profile ritterm
I have a host (Q6600, WinVista, GTX260) that typically finishes a WU in a little over 30 minutes. Today I found it still crunching this WU after more than 9 hours. It seems to be "stuck" at 97.66% completion with a time to completion estimate of ~15 minutes. 90 minutes later it had made no progress other than elapsed time.

Should I just abort it or is there a chance it will finish and I'll get obscene amounts of credit for it? :D I've since suspended this WU and the next one seems to be running "normally".

MarkR


Next 20

Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.