Welcome to MilkyWay@home

Posts by Brickhead

1) Message boards : News : Award Badges Going Live Soon! (Message 66602)
Posted 11 Sep 2017 by Brickhead
Post:
A question more relevent for 2016 but will you be adding badge pairs for extened years of service after platinum?

(bronze),(silver),(gold),(platinum),(rarer metal),(unobtainium)

or adjusting the service ranges for (bronze),(silver),(gold),(platinum) in the future?


We'll add more as the need arises - we'll get clever. :)

The need is already here. Since the number of years are displayed in writing, I don't quite see the need for different background colours. Fitting the additional years (in Roman numbers) within the available space ought to be easy up to XVII or so. (Beyond that, just un-tilt the square.)
2) Message boards : News : Bad Batch of Runs (Message 66423)
Posted 9 May 2017 by Brickhead
Post:
Thanks! (Those were becoming a bit of a nuisance.)
3) Message boards : News : Scheduled Server Maintenance Concluded (Message 66412)
Posted 8 May 2017 by Brickhead
Post:
Jake, what are these being generated from scratch (not quota-fill) today?

de_modfit_19_3s_146_bundle5_ModfitConstraintsWithDisk_1...

They all fail, and the names don't fit neither your list nor the old run.
4) Message boards : Number crunching : tasks de_modfit_fast_19_3s_140 all failing (Message 66340)
Posted 3 May 2017 by Brickhead
Post:
I think you'll find the answer in the news: https://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=4127

Background (also in the news): https://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=4125
5) Message boards : Number crunching : M@H seems to be freezing the progress of other tasks (Message 66295)
Posted 12 Apr 2017 by Brickhead
Post:

Limit yourself to 1 or 2 tasks running on the gpu or else all the instances will be spinning on multiple cpu's waiting for the gpu to finish a call.

So I should cancel my Citizen Science Grid project and run only Milkyway@Home, is that what you mean?

He didn't say that. The way I understand Tom's advice, you should NOT run more than 1 or 2 GPU tasks at once, then all but 1 or 2 CPU cores will be free for other projects to use.

Looking at computer ID 706124, that one has 8 CPU cores (4 physical + 4 more logical, like you said) and 1 GPU. If you're not using an app_config.xml file to tell MW otherwise, MW will run GPU task one after the other, which should keep one CPU core occupied - in that case you're following Tom's advice already. Your other project should be able to use the other 7 cores, unless it expects access to all 8 and refuses to run when one of them is unavailable.

If all of the CPU cores are busy, can you see which processes are using all that CPU time? (Task Manager, Processes, sort by CPU.)
6) Message boards : Number crunching : (RX 480) GPU tasks looping a couple of times before completing (Message 66282)
Posted 8 Apr 2017 by Brickhead
Post:
That's normal behaviour. There are several computing tasks bundled into each WU nowadays, but their apparent progress isn't one linear progression. Instead, each step starts at 0% and finishes at 20% more than the previous step, so you'll se 0..20, then 0..40, 0..60, 0..80, and finally 0..100%.

This is actually a small ovesight partially rectified by the sequence of finish values, but we're still waiting for a similar sequence of start values (albeit not too impatiently, it's just a cosmetic issue).
7) Message boards : Number crunching : Nvidia GPU compute client only doing 1/5 of the required work (Message 66270)
Posted 3 Apr 2017 by Brickhead
Post:
However it got there, you are using a brand new nvidia driver that is known to have caused failures in MW.

http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=4101&postid=66257
8) Message boards : Number crunching : Nvidia GPU compute client only doing 1/5 of the required work (Message 66263)
Posted 2 Apr 2017 by Brickhead
Post:
I'd venture a guess at Windows 10 auto-updating drivers. Have a look at this thread:

https://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=4101
9) Message boards : Number crunching : Thousands of validation errors, no good work? (Message 66260)
Posted 1 Apr 2017 by Brickhead
Post:
The last (I think) task marked "MilkyWay@Home Anonymous platform (CPU)" was returned 1 Apr 2017, 18:51:57 UTC. Status "Abandoned" indicates that this was when you reattached to MW on this computer.

After that, all non-nbody tasks sent to that computer are marked "MilkyWay@Home v1.40", and the one returned 1 Apr 2017, 21:31:21 UTC does indeed say "Completed and validated".

Looks to me like you've sorted the problem :)
10) Message boards : Number crunching : Thousands of validation errors, no good work? (Message 66243)
Posted 24 Mar 2017 by Brickhead
Post:
Clients using Anonymous platform, where the application and settings are specified locally, are not subject to automatic updates from the MW servers.

Have your checked whether the MW application on this host is the current latest release? My guess is that it might be obsolete.

There are a few other hosts with this problem (all of them use Anonymous platform AFAICT), but the vast majority are OK.
11) Message boards : Number crunching : No new WUs at all since Jan 27 2017 ??? (Message 66217)
Posted 26 Feb 2017 by Brickhead
Post:
What does the BOINC event log say? (BOINC Manager, Tools, Event Log)
12) Message boards : Number crunching : Hosts with only invalid results (Message 66210)
Posted 20 Feb 2017 by Brickhead
Post:
CPDN has (for more than a decade) a dynamic value per computer - "Maximum daily WU quota per CPU". To me, this indicates that the throttling mechanism might already be present in the BOINC server software.
13) Message boards : Number crunching : Hosts with only invalid results (Message 66204)
Posted 18 Feb 2017 by Brickhead
Post:
Do the BOINC servers throttle the flow of new work to hosts that produce nothing but validate errors? If not, should they?

Not a big deal, but it's a bit annoying to see good results invalidated by a few wingmen that seem to do nothing else.

Example:
https://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=606779
14) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61614)
Posted 24 Apr 2014 by Brickhead
Post:
Btw you could underclock your 6970s grx RAM & it would hardly affect WU times if at all.

I know, and I already have - that's why I need to manually apply a preset after each reboot. Stock clocks for those cards are 880/1375 (as opposed to my 950/900).
15) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61570)
Posted 21 Apr 2014 by Brickhead
Post:
Hi Brickhead, good to see you here :) & thanks for the times but I take it you didn't read my last post here? or the one at Picard's forum?

Actually I did, but seing as I had 213.76 cs WUs aplenty, my brain refused to cooperate.

Interesting to see the different output from the different credit WUs, unfortunately as I mentioned in my last post here the 'long' 213.76 credit WUs that I was using for the benchmark are no longer being released*. And I'm pretty sure the 213.76 times your showing are for the short versions, so sadly your too late & I can't use your data for the table :(.

If the run time comparison needs WUs no longer available, isn't that a bit like judging which dinosaur meat tastes the best? (Not that I wouldn't like to know;)

I thought I'd found a bunch for your 6970 but their seem to be some larger fluctuations in times their, were you altering clock speeds or playing a game along side those longer time WUs? Or watching videos? Anyway the WUs I was looking at run from 133-148s (mostly 140-145s). Visually I'd say the average was ~143s but I'd like to hear your answer before I commit to that time ;).

That particular computer does nothing else while crunching, except the odd Windows update, and it has been running at 950/900 for years now. But the first 3 WUs after a logout or reboot may be interrupted and restarted once or twice, not only due to the logout, but sometimes again because the memory clock is too low for the driver to remember and must be manually retrieved from a preset.
16) Message boards : Number crunching : Benchmark results - times wanted for any hardware, CPU or GPU, old or new! (Message 61560)
Posted 20 Apr 2014 by Brickhead
Post:
AMD Radeon HD 5870
GPU 868 MHz, memory 900 MHz, 2 cards, no CPU tasks

MilkyWay@Home v1.02, credit = 79.93 cs
34.09 s run time (single sample), 2.3447 cs/s

MilkyWay@Home v1.02, credit = 106.88 cs
45.17 s run time (50 sample avg), 2.3661 cs/s

Milkyway@Home Separation (Modified Fit) v1.30, credit = 213.76 cs
126.67 s run time (50 sample avg), 1.6876 cs/s

AMD Radeon HD 6970
GPU 950 MHz, memory 900 MHz, 3 cards, no CPU tasks

MilkyWay@Home v1.02, credit = 79.93 cs
28.23 s run time (4 sample avg), 2.8319 cs/s

MilkyWay@Home v1.02, credit = 106.88 cs
35.48 s run time (50 sample avg), 3.0126 cs/s

Milkyway@Home Separation (Modified Fit) v1.30, credit = 213.76 cs
101.67 s run time (50 sample avg), 2.1025 cs/s
17) Message boards : Number crunching : Happy CPU (Message 56110)
Posted 5 Nov 2012 by Brickhead
Post:
On multiprocessors, use at most
Enforced by version 6.1+ 80% of the processors


This is the setting I'm using that seems to be working. Checking the CPU threads, they seem to fluctuate between 50% to 100%, but sometimes lingering lower or at 100%.

That setting does exactly what others have suggested; limit the number of cores used. Windows' SMP scheduler will to some extent switch the processes back and forth if it can, making the overall load appear more evenly spread than it actually is at any given instant.

If you're not happy with the responsiveness of the computer without that 80% limit, what you've done is the proper way to reserve some of the CPU.
18) Message boards : Number crunching : Catalyst issues (Message 55813)
Posted 15 Oct 2012 by Brickhead
Post:
Just apply the stored profile, and the slider scale should expand to encompass the mem clock stored in it. (You might have to apply it twice to actually set the mem clock too.)
19) Message boards : Number crunching : Hi I'm back, but n-body tasks run for 6 seconds ? (Message 52499)
Posted 14 Jan 2012 by Brickhead
Post:
Is there a limit as to how many simultaneous n-body tasks one host can run or queue? I see only one task downloaded at a time, run for a few seconds, and then all CPU cores are idle for the rest of that minute, repeat.

I can of course use the CPU for something else, but then the resource share - and ultimately running n-body at all - is a moot point.
20) Message boards : Number crunching : Host with persistent split personalities (Message 50871)
Posted 29 Aug 2011 by Brickhead
Post:
Some time around the big bang a few months ago, one of my computers was assigned a new host ID, which normally would be easy to sort. In this case however, for some reason the merge function complains about overlapping lifespans. Does anyone know if it's at all possible to circumvent this nuisance and get the IDs merged?

This is the (one) host I'm referring to:
http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=280260
http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=280532

Cheers


Next 20

©2024 Astroinformatics Group