Posts by DanNeely

1) Message boards : News : increased WU limits (Message 47926) Posted 16 Apr 2011 by DanNeely Post: [Here's the link. I'm using 6.12.22 on everything. The preferences have changed. set "While processor usage is less then" to 0 for behavior similar to earlier BOINC versions: http://boinc.berkeley.edu/download_all.php Thanks. I've got it installed on my main box and it seems to be behaving fairly well, although until MW glitches again and I run out of work I can't test the new features I care about. I'm going to hold off a bit on installing it elsewhere since my CPU only boxes don't have any trouble getting several days of work in normal circumstances. I noticed some UI changes, is there anywhere I can find a consolidated list of any other major changes since 6.10.x; or do I need to wade through the entire 6.11/12 thread in the boinc dev forum?
2) Message boards : News : increased WU limits (Message 47886) Posted 15 Apr 2011 by DanNeely Post: How's the server doing under load so far? I'd really like to be able to get at least an hours work for my 5870 so it stays on MW after a maintenance window triggers a 1 hour backoff. When that happens and my MW queue runs dry boinc turns to collatz (backup project) and gets the better part of a day worth of work and decides to crunch all of it before returning to MW. It's possible the latter was due to my debt levels getting messed up. They looked off when I checked earlier today, so I zeroed them out in client_state.xml and am waiting to see what happens after the next outage. It'll do the same thing because backup projects aren't supported in 6.10.xx. Hate to sound like a broken record... :) They aren't? The DL only when other projects are out of work portion is working. What version do I need to get full support?
3) Message boards : News : increased WU limits (Message 47871) Posted 15 Apr 2011 by DanNeely Post: How's the server doing under load so far? I'd really like to be able to get at least an hours work for my 5870 so it stays on MW after a maintenance window triggers a 1 hour backoff. When that happens and my MW queue runs dry boinc turns to collatz (backup project) and gets the better part of a day worth of work and decides to crunch all of it before returning to MW. It's possible the latter was due to my debt levels getting messed up. They looked off when I checked earlier today, so I zeroed them out in client_state.xml and am waiting to see what happens after the next outage.
4) Message boards : Number crunching : Change in behavior with mismatched GPUs (Message 47672) Posted 12 Apr 2011 by DanNeely Post: Thanks, I figured it'd probably be a pita; but was hoping to be proven wrong.
5) Message boards : Number crunching : Change in behavior with mismatched GPUs (Message 47654) Posted 11 Apr 2011 by DanNeely Post: Thanks. That has me running MW successfully again. Not sure if it'd be worthwhile on a watts/credit basis, but is there any way I could do similar at the app level so I could keep MW on the 5870 as my primary project, while letting collatz serve as both the 5870's backup and as the 5450's primary?
6) Message boards : Number crunching : Change in behavior with mismatched GPUs (Message 47630) Posted 11 Apr 2011 by DanNeely Post: ok, dnetc's broken in general and is apparently not related to whatever is causing my problems with MW now. I assume it's related to the new apps based on the timing; but since I was away during the rollout period I can't confirm that is when it stopped working.
7) Message boards : Number crunching : Change in behavior with mismatched GPUs (Message 47591) Posted 11 Apr 2011 by DanNeely Post: PS I noticed dnetc not working a week or so back, but don't recall if it was before or after my driver update. It was only a backup project though, so I didn't try and do any troubleshooting at the time.
8) Message boards : Number crunching : Change in behavior with mismatched GPUs (Message 47590) Posted 11 Apr 2011 by DanNeely Post: I have a 5870 and a 5450 in my system. Owing to a boinc bug my client thought I had two 5870's and would run two WUs at the same time. At the time my runtime results corresponded to running 2 WUs on my 5870 concurrently (and 0 on the 5450), but without the need to fiddle with app_info files with each new app. It also worked for all 3 GPU project I ran: milkyway, collatz, and dnetc. A fringe benefit was that it resulted in slightly lower GPU loading and kept my desktop from lagging. Recently something changed though. Boinc is now apparently trying to run WUs on the 5450. The strongest indication I have of this is with collatz where I see most WUs completing in about 25 minutes, with a few taking ~8.5 hours. The 20x speed factor matches well with the 20x difference in the number of cores the cards have. MW and dnetC both have one task run properly and a whole bunch more crash within a second or two of starting. I know MW needs DP and won't run on the 5450, and I believe the same is true of dnetc, but since their page is down I can't confirm it. In a thread from 6 months ago, it was suggested I use a cc_config.xml file to control the behavior. At the time I didn't bother since nothing was causing a problem. I've tried doing it today, but can't get it to work. I fixed the typo in the example file and put in in C:\programdata\boinc\ with both 0 and 1 set, but I continue to see both cards being used in MW/collatz. I'm currently running the 11.3 ati drivers. This is a recent update, but IIRC MW was working with the 11.3 driver prior to the recent upgrades here. <cc_config> <options> <use_all_gpus>0</use_all_gpus> </options> </cc_config> old thread on the issue: http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2016&nowrap=true#43312
9) Message boards : Number crunching : Aaargh! Servers are out of new work!(2)" (Message 46409) Posted 2 Mar 2011 by DanNeely Post: I noticed something: every time a backlog forms with the queue for results waiting for validation, the server runs out of work. Is there something wrong with the validator? Results from the validator are used to determine what to set the new WUs up as; I assume that means if validation stops so does WU creation.
10) Message boards : News : Nvidia OpenCL updated (Message 46227) Posted 13 Feb 2011 by DanNeely Post: I'm seeing 100% failure with win7-64 and GTX260s/
11) Message boards : Number crunching : super charged? (Message 43863) Posted 16 Nov 2010 by DanNeely Post: 14.11.2010 10:00:19 Milkyway@home Message from server: (won't finish in time) BOINC runs 88.1% of time, computation enabled 100.0% of that According to this, boinc thinks you have more WUs than you can complete before they expire. When this happens it stops DLing any new work until it clears the backlog. IN your case the backlog is almost certainly Einstein WUs. IF this is actually the case (Are your Einstein WUs finishing in about the same amount of, or more, time as the estimates?), you should abort some of the Einstein task. IF it turns out boinc is just confused about how long they'll take it will self correct after a few days and then run MW primarily or exclusively for a while to make up the deficit it's currently undergoing. Due to vargarities of the scheduler even after you're caught up boinc is unlikely to consistently have exactly half of your cores running each application at any given time.
12) Message boards : News : updated the server side daemons to deal with the credit issue (Message 43754) Posted 11 Nov 2010 by DanNeely Post: The other thing I've discovered that will kick your card into 2d mode is most web based video (eg youtube); fortunately this recovers as soon as the offending tab is closed.
13) Message boards : Number crunching : Does AMD cripple DP performance? (Message 43648) Posted 9 Nov 2010 by DanNeely Post: I'm looking at the DP performance between the FireStream and Radeon line and they both have the same 1/5 SP performance. Why would the FireStream GPU have the same "limitation"? Because unlike nVidia ATI doesn't cripple DP performance on consumer cards to prop up the premiums on their workstation cards (although if they decide to compete against Tesla in the super computer market I would be surprised if they don't change their tune). What they don't do, and what nVidia hasn't done outside of the GF100 chip (and presumably the GF110), is to support DP in hardware to the maximum amount possible (SP/2). The reason ATI doesn't do it, and nVidia hasn't done it on their other chips is that the 99% of their customers who don't take advantage of GPGPU never use it which means that the die area that is being taken up to support FP64 (and some is needed for setup/teardown activities and linking 2 32 bit pipes into one 64bit pipe even though most of the FPU is shared) instead of being used to place more shaders/texture units/rops/cache/etc that the 99% of people who buy them for gaming performance only care about. Alternatively they could make the dies slightly smaller and cheaper for a higher profit margin, or sell the cards for less to gain market share. That said however, nVidia is nerfing FP64 more than is strictly needed to protect their Tesla cards. The C2050 (Tesla equivalent of a GTX480 with 3GB ram) is priced at $2150 by the only company I could find selling it. With 480's running $450-500, FP32/6 wouldn't seriously threaten the Tesla market. After factoring in the power consumption, cooling costs, and rack space even FP32/4 would probably be safe. However until a mass marketish consumer need for GPGPU-FP64 support emerges I wouldn't hold my breath on them loosening up much.
14) Message boards : Number crunching : Software confusion with multiple GPUs (Message 43319) Posted 30 Oct 2010 by DanNeely Post: I should add that I'm assuming that resource conflicts will result in 2 WU's running concurrently taking slightly longer to complete than the same 2 WU's running sequentially. IF this isn't the case other than looking stupid I suppose it's not a major problem,
15) Message boards : Number crunching : Software confusion with multiple GPUs (Message 43317) Posted 30 Oct 2010 by DanNeely Post: I know the 5450 won't run MW, so 1 at a time would be my default objective so they're not competing with each other for hardware resources. I'd like to know where the bug is though so I can report it to the correct people. My account lists my computer as having 2 5800 series cards not one 58xx and one 54xx; is this an MW problem or a problem with the boinc client. http://milkyway.cs.rpi.edu/milkyway/hosts_user.php
16) Message boards : Number crunching : Software confusion with multiple GPUs (Message 43312) Posted 30 Oct 2010 by DanNeely Post: I have both a 5870 and a 5450 card in my PC. Since I added the 2nd card boinc has been running 2 MW WU's at a time, thinking both are getting 100% of a GPU; the fact that their runtimes move in lockstep with each other and are 2x as high makes it clear that they're both running on my 5870. Is there any way I can apply a cluebat to the system to straiten it out?
17) Message boards : Number crunching : How do I reset ATI GPU speed after gaming? (Message 43216) Posted 28 Oct 2010 by DanNeely Post: That was my understanding of why it was dropping from 890 to 400. The problem is that once it drops to 400, I can fiddle with settings and click the apply button as often as I want in CCC/overdrive but the GPU will not leave 400mhz until I reboot.
18) Message boards : Number crunching : How do I reset ATI GPU speed after gaming? (Message 43212) Posted 28 Oct 2010 by DanNeely Post: That works as an explanation, but not being able to recover from an overheat situation without a reboot is fairly horrid design.
19) Message boards : Number crunching : How do I reset ATI GPU speed after gaming? (Message 43191) Posted 26 Oct 2010 by DanNeely Post: I use CCC/MSI Overdrive to set my 5870 to run at a strait 890mhz at startup. It stays there and cranks out WU's in about 90s until I do a gaming session; after I'm done gaming the 5870 remains stuck at 400mhz and I can't find a way to kick its operating speed back up to 890 without rebooting. There has to be an easier way...
20) Message boards : Number crunching : Aaargh! Server out of new work! (Message 42598) Posted 5 Oct 2010 by DanNeely Post: Would it be possible to split the server-status that we can see which app has work and which does not? Einstein at home does a full breakout on their server status page, so it's definately doable. Not sure how much of their stuff is custom though. http://einstein.phys.uwm.edu/server_status.html[/url]

Next 20