Posts by miw
|
1)
Questions and Answers :
GPU applications :
CUDA With 9400GT: BOINC only sees 128MB Memory
(Message 886248)
Posted 18 Apr 2009 by  miw
Post: The difference between the systems was probably that the GeForce 9400 system hadn't reported a SETI unit since it was on driver version 168.49. It shows 182.50 now, since I just forced a report. Strangely enough, that card did CUDA just fine on version 168.49 back on an older version of BOINC that did not check the amount of memory on the card. (Even though it ostensibly was unsupported.) I'd still be interested to hear if anyone has had a problem with BOINC not reporting the right amount of memory for a CUDA card. Edit: The old driver was the one that came with the card most likely. the latest comes from nvidia. Not that it matters. The card is obviously a reference card, to the point that it does not have any card vendor branding on it. --Mark |
|
2)
Questions and Answers :
GPU applications :
CUDA With 9400GT: BOINC only sees 128MB Memory
(Message 886240)
Posted 18 Apr 2009 by miw
Post: Yes. You may well be right. However, I've tried many of the driver versions starting right back at about 181.20 and get the same result. There was an earlier version of BOINC that did not check video card memory and CUDA ran fine with that. I'd still be keen to know if anyone else has had issue with BOINC getting an incorrect reading of the RAM from the video card, regardless of where the source of the problem is. If nobody else has seen it, then I would actually suspect some issue with the particular brand of 9400GT that I have.... (I thnk it is Gainward, but am not sure. Looking at the card gives no clue and I threw the box out ages ago.) Note that the driver correctly (or otherwise?) reports 512MB to windows. --Mark |
|
3)
Questions and Answers :
GPU applications :
CUDA With 9400GT: BOINC only sees 128MB Memory
(Message 886237)
Posted 18 Apr 2009 by miw
Post: Folks, I'm running: Windows XP SP3 Nvidia Drivers 182.50 BOINC 6.6.23 In the startup, It says: Cuda Device: GEForce 9400 GT (driver version 18250, CUDA version 1.1, 128MB, est. 3GFLOPS) But this 9400GT has 512MB RAM and windows confirms that it has 512MB in the control panel. Of course, since 128 < 256, CUDA will not run because of this. Has anyone else seen this? --Mark |
|
4)
Message boards :
Team Recruitment Center :
The Great Aussie Assault Number four
(Message 473095)
Posted 5 Dec 2006 by miw
Post: Anyone else noticed a change in the work unit deadlines? I have some now that are due 20 days before ones I already had. Very strange. Yes. The time to deadline varies according to the AR (which affects the expected crunching time.) Go Aussie! (now, If I could just unwedge this Linux box from crunching other projects..... :-) ) |
|
5)
Message boards :
Number crunching :
BOINC/SETI causes trouble on Core Duo Laptop
(Message 457477)
Posted 14 Nov 2006 by miw
Post: Found the problem. It's called "Thinkvantage Away Manager". It's a piece of IBM bloatware that is supposed to improve performance by stopping some tasks from running when the CPU is busy. Seems to me it chains itself into some software interrupts and then refuses to run when CPU is 100%. Yes. and as you can imagine, if it is a work laptop, then there is limited sympathy or help given by the corporate IT people. Solution is "stop running BOINC" because that makes the symptom go away, even if the problem is a poorly-written piece of software that comes with the PC. |
|
6)
Message boards :
Number crunching :
Run at 100% causes Windows to lock
(Message 456656)
Posted 13 Nov 2006 by miw
Post: I just installed BOINC 5.4.11 a few days ago on my WinXP Pro SP2 machine Thinkpad laptop. It's been running at 100% since installation. Your problem is called "Thinkpad Away Manager". Uninstall it and things will work find. See: http://setiathome.berkeley.edu/forum_thread.php?id=35324&nowrap=true#456654 |
|
7)
Message boards :
Number crunching :
BOINC/SETI causes trouble on Core Duo Laptop
(Message 456654)
Posted 13 Nov 2006 by miw
Post: @ miw Found the problem. It's called "Thinkvantage Away Manager". It's a piece of IBM bloatware that is supposed to improve performance by stopping some tasks from running when the CPU is busy. Seems to me it chains itself into some software interrupts and then refuses to run when CPU is 100%. Anything chained behind it in the SWIs waits forever to run. Uninstalled it and everything is fine now.... |
|
8)
Message boards :
Number crunching :
BOINC/SETI causes trouble on Core Duo Laptop
(Message 455015)
Posted 11 Nov 2006 by miw
Post: @ miw Well, I have not found a decent workaround yet, since setting the Laptop (which is also a Thinkpad - X60) to not work while user is active means it barely crunches during the day. For the time being, what I do is when I want to open an Excel or Word file or visit a link out of a mail message or hibernate or whatever, I pull the power pack out and go to battery. BOINC suspends, and everything works. Then I put the power back in. However, knowing that Eric also has the same problem with his thinkpad gives me some ideas, as well as letting me know I am not going crazy. :-) I'm gonna hunt down all the extra services that IBM install to run thinkpads and see if I can find one of them causing the problem.... |
|
9)
Message boards :
Number crunching :
BOINC/SETI causes trouble on Core Duo Laptop
(Message 454155)
Posted 9 Nov 2006 by miw
Post:
Yes. It is a partial solution. At least the processes do actually start in this case. Unfortunately, the throttling algorithm seems to be a duty cycle with a 1s granularity. That is, if you set it to use no more than 95%, then BOINC runs for 19s and then suspends for 1s. If you set it to use no more than 98%, then it runs for 49s and pauses for 1s. So if I set it to use 95%, then I still have to wait 10s on average for my process to launch. It would be more intereating to find out why BOINC is blocking DDE messages. Is it a problem with BOINC or with the particular windows setup on my PC. If anyone knows about DDE and could suggest things I could do to debug this, then I would be most grateful. |
|
10)
Message boards :
Number crunching :
BOINC/SETI causes trouble on Core Duo Laptop
(Message 454152)
Posted 9 Nov 2006 by miw
Post:
1 Processor, 2 cores, 1GB RAM. Definitely not thrashing. Except for this process launch problem which I think may be linked to DDE strangeness, everything is fine. |
|
11)
Message boards :
Number crunching :
BOINC/SETI causes trouble on Core Duo Laptop
(Message 453573)
Posted 8 Nov 2006 by miw
Post: Check to see what priority the boinc executables are running at... Running at low priority, just like on my other BOINC installations. But the effect is as if it were running at "real time" priority. :-( |
|
12)
Message boards :
Number crunching :
BOINC/SETI causes trouble on Core Duo Laptop
(Message 453560)
Posted 8 Nov 2006 by miw
Post: This is a strange issue I have not seen on any other platforms I have (including some dual-core Athlons.) Platform: Core Duo T2400 1.83GHz Laptop. OS: Windows XP SP2 BOINC 5.7.2, Boinc 5.6.5, Boinc 5.4.11 Symptoms: When BOINC/Seti (or any other project) is running, I can't hibernate, can't open word or excel files by clicking on the file. Can't open a web page by clicking on a link in Thunderbird (whether or not Firefox is running at the time.) As soon as boinc stops running either by being paused or suspended, or even when I use CPU throttling (which actually (but may be irrelevant)just pauses the app/s regularly) then word/excel document opens, the link is opened, or the laptop suspends. In other words, the action I tried to do goes into some resource wait state and does not go forward until the CPU drops below 100% - it is waiting for BOINC. This of course should not happen. Nothing should have to wait for BOINC. Happens with all recent versions of BOINC that I have tried. Edit: This behaviour does not happen for opening all file types. It happens for Word and Excel, but not PPT and PDF. One thing I can see that is consistent across the file types is that those file types that have a problem have a DDE message in the "open" section under "file types", whereas those that do not have a DDE message. (although they might use DDE) any thoughts? |
|
13)
Message boards :
Number crunching :
Seti causes Some apps not to start on Intel Dual Core Laptop?
(Message 448409)
Posted 2 Nov 2006 by miw
Post: Folks, I have a strange problem here.... I've got a laptop running an Intel Core Duo T2400 processor with Windows XP SP2. When BOINC is running, some apps take a *very* long time to start if you launch them by clicking on files. (They start just fine if you run the program from the start menu or whatever.) Powerpoint is fine. Word takes ages. Excel takes ages. Opening a web page by clicking on a link in an email in Thunderbird takes ages (so it's not just microsoft) Acrobat is fine. Now, what happens when you click on the file is the folder window freezes up for minutes. If you look in task manager the app is not running yet, so it seems to be a wait for something before the app actually get launched. If you stop Boinc it works fine. If you set boinc to throttle the CPU with 5.6.x or 5.7.x Boinc client, then it waits until boinc gets throttled and then the app starts fine. (e.g. if you set it to throttle to 95%, then it will sit waiting for 19 seconds and then go through in the 1 second that BOINC gets stopped.) It's almost as if launch is not allowed while the CPU is at 100%. When I get a bit of time I'll test it out with some other app that takes 100% of CPU (like prime). Any thoughts as to why this is happening and how I can stop it from happening? It's quite annoying to have to turn boinc off when I want to work. (Yes. I could set it to not run while I am active, but that will have a similar effect to turning BOINC off in the daytime.) |
|
14)
Message boards :
Number crunching :
New version?
(Message 441641)
Posted 22 Oct 2006 by miw
Post:
If you click on the link for "all versions" you'll find 5.6.5 there, flagged as "may be unstable". I think this version was "never to be released" because they were going straight to version 5.8(?) because there was not enough resource to properly test both and the improvements in 5.6 are included in 5.8 anyway. I also got that message, just once. Probably a hiccup? I've am trying 5.6.5 out on one of my machines and so far no problems, but YMMV. |
|
15)
Message boards :
Number crunching :
New Dual FX 62 Processors
(Message 429728)
Posted 2 Oct 2006 by miw
Post:
One thing that might be a bit of a wildcard in all of this is the AMD 4x4 programme. Last time I looked, they were talking about having Mobos out around the end of this year that would accept 2x sktAM2 CPUs. that would mean 4 cores on the board this year or January and 8 cores sometime mid-2007 when you can upgrade to 4-core CPUs. Basically it means you can have 2 CPUs on the motherboard without having to go into the expensive server chips. |
|
16)
Message boards :
Number crunching :
Athlon 64 x2's and v2.10 of crunc3rs app
(Message 277251)
Posted 7 Apr 2006 by miw
Post: Yup. That fixed it! I was having the same problem with the latest Crunch3r app and an Athlon 64 X2 4800+. If no CPU affinity set, then I was losing about 50% of the cycles. In fact this machine had a lower RAC than my A64 3000+ with the same projects. If I set cpu_affinity, then SETI stopped altogether. Now with the older version and cpu_affinity set, it is crunching twice the WUs it was before, and so far I think the other projects are doing better as well (because cpu_affinity is set.) --Mark Got it! |
|
17)
Message boards :
Number crunching :
My God.....It's full of stars!!
(Message 263896)
Posted 18 Mar 2006 by miw
Post: Hmm..... Donated $50 about a year ago, but no star. Guess I'll just have to try again. :-) aaah. that's better........ |
|
18)
Message boards :
Number crunching :
My God.....It's full of stars!!
(Message 263874)
Posted 18 Mar 2006 by miw
Post: Hmm..... Donated $50 about a year ago, but no star. Guess I'll just have to try again. :-) |
|
19)
Message boards :
Number crunching :
Optimized Boinc CC v5.3.6 + modifications
(Message 263871)
Posted 18 Mar 2006 by miw
Post: I've been running tx36 for some time now, without any significant problems. However, I notice that even after well in excess of 50 WUs going through, calibration is still getting blocked by the negative calibration limit for predictor and einstein: e.g. -------- 2006-03-18 16:18:13 [Einstein@Home] CC calibration: blocked [negative calibration limit] 33.69 >> 17.03 (time: 5452s >> 2688s / Gfpops: 2.43 >> 2.70) ------- Calibration is working nicely for SETI (which has Crunch3r optimised client). It's not causing me major heartburn, because in the end I request more credit than I get, but I was wondering if others have seen this and if it goes on forever or eventually calibrates itself away. cheers, |
|
20)
Message boards :
Number crunching :
Truxsoft/Crunch3r wierdness
(Message 263866)
Posted 18 Mar 2006 by miw
Post: Has anybody else noticed the following wierdness? Machine: AMD Athlon64 X2 4800, Windows XP Running Truxsoft tx36 calibrated client BoincManager 5.2.13 Crunch3r optimised Seti client, AMD A64, SSE2,SSE3 version When I have Truxsoft <cpu-affinity/> option set, most seti results race up to about 17% in just under a minute and then sit there grinding away without making progress forever. A few (maybe 1 in 10) run to completion. I gave one unit about 36 hours and another couple in excess of 6 hours to make progress but no dice. None of my other projects (including optimised Albert C37, standard Predictor and Climate Prediction) seem to have any problems on this machine, nor are there any problems for Crunch3r on my single-CPU machines. I just disabled CPU-affinnity and all seems to be well on the dual-core box now. cheers, |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.