Message boards :
Number crunching :
AP units initial time over 400 hours- How to get additional SETI work?
Message board moderation
Author | Message |
---|---|
Richard Walton Send message Joined: 21 May 99 Posts: 87 Credit: 192,735 RAC: 0 |
About the AP units- They come in to my computer reading at over 400 hours to complete initially. Generally, the real time is just over 100 hours. The problem is that if I want to run just SETI at this point, when I get one of those AP units in, I do not get any further work until I am well into that AP unit, usually meaning that one processor is idle while waiting. At times, I go off to other things for a while, but there are times I would rather just run Seti. I have tried to make BOINC get more work by increasing the cache; but that 400 hours of AP unit is over 14 days work as far as BOINC knows and I can not get any more. Why is the AP unit reading so huge? I have done a number of the larger ones now. All have been just over 100 hours real time, and they all say higher number of hours initially. |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
The time estimate for how long the WU will take is based on the Task Duration Correction Factor (TDCF, or DCF). On systems that only crunch APs, the time estimate is pretty close to accurate. Mixing APs and MBs on the same system, especially if there is CUDA running, will throw the DCF way off, and the estimates will be all wrong. The simplest solution is that if you don't want the AP units, go into your SETI@Home Preferences and change the allowed work for a venue to SETI@Home Enhanced (MB), and uncheck AP, AP_v5, and "allow work for other apps". That will ensure that you only get MB tasks from now on. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
The problem is that if I want to run just SETI at this point, when I get one of those AP units in, I do not get any further work until I am well into that AP unit, usually meaning that one processor is idle while waiting. Astropulse is SETI, so when you have Astropulse, you are doing SETI. Multibeam looks for narrowband evidence of ET, Astropulse looks for wideband evidence of ET. I think your actual question is: why do I have an idle core on my multi core system? To talk about that, we'd have to know what version of BOINC you are using. The times are off because the estimates are estimates. Over time, your BOINC client will get better at correcting the estimates, but they'll likely always be a little off. |
Richard Walton Send message Joined: 21 May 99 Posts: 87 Credit: 192,735 RAC: 0 |
I am on BOINC 6.4.6 I do not mind AP. It would just be nice if I could get other SETI work on the idle processor when I have a new AP unit come in at over 400 hours that continues to take only a bit over 100 hours to do when it gets down to it. Why not even two AP units? I can do two in the same time it takes for one. It just seems that since the new improved AP, once I get one, that is it until I get a few days into the AP unit. |
Andy Williams Send message Joined: 11 May 01 Posts: 187 Credit: 112,464,820 RAC: 0 |
Mixing APs and MBs on the same system, especially if there is CUDA running, will throw the DCF way off, and the estimates will be all wrong. Now that you mention it, I don't understand why the problem isn't worse. I just checked the DCF on my CUDA machines. It is as low as 0.02. The same machines run AP on the CPUs, but the cache doesn't get out of control. Is BM (6.6.20 in my case) smart enough to remember the time CPU tasks actually take and schedule downloads for the cache appropriately? It would seem the long term solution would be to have two DCFs, one for CUDA and one for the CPU(s). -- Classic 82353 WU / 400979 h |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
Mixing APs and MBs on the same system, especially if there is CUDA running, will throw the DCF way off, and the estimates will be all wrong. I'm not involved in any of the CUDA stuff, or any of the 6.6.x testing, but I believe I remember hearing that starting with 6.6.17 (?), the work fetching is split into three categories..or at least two.. CUDA, CPU (MB), and AP (CPU). Three separate "caches" to fetch work for. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
Richard Walton Send message Joined: 21 May 99 Posts: 87 Credit: 192,735 RAC: 0 |
Mixing APs and MBs on the same system, especially if there is CUDA running, will throw the DCF way off, and the estimates will be all wrong. No CUDA here. I would like to, but just have a lowly CUDA free laptop- Maybe I should go back to pre-CUDA Boinc version? But, I thought the "works" were the same except for the addition of CUDA choice. |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
I am on BOINC 6.4.6 I don't know if you're adverse to beta testing. I understand that a big change in the current development versions is to do scheduling on a per-core basis instead of a per-system basis. In other words, BOINC would recognize your overworked core, and your idle core, and get more work. This has nothing to do with the science application, by the way, one of the major BOINC features is that the science application may completely ignore scheduling. |
Richard Walton Send message Joined: 21 May 99 Posts: 87 Credit: 192,735 RAC: 0 |
I am on BOINC 6.4.6 How do I sign up? I mean, sure- I could do this, but are there requirements other than downloading a test version of BOINC and telling someone if I have a problem? |
Aurora Borealis Send message Joined: 14 Jan 01 Posts: 3075 Credit: 5,631,463 RAC: 0 |
You don't necessarily have to join Boinc Alpha testing to use the alpha versions. You could just join the mailing list to stay current (I don't have the URL handy), but be ready for a lot of mail especially when a new version is put out for testing. If you would prefer you can just report problems to the Boinc Message Boards. You will need to register to be allowed to post. The latest Boinc version can found on the download page from All versions or from Download Index page which also list past versions. Boinc V7.2.42 Win7 i5 3.33G 4GB, GTX470 |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
You can get the current pre-release client here. You should sign up for the BOINC Alpha project as well -- crunching for that project is very optional. |
Richard Walton Send message Joined: 21 May 99 Posts: 87 Credit: 192,735 RAC: 0 |
You don't necessarily have to join Boinc Alpha testing to use the alpha versions. You could just join the mailing list to stay current (I don't have the URL handy), but be ready for a lot of mail especially when a new version is put out for testing. If you would prefer you can just report problems to the Boinc Message Boards. You will need to register to be allowed to post. Thanks. I just downloaded version 6.6.20, and within seconds of restarting computation I got an additional 15 tasks. So far, so good. . . |
daysteppr Send message Joined: 22 Mar 05 Posts: 80 Credit: 19,575,419 RAC: 53 |
You may also wish to use the optimized apps if you are not already using them. They can cut down the processing times dramatically. Unknown if they work in Alpha/Beta though. Sincerely, Daysteppr |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
You may also wish to use the optimized apps if you are not already using them. They can cut down the processing times dramatically. They're fine if you're testing BOINC. If you are part of the SETI Beta effort, that tests new science applications, and if you run "optimized" on SETI Beta you can give false results -- best to ask before you do so. |
©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.