Message boards :
Number crunching :
The Neverending WU?
Message board moderation
Author | Message |
---|---|
Alan Smith Send message Joined: 19 Sep 00 Posts: 106 Credit: 26,816 RAC: 0 |
When I started cunching an Astropulse v5 WU, the estimated time was 124 hrs. After crunching for about 19 hours, the time remaining is 121 hrs. 40 min. Should I be concerned? At this rate it won't finish for over 700 hours? Can this be right? |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
Try to upgrade both your hosts to opt AP version. |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
When I started cunching an Astropulse v5 WU, the estimated time was 124 hrs. BOINC Manager's estimates use the Duration Correction Factor the host has established doing previous work. As AP_v5 work progresses the estimate gets better since actual performance is blended in gradually, so although crunch time will be longer than 124 hours it won't be as much more as you've guessed. If progress is above 5 or 10 percent, simply divide the amount of time which has been used so far by the progress to get a reasonably close full run time estimate, the progress indication is pretty good for AP work. Joe |
Alan Smith Send message Joined: 19 Sep 00 Posts: 106 Credit: 26,816 RAC: 0 |
When I started cunching an Astropulse v5 WU, the estimated time was 124 hrs. OK, so after 27 hours it's at 16.1 percent, so I figure it'll take about 168 hours. At least that's better than 700! This was the first AP v5 I've crunched on this machine. I tried one during the 'drought' just to keep busy. I'm guessing by the time it finishes things will be back to normal, so I set the prefs not to DL anymore. I'll wait until I can get one of those IBM petaflop computers. ;-) As for the suggestion Raistmer made that I change to the opt apps, I would if I wasn't sure thar I'd screw something up. I'll just wait until the 'opt' app is the 'standard' app. Thanks... |
skildude Send message Joined: 4 Oct 00 Posts: 9541 Credit: 50,759,529 RAC: 60 |
That still seems very slow. I had an AMD sempron 2500+ that finished ap5.03 in about 160 hrs. I upgraded a bit to an amd xp 3000+ and it finishes them in about 100 hours. In a rich man's house there is no place to spit but his face. Diogenes Of Sinope |
Alan Smith Send message Joined: 19 Sep 00 Posts: 106 Credit: 26,816 RAC: 0 |
That still seems very slow. I had an AMD sempron 2500+ that finished ap5.03 in about 160 hrs. I upgraded a bit to an amd xp 3000+ and it finishes them in about 100 hours. Is an ap5.03 the same as an AP v5? This machine crunched an original AP WU in about 110 hours. |
W5GA, W5TAT, W8QR, K6XT Send message Joined: 25 Sep 99 Posts: 42 Credit: 23,144,377 RAC: 6 |
That still seems very slow. I had an AMD sempron 2500+ that finished ap5.03 in about 160 hrs. I upgraded a bit to an amd xp 3000+ and it finishes them in about 100 hours. Using the stock SETI AP app my Q6600 shows around 120 hours. The Lunatics AP distribution takes 12-14 hr per AP WU and yields around 1,250 credits. Well worth the small frustrations to get it running. Always assuming there is work available -- which there does not seem to be today. I'm out of SETI and AP work. |
Cosmic_Ocean Send message Joined: 23 Dec 00 Posts: 3027 Credit: 13,516,867 RAC: 13 |
That still seems very slow. I had an AMD sempron 2500+ that finished ap5.03 in about 160 hrs. I upgraded a bit to an amd xp 3000+ and it finishes them in about 100 hours. Yes. There was some minor confusion for me at first because the "old" AP made it to 5.00 (which you can see for yourself, here), and then the "new" version got a suffix of _v5 to make it a completely different app, thus avoiding the 4.35/5.00 fiasco a few months ago where those two versions would not validate against each other. Linux laptop: record uptime: 1511d 20h 19m (ended due to the power brick giving-up) |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
That still seems very slow. I had an AMD sempron 2500+ that finished ap5.03 in about 160 hrs. I upgraded a bit to an amd xp 3000+ and it finishes them in about 100 hours. Yes, the 5.03 build is officially "astropulse_v5" with friendly name "Astropulse v5". An AP at 110 hours and an AP_v5 at ~170 hours makes sense, the additional processing is about that much in stock. Upgrading to optimized would reduce the time for AP_v5 by about a factor of 4, and times on S@H Enhanced work by about a factor of 2. There's been a lot of discussion of special cases here which may make that change seem complicated, but in my view it consists of simple steps: 1. Download and unpack a combination package suitable for the system (check system capabilities with CPU-Z if needed). 2. Find where BOINC has put the projects\setiathome.berkeley.edu folder. 3. Copy the files from the unpacked "Files to install" folder to that project folder. 4. Exit and restart BOINC. If you're more comfortable with a detailed sequence with images, see the FAQ: SETI: Installing an optimised application written by Richard Haselgrove and Mark Sattler. Although written for S@H Enhanced it applies equally to the combined packages. I'm not urging you to go optimized, it implies a future obligation to watch for changes which not everyone is willing to accept, and it's definitely more complex than simply allowing BOINC to get the latest stock applications as they are released. But it isn't really difficult either. Joe |
Alan Smith Send message Joined: 19 Sep 00 Posts: 106 Credit: 26,816 RAC: 0 |
Just curious, but if the optimized apps are so much more efficient, why aren't they the standard apps? |
Raistmer Send message Joined: 16 Jun 01 Posts: 6325 Credit: 106,370,077 RAC: 121 |
I'm not urging you to go optimized, it implies a future obligation to watch for changes which not everyone is willing to accept IMHO if someone worries about how its host going, better to put this care in more concrete form by installing opt apps and care how they going by doing timely updates. "Set and forget" kind of person will not start threads here worrying about why his host goes so slow ;) |
Josef W. Segur Send message Joined: 30 Oct 99 Posts: 4504 Credit: 1,414,761 RAC: 0 |
Essentially the project doesn't have the time, staff, and money to make and test separate builds for hosts with varying capabilities. That means optimizations need to be either applicable to any kind of host or have some method of finding out if they should work on a specific host. The source is open, so many optimizations may eventually be fed back into it, S@H Enhanced on CPUs would be much slower without some code which has been contributed. That feedback process takes time and effort too, though. I did submit some fairly basic speed improvements for Astropulse, and there are more which I'll get around to some day soon, I hope. Those will probably get into a 5.05 or later build, the 5.04 being tested at the SETI@home/Astropulse Beta project doesn't have them. Joe |
Paul D Harris Send message Joined: 1 Dec 99 Posts: 1122 Credit: 33,600,005 RAC: 0 |
They are not developed by seti but are by independent individual developers not affiliated by seti staff similar in a way that a person modifies their car for better performance. |
Alan Smith Send message Joined: 19 Sep 00 Posts: 106 Credit: 26,816 RAC: 0 |
WHEW! Well, my 'Neverending WU' finally ended. The method for the estimate I got earlier from Josef W.Segur turned out to be surprisingly accurate. We estimated 168 hours (WAY down from my initial est of 700 hrs) and the actual ended up to be 166 hours. Did'ja ever notice just how long 166 hours is when your waiting for something? ;-) Guess the Astropulse jobs gotta go My computer is simply too slow I'm still gonna crunch M.B.s by the bunch But opt apps I just do not know Thanks to all... |
©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.