Message boards : Number crunching : Target CPU run time
Author | Message |
---|---|
Insidious![]() Send message Joined: 10 Nov 05 Posts: 49 Credit: 604,937 RAC: 0 |
Hi folks, I'll be happy to set this preference wherever the project team feels it will do the most good. I have reasonably fast CPUs (AMD X2 @ 2.5GHz) and lots of RAM available. (2GB on two machines and 1GB on the other) I have NO idea what so ever where to set target time to provide the most benefit to Rosetta@Home. Can any one provide guidance to me. Thanks in Advance -Sid Proudly crunching with TeAm Anandtech |
![]() ![]() Send message Joined: 30 Dec 05 Posts: 1755 Credit: 4,690,520 RAC: 0 |
It is really more of a personal preference... otherwise they wouldn't make it so flexible. The less you have to hit the servers the better. Having enough work on hand to assure you don't find yourself idle when you lose internet access or servers are under maintenance is always a good idea. My suggestion is to set the runtime preference to a value where you complete a result per day on each CPU. So if you run R@H with a 60% resource share on a machine that is on 24 hrs, I might set to 16hrs. Or, if you're like me and run 100% R@H, I use 24hr WU pref. Be sure to review the ways that changing the preference can throw off BOINC for a short time, until it's crunched and reported a few of the WUs under the new runtime pref. Make change gradually to assure a smooth transition. For others that are curious, the above is not based on the high memory nor the faster CPUs that Insidious has. Rosetta work is not of a fixed size. If you have a slower machine, it just gets less of the possible models of the WU completed. But still tries to run within your runtime preference. The only fixed portion to it is that a WU must complete at least one model, regardless of how long that takes on your machine, and regardless of your runtime pref. So, if you set the pref. very low, you are more likely to see WUs take significantly longer then your pref. For example a 1hr pref. will often see docking work run for 2 or 3 hours. It takes that long to complete the first model. Add this signature to your EMail: Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might! https://boinc.bakerlab.org/rosetta/ |
Michael G.R. Send message Joined: 11 Nov 05 Posts: 264 Credit: 11,247,510 RAC: 0 |
Thanks for the explanation, Feet1st. |
Insidious![]() Send message Joined: 10 Nov 05 Posts: 49 Credit: 604,937 RAC: 0 |
Thanks Feet1st !!!!! Exactly the kind of help I was in need of. -Sid Proudly crunching with TeAm Anandtech |
![]() ![]() Send message Joined: 15 Dec 05 Posts: 761 Credit: 285,578 RAC: 0 |
... The only fixed portion to it is that a WU must complete at least one model, ... and the only variable point to know is that the app will only stop at the end of a "decoy", so don't expect the times to fit exactly to what you asked. On a slow box some WU can vary a lot. For example we recently had WU that took over four hors per decoy on my 667MHz box. With a stting of 24hrs these could end anywhere from 19hrs to 27hrs. At the other extreme we have had WU that fitted in over 200 decoys/hour on a fast machine - these WU on that machine would end to within a few seconds of the user pref. River~~ ![]() |
![]() ![]() Send message Joined: 30 Dec 05 Posts: 1755 Credit: 4,690,520 RAC: 0 |
...With a stting of 24hrs these could end anywhere from 19hrs to 27hrs. In my experience with 24hr runtime preference, it will cut out early, rather then late. So it is very rare to see anything over 24hrs. ...so the key point here is that slow machines still may crunch a 16hr work unit in 16hours. But, as River points out, they must end on a completed model, and must complete at least one model. So, if you have a fast machine with a 16hr preference, it may crunch 5 models on a difficult docking WU. A slower machine will still complete in the 16hrs, but it may only compute 1 or 2 models. This is why a low runtime preference on a slow machine gives you more disparity in runtimes. Add this signature to your EMail: Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might! https://boinc.bakerlab.org/rosetta/ |
![]() ![]() Send message Joined: 15 Dec 05 Posts: 761 Credit: 285,578 RAC: 0 |
...With a stting of 24hrs these could end anywhere from 19hrs to 27hrs. Yes, it won't start another decoy unless it is confident it can finish, and it makes that decision on the basis of the average (or max ??) decoy seen so far this run. The reason I wonder if it goes on the max seen so far is that, as Feet1st correctly says, they almost always run short rather than long. But it is still possible that the last decoy runs for so much longer than the average seen so far, in which case it will overrun. The 27hr WU I had had only run four decoys, so plenty of room for statistical variation. |
cloaked_chaos Send message Joined: 9 Nov 05 Posts: 14 Credit: 80,818 RAC: 0 |
Thanks Feet1st !!!!! Insidious, I was just wondering which optimized client you are using on your X2 3800's. I don't know if you will see this, but this was your most recent posting. If you don't feel like telling everyone, then please email it to me at motoguzzie123@aol.com |
FluffyChicken![]() Send message Joined: 1 Nov 05 Posts: 1260 Credit: 369,635 RAC: 0 |
Thanks Feet1st !!!!! It looks like he is using 5.5.0 on two of them (most probably crunch3rs old one) and 5.4.11 on the 3rd one (looks like standard client) Team mauisun.org |
Insidious![]() Send message Joined: 10 Nov 05 Posts: 49 Credit: 604,937 RAC: 0 |
Thanks Feet1st !!!!! Hi Cloaked Chaos Fluffy Chicken has it right. It's the old Crunch3R 5.5 on two of them. I can't get to the third machine as I gave it to my son for a graduation present and he has it with him at college. Are you in need of an optimized client? Post here if so, and I'll try to help. -Sid Proudly crunching with TeAm Anandtech |
FluffyChicken![]() Send message Joined: 1 Nov 05 Posts: 1260 Credit: 369,635 RAC: 0 |
You can get various newer client than 5.5.0 (and the official 5.4.11) from dumas. Well as long as you can use SSE2 (which if it's the A64's then you can) https://boinc.bakerlab.org/rosetta/forum_thread.php?id=2549 If you want crunchers then do a search Team mauisun.org |
cloaked_chaos Send message Joined: 9 Nov 05 Posts: 14 Credit: 80,818 RAC: 0 |
I have been using crunch3r's optimizers for a long time. I was wanting a newer version since my latest archived version is for 5.2.13 (both 32bit/P3 and 64bit/P4 versions) |
FluffyChicken![]() Send message Joined: 1 Nov 05 Posts: 1260 Credit: 369,635 RAC: 0 |
I have been using crunch3r's optimizers for a long time. I was wanting a newer version since my latest archived version is for 5.2.13 (both 32bit/P3 and 64bit/P4 versions) Looking at the ptojects you are running there is no point in using them if it's to claim increades credit, especially as it seems you've switched to a dedicated Rosetta@home. They all use a different benchmarking system to the internal boinc one. So Dumas's maybe your better choice since he has (afaik) also compiled wxwidget and boincmanager against SEE2 and 32/64 bit. Team mauisun.org |
Message boards :
Number crunching :
Target CPU run time
©2024 University of Washington
https://www.bakerlab.org