bigger WUs
Message boards :
Number crunching :
bigger WUs
Message board moderation
Previous · 1 · 2
| Author | Message |
|---|---|
 ML1 Send message Joined: 25 Nov 01 Posts: 21790 Credit: 7,508,002 RAC: 20 
 |
Well it has been suggested to the devs, and they thinking about it, compressed WUs that are batched and stored in a compressed state. Say 20 WUs at a time... Or even a x20 sized WU all in one gulp!? At present, there is a time overlap at the start and end of each WU between the WUs so that the signal analysis can be continuous. Processing 20 WUs end-to-end as one 20xWU would eliminate 19 out of 21 time overlaps within that 20xWU compared to 20 normal WUs. We get more efficient processing whilst also getting longer WUs to reduce the interactions with the Berkeley servers. Is this a good idea to be tried as soon as the s@h-enhanced client is running? (Dial-up users would have to be given the option to stay with normal (small) WUs.) Regards, Martin See new freedom: Mageia Linux Take a look for yourself: Linux Format The Future is what We all make IT (GPLv3) ID: 152356 ·  |
Jord  Send message Joined: 9 Jun 99 Posts: 15184 Credit: 4,362,181 RAC: 3 
|
"unoptimized-client" The un-optimized client is the Boinc Client as you run it. Some people have brought out "optimized clients", which were compiled using other compling programs than the one Berkeley uses. Then there are un-optimized science applications versus the optimized ones. Same trek there, but for that the optimized science apps run faster than the un-optimized ones. Combine the 4.45 optimized clients and 4.18 (4.11) optimized SA's together and you should run faster units. I must say, I run one of the optimized SA's and normal client 4.72, and I don't see much difference in times. But that's something completely different and for CC4.73 of Boinc to solve. ;) ID: 152381 · |
[B@H] Ray  Send message Joined: 1 Sep 00 Posts: 485 Credit: 45,275 RAC: 0 
|
"unoptimized-client" Diferant CPU's have diferant insteruction sets built into them. Each optomized program is designed to make up of a certain instruction set rather than the one size fits all that you get when you join. They give the same results but faster as you get one designed for the type of CPU that you have. I have a P4 2.4 Gig SSE2 processor, runs a unit in about 3¼ hours with the standard client. With one optomized for SSE2 thay take about 1 hour 40 Min. to finish up. Some newer processors do them in about a hour. Ray  Pizza@Home Rays Place Rays place Forums ID: 152396 · |
ghstwolf  Send message Joined: 14 Oct 04 Posts: 322 Credit: 55,806 RAC: 0
|
Some of us are already testing Seti Beta units. They take around 133,000 - 140,000+ seconds on a P4, are you sure you want them to be enabled in here? ;) It sounds like by production time that should be down to about 100,000 seconds. Raise the WU size to 500Kb, while your at it, and it becomes a 40hr WU. Of course I've made a couple assumptions here (room for general optimization and linear time increase with WU size), but that sounds pretty good to me. With my current allotment, 1 WU will take about 4 days. Works for me.  Still looking for something profound or inspirational to place here. ID: 152402 · |
W-K 666  Send message Joined: 18 May 99 Posts: 19744 Credit: 40,757,560 RAC: 67
 |
I'm running both Seti and Seti Beta with optimized clients. My average crunch times are: Seti - 1hr:20min Claimed credits - 14 approx Seti Beta - 12h:30min Claimed credits - 155 approx Andy ID: 152474 · |
Astro  Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
Yes, Andy, I'm seeing the same ratio between seti and seti/beta: 1:04 seti (optimized) 10:40 seti/beta (optimized) a 10x ratio  ID: 152576 · |
|
W-K 666 Send message Joined: 18 May 99 Posts: 19744 Credit: 40,757,560 RAC: 67
|
What I forgot to say is that if you monitor the progress, at all, the first 50% takes about 80% of the time, 10hr in my case and then does the final 50% in 2h:20m. And during this it may predict that the total completion time is over 20hrs. Andy ID: 152620 · |
|
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
What I forgot to say is that if you monitor the progress, at all, the first 50% takes about 80% of the time, 10hr in my case and then does the final 50% in 2h:20m. And during this it may predict that the total completion time is over 20hrs. Andy, I was at the beta site yesterday, and if I recall correctly, JM7 has found this problem and submitted a fix. I'm 70% sure of this, but to lazy to go and reread it. tony ID: 152635 · |
|
W-K 666 Send message Joined: 18 May 99 Posts: 19744 Credit: 40,757,560 RAC: 67
|
.... JM7 has found this problem and submitted a fix. I'm 70% sure of this, but to lazy to go and reread it. Hi Tony, I'd also read that JM7 had spotted the error and was working on a fix, but must admit I havn't been there for some time. Andy P.S. Ref Bob Newhart - ever thought of being a driving instructor? ID: 152652 · |
|
Astro Send message Joined: 16 Apr 02 Posts: 8026 Credit: 600,015 RAC: 0 |
P.S. Ref Bob Newhart - ever thought of being a driving instructor? Every time I try to "instruct" my wife, I get a beating. Each time I tried with my daughter, I got "quit being so technical"(doesn't every driver NEED to know what happens during each of the 4 strokes of an engine?). Maybe it'd be different with strangers. ID: 152655 · |
John McLeod VII  Send message Joined: 15 Jul 99 Posts: 24806 Credit: 790,712 RAC: 0
|
What I forgot to say is that if you monitor the progress, at all, the first 50% takes about 80% of the time, 10hr in my case and then does the final 50% in 2h:20m. And during this it may predict that the total completion time is over 20hrs. Not for this problem. For the problem where the estimated time did stopped shrinking after each result long before it should have.   BOINC WIKI ID: 152778 · |
©2025 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.