Message boards :
Number crunching :
Here's a question about the P4 HT machines
Message board moderation
Author | Message |
---|---|
Cosmos Kramer Send message Joined: 24 May 01 Posts: 22 Credit: 2,740,194 RAC: 176 |
I notice that it would process 2 units at once. I know with Seti Driver, processing 2 at once would create more results, but the time it took to crunch each were longer than with just one at a time. I was wondering if this would react similar? I am only running one now, to see how much difference there is. <img src="http://www.setisynergy.com/images/stats/197876sah.png"></img> |
Cosmos Kramer Send message Joined: 24 May 01 Posts: 22 Credit: 2,740,194 RAC: 176 |
Just noticed another problem. When I try to use Nero to read/burn something, BOINC stops, even though I have it selected to run all the time. <img src="http://www.setisynergy.com/images/stats/197876sah.png"></img> |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
Hawk, HT = 2 Pretend CPUs on One real CPU (as you said). When it is setup to pretend to be two, WUs run in parallel, but they share the same CPU Level 2 Cache, and RAM...and so are not quite twice as fast as one CPU. To do the test you are discussing, I suggest you restart the machine, and go into the BIOS and turn off Hyperthreading for a true "one CPU" one thread test. ...or...you could do 3 tests. 1. HT on - 2 Threads (normal) 2. HT on - 1 Thread (ie use at most 1 CPU on preferences) 3. HT off - 1 Thread I, for one, would be interested to know your results. |
Cosmos Kramer Send message Joined: 24 May 01 Posts: 22 Credit: 2,740,194 RAC: 176 |
I have an HP machine, so I am thinking I can't turn HT off. I don't recall seeing it in the Bios but I can check later. I would still like advice on the problem with the crunching stopping while Nero is running. I normally leave my pc on 24/7 crunching, so I hate to see it stop for any reason. <img src="http://www.setisynergy.com/images/stats/197876sah.png"></img> |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
> Just noticed another problem. When I try to use Nero to read/burn something, > BOINC stops, even though I have it selected to run all the time. > BOINC (science application) is set to run at the absolute lowest priority...and Nero is set to run at highest (when it sets up read from HDD and writes to CD/DVD burner) so BOINC might not get much CPU time. I've found with just 4x DVD burning BOINC gets some of the CPU time...but I was using Easy CD creator 5.x. |
Cosmos Kramer Send message Joined: 24 May 01 Posts: 22 Credit: 2,740,194 RAC: 176 |
Just noticed that even though I set BOINC and Nero priority at normal under task manager, NERO is still stealing 98-99% of the CPU. Argh. <img src="http://www.setisynergy.com/images/stats/197876sah.png"></img> |
Cosmos Kramer Send message Joined: 24 May 01 Posts: 22 Credit: 2,740,194 RAC: 176 |
I do a lot of reading/burning, so this ain't good. <img src="http://www.setisynergy.com/images/stats/197876sah.png"></img> |
STE\/E Send message Joined: 29 Mar 03 Posts: 1137 Credit: 5,334,063 RAC: 0 |
Just noticed that even though I set BOINC and Nero priority at normal under task manager, NERO is still stealing 98-99% of the CPU. Argh. ========== It's either that or NERO simply is not going to run & you will end up locking up your Computer ... |
Cosmos Kramer Send message Joined: 24 May 01 Posts: 22 Credit: 2,740,194 RAC: 176 |
Nero always ran with the old Seti, but alas... Anyway, I found a workaround and DVD Shrink seems to work with Boinc pretty well and it has a burner also. Well, it uses the Nero burning ROM, but it seems to work. Kewl. <img src="http://www.setisynergy.com/images/stats/197876sah.png"></img> |
EdwardPF Send message Joined: 26 Jul 99 Posts: 389 Credit: 236,772,605 RAC: 374 |
This may - or may not - help in regards to undrtstanding HT and SETI-classic. I have 2 identical 3.0Gh P4's with 1Gb memory Before BOINC, I ran one with HT on and the other with HT off ( for a short test). I understand that the mix of WU's were not identical and don't count the same as in BOINC but it's the best test I have ... WU's Total CPU time Avg CPU time W/HT on 38 123h37m 3h15m11.6s W/HT off 23 55h33m 2h24m57.2s Just fyi ... EdwardPF > Hawk, > > HT = 2 Pretend CPUs on One real CPU (as you said). > > When it is setup to pretend to be two, WUs run in parallel, but they share the > same CPU Level 2 Cache, and RAM...and so are not quite twice as fast as one > CPU. > > To do the test you are discussing, I suggest you restart the machine, and go > into the BIOS and turn off Hyperthreading for a true "one CPU" one thread > test. > > ...or...you could do 3 tests. > 1. HT on - 2 Threads (normal) > 2. HT on - 1 Thread (ie use at most 1 CPU on preferences) > 3. HT off - 1 Thread > > I, for one, would be interested to know your results. > |
Cosmos Kramer Send message Joined: 24 May 01 Posts: 22 Credit: 2,740,194 RAC: 176 |
Interesting. I would have thought HT on would be faster. <img src="http://www.setisynergy.com/images/stats/197876sah.png"></img> |
Cosmos Kramer Send message Joined: 24 May 01 Posts: 22 Credit: 2,740,194 RAC: 176 |
Well I'm not seeing an option in the Bios to turn off HT. Are there any programs that might do it? I'm thinking no, but I would like to test it. <img src="http://www.setisynergy.com/images/stats/197876sah.png"></img> |
Hans Dorn Send message Joined: 3 Apr 99 Posts: 2262 Credit: 26,448,570 RAC: 0 |
> Interesting. I would have thought HT on would be faster. > HT is faster, actually. Since boinc runs on 2 virtual CPUs, doing 2 WUs in parallel, you have to divide the CPU time by 2 to get the real performance. Regards Hans |
SURVEYOR Send message Joined: 19 Oct 02 Posts: 375 Credit: 608,422 RAC: 0 |
WITH MY 3.40G HT i COMPLETE TWO WU IN ABOUT 3HR 30MIN TO 3HR 45MIN WITH MY 2.53 G NO HT i COMPLETE ONE WU IN ABOUT 3HR 12MIN TO 3HR 45MIN Fred BOINC Alpha, BOINC Beta, LHC Alpha, Einstein Alpha |
Helli_retiered Send message Joined: 15 Dec 99 Posts: 707 Credit: 108,785,585 RAC: 0 |
> Since boinc runs on 2 virtual CPUs, doing 2 WUs in parallel, you have to > divide the CPU time by 2 to get the real performance. That is wrong. You crunch 2 WUS at the same time with HT, not one behind the other. Thus each WU needs 3h15. Example: Two Sprinters arrive after 10 seconds at the same time at the goal, That means that each Sprinter needed 10 seconds for the distance and not 5sec. |
Dunc Send message Joined: 3 Jul 02 Posts: 129 Credit: 2,166,460 RAC: 0 |
> > Since boinc runs on 2 virtual CPUs, doing 2 WUs in parallel, you have to > > divide the CPU time by 2 to get the real performance. > > That is wrong. You crunch 2 WUS at the same time with HT, not one behind > the other. Thus each WU needs 3h15. > > Example: Two Sprinters arrive after 10 seconds at the same time at the goal, > That means that each Sprinter needed 10 seconds for the distance and not 5sec. > If it takes Boinc 3 hours to crunch 2 wus at 3 hours each, that is the equivalent of 1 hour and 30 minutes each as you have produced 2wus in 3 hours. I have a machine with 2 actual, and 2 virtual processors. Each wu takes about 3 hours. Therefore the 'real' time per wu is 45 minutes each. Or 32 wus per day. Dunc |
Helli_retiered Send message Joined: 15 Dec 99 Posts: 707 Credit: 108,785,585 RAC: 0 |
LOL |
7822531 Send message Joined: 3 Apr 99 Posts: 820 Credit: 692 RAC: 0 |
If memory serves, the difference between a hyper-threading and non-hyper-threading computer lies in the processor's firmware, which is on the CPU - So I have to agree. Hyper-threading aside, a uni-processing computer has only one physical processor, and that's what ought to count. So if you have one workunit completed in three hours by one processor pretending to be two, it's still three hours for the one physical processor. Just my two cents. |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
I love these discussions that confuse processing time with throughput ... |
7822531 Send message Joined: 3 Apr 99 Posts: 820 Credit: 692 RAC: 0 |
My apologies, then. Hyper-threading being considered a doubling of processing as opposed to a processing method is what is causing my confusion. |
©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.