The Great Crunchoff Grandstand
Message boards :
Number crunching :
The Great Crunchoff Grandstand
Message board moderation
| Author | Message |
|---|---|
|
Alinator Send message Joined: 7 Jun 08 Posts: 464 Credit: 56,639,936 RAC: 0  
|
I see CFL has started the Documentation thread for his crunchoff experiemnt. I suggest that we try to keep that one for updates from him and direct observations and questions about the experiment. Please use this one for all other comparisons and commentary about the run. <edit> @ CFL: I saw your last post last night about the problems getting it all set up. It's just goes to prove the old adage, "Nothing is easy. If it was, everyone would be doing it!". Anyway, it looks like we are off to the races now! :-D <edit2> I just noticed Kenzie's running a crunchoff as well, so I guess we can use the Grandstand to cheer her on too! ;-) Alinator |
 Kevint Send message Joined: 22 Nov 07 Posts: 285 Credit: 1,076,786,368 RAC: 0 
|
I see CFL has started the Documentation thread for his crunchoff experiemnt. I have set the cache on both boxes to 8 days. Not that MW will use it, but with the problems that SETI has, I thought it might be wise to at least have a cache of SETI there. . 
|
GalaxyIce Send message Joined: 6 Apr 08 Posts: 2018 Credit: 100,142,856 RAC: 0 
|
Anyway, it looks like we are off to the races now! :-D I bet a tenner in MW to win :/ |
|
Alinator Send message Joined: 7 Jun 08 Posts: 464 Credit: 56,639,936 RAC: 0
|
FWIW: I set up all my hosts to run my project set with equal resource shares several months ago to collect comparative data. In general, it is not as rigorous a test as CFL's, since all these machines have other assigned duties on my personal network and thus are not what I would classify as 'dedicated' crunch boxes. Anyway, here are the current details (2/22/09) for them, but they aren't hidden on any project they run. So if you're from Missouri, you can check them yourself. ;-) Application is stock unless otherwise noted. Unit 1: 400 MHz G3 iMac running Panther, 512 MB RAM CPCS: MW: 0.001255 EAH: 0.000431 SAH: 0.000332 LC: No compatible application Unit 2: 1.83 GHz T2400 (CD Yonah) running XP Pro SP3, 2 GB RAM MW: 0.004569 EAH: 0.004191 SAH: 0.004789 (AKv8 SSE3) LC: 0.002666 Unit 3: 2.66 GHz P4 (Northwood) running XP Pro SP3, 1 GB RAM MW: 0.005952 EAH: 0.003282 SAH: 0.004758 (AKv8 SSE2) LC: 0.002319 Unit 4: 550 MHz PIII (Katmai) running 2K Pro SP4, 384 MB RAM MW: 0.001541 EAH: 0.000652 SAH: 0.000701 (AKv8 SSE) LC: 0.000753 Unit 5: 450 MHz PII (Deschutes) running XP Pro SP3, 384 MB RAM MW: 0.001579 EAH: 0.000467 SAH: 0.000411 (KWSN 2.4 MMX) LC: 0.000637 Unit 6: 500 MHz K6-2 running 2K Pro SP4, 256 MB RAM MW: 0.001568 (Gipsel 0.19)* EAH: 0.000287 SAH: 0.000196 (KWSN 2.4 MMX) LC: 0.000567 Unit 7: 450 MHz K6-3 running 2K Server SP4, 384 MB RAM MW: 0.001129 ** EAH: 0.000354 SAH: 0.000175 (KWSN 2.4 MMX) LC: 0.000548 Unit 8: 300 MHz K6 running NT4 Server SP6a, 192 MB RAM MW: 0.000857 EAH: No compatible application SAH: 0.000126 (KWSN 2.4 MMX) LC: 0.000335 * Gipsel app used for dual boot 9x compatibility ** either Gipsel or zslip 0.19 used for same reason as before, just not sure which right now. ;-) The other two hosts I have are K6-2/500's, generally similar to but not identical to the one listed here. The one I chose is the one which consistently performs the best of the three. Alinator |
|
GalaxyIce Send message Joined: 6 Apr 08 Posts: 2018 Credit: 100,142,856 RAC: 0
|
** either Gipsel or zslip 0.19 used for same reason as before, just not sure which right now. ;-) zslip hosts a copy of Gipsel's 0.19 - they are both Gipsel's ;) |
|
Alinator Send message Joined: 7 Jun 08 Posts: 464 Credit: 56,639,936 RAC: 0
|
OK, that's what I thought, but mentioned it for thoroughness. ;-) Alinator |
caspr Send message Joined: 22 Mar 08 Posts: 90 Credit: 501,728 RAC: 0 
|
@ Alinator: You mention Missouri, is that because it's "The Show Me State?" Or are you from Mo.? A clear conscience is usually the sign of a bad memory  |
|
Alinator Send message Joined: 7 Jun 08 Posts: 464 Credit: 56,639,936 RAC: 0
|
Both. ;-) |
speedimic Send message Joined: 22 Feb 08 Posts: 260 Credit: 57,387,048 RAC: 0
|
LOL, seems I kicked off something here... Anyway, my 50/50 test box if ready too: It's also a Q6600 with 2GB RAM with a fresh BOINC install, running Seti stock and MW stock at 50/50 rate. Host at Milkyway same Host at Seti On the Seti side it's set to crunch Multibeam only, but from my experience it'll take at least 4 weeks before the Seti RAC get kind of stabe. [edit] just to have it in one place: CampaignForLiberty's post Kenzie's post [/edit] mic. 
|
|
Alinator Send message Joined: 7 Jun 08 Posts: 464 Credit: 56,639,936 RAC: 0
|
That's why we're using reported CPCS. It isn't subject to availability issues for the most part. Of course you do have to able to get and report some work, which hasn't been easy at SAH this weekend. ;-) Anyway, the more the merrier for this kind of experiment. :-) Alinator |
|
speedimic Send message Joined: 22 Feb 08 Posts: 260 Credit: 57,387,048 RAC: 0
|
That's why we're using reported CPCS. It isn't subject to availability issues for the most part. Of course you do have to able to get and report some work, which hasn't been easy at SAH this weekend. ;-) Strange enough, I didn't have any problem getting/reporting Seti WUs on this host - all other Seti-host are still trying to upload friday's work... mic.
|
|
Brian Silvers Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0
|
Unit 2: I picked those two hosts out of your list because they are most indicative of the more popular CPUs out there now... One thing that you didn't put as a disclaimer is that EAH has a "stock" application that is auto-detecting what feature set the processors have, so the "stock" there is running the "_2" (SSE2) application on both systems, unless you have forced it to run at a lower level. With that in mind, that validates what I've suspected for quite some time now, that EAH is significantly underpaying for most SSE2 and higher machines when compared to what you can get at SETI. As for here, I didn't think SETI was going to be paying out more. In fact, I figured MW would be about 1.6-1.8X. Right now, with only 1 reported result at SETI, the MW system is showing about 1.9X, but I believe the task reported at SETI was one of the lower paying tasks, so the multiple should go down over time. If that holds true, then we get into whether or not more actual work is being done by the science application to where it deserves to be gaining a higher payout, and then the additional "alpha correction"... |
|
Alinator Send message Joined: 7 Jun 08 Posts: 464 Credit: 56,639,936 RAC: 0
|
Strange enough, I didn't have any problem getting/reporting Seti WUs on this host - all other Seti-host are still trying to upload friday's work... Yep, I guess if you were doing Astropulse you could get through. I'm just doing Multibeam currently, so I was dead in the water except for running cached work. They seem to be getting things cleared now. I was able to force an upload/report for a task on Unit 6 which had gone over deadline while cooling its heels in the upload queue for two days! Whew! ;-) Alinator |
Cori Send message Joined: 27 Aug 07 Posts: 647 Credit: 27,592,547 RAC: 0  
|
That's why we're using reported CPCS... Sorry, I don't get it right now... what is CPCS standing for? (I am no native English speaker so some abbreviations just puzzle me. ;-))) Lovely greetings, Cori  
|
|
Alinator Send message Joined: 7 Jun 08 Posts: 464 Credit: 56,639,936 RAC: 0
|
Point taken about EAH, and it is the official 'wrapper' stock app which I'm currently running. :-) If I have been following the discussion correctly here, MW is planning to have something similar when they get to the same point of development. Moral of the story is, you have to make sure to specify what app you're running and any user tweaks you may have put for the other projects in these discussions. ;-) <edit> I see I neglected to mention BOINC CC version. Unit 1: 5.10.38 Unit 8: 5.8.16 All the rest are 5.10.13. Alinator |
|
Alinator Send message Joined: 7 Jun 08 Posts: 464 Credit: 56,639,936 RAC: 0
|
That's why we're using reported CPCS... Sorry.... CPCS = Credit Per Cpu Second. Alinator |
|
Cori Send message Joined: 27 Aug 07 Posts: 647 Credit: 27,592,547 RAC: 0
|
That's why we're using reported CPCS... Thanks, mate!!! :-) I got up to "credits per..." and then the puzzle started. *LOL* Lovely greetings, Cori
|
|
Brian Silvers Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0
|
That's why we're using reported CPCS... Should we tell them that you love to use TLAs? ;-) |
|
Cori Send message Joined: 27 Aug 07 Posts: 647 Credit: 27,592,547 RAC: 0
|
That's why we're using reported CPCS... *ROFL* Funny enough I know TLA (Three Letter Acronym). :-D Lovely greetings, Cori
|
|
C Send message Joined: 22 Nov 07 Posts: 36 Credit: 1,224,316 RAC: 0 
|
That's why we're using reported CPCS... Based on the size of the numbers, I think it means Credits Per CPU-Sec, or credits per cpu-second. C Team MacNN |
©2023 Astroinformatics Group