Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Thread Type: Sticky Thread Total posts in this thread: 124
|
![]() |
Author |
|
Jurgen Rapp
Cruncher Joined: Jan 30, 2005 Post Count: 13 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Hi and thanks for sharing. Small note. You don't have to enter exact percent fractions. e.g. 92% will round the allowed cores/threads to the nearest whole core, so with a quad setting 99 or 76% or 75% percent will all restrict cores to 3. Changes of allowed cores/threads is immediate upon a client-server connection when updating the website profiles or after doing an OK in Local preferences. On 6.10 and earlier clients there will be a benchmark. With 6.12 that's no longer the case (just tested). Happy Crunching. --//-- On my i7 920 (8 threads) 87% will run 6 tasks and 87.5% and 88% will run the expected 7 tasks so I'm not sure I would agree on the "nearest whole core" :) But I do understand how it works, but you have to set a value above the % for a given number of cores. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
On my i7 920 (8 threads) 87% will run 6 tasks and 87.5% and 88% will run the expected 7 tasks so I'm not sure I would agree on the "nearest whole core" :) Trivial pursuit: On an Intel 980x, 83.3% = 9 cores/threads; 83.4% = 10 cores/threads.But I do understand how it works, but you have to set a value above the % for a given number of cores. Benchmarking results made me (briefly) consider single-stepping from one core/thread up to 12 just to see the difference: 16-Oct-2011 13:33:09 [---] Benchmark results: 16-Oct-2011 13:33:09 [---] Number of CPUs: 9 16-Oct-2011 13:33:09 [---] 3377 floating point MIPS (Whetstone) per CPU 16-Oct-2011 13:33:09 [---] 8076 integer MIPS (Dhrystone) per CPU 16-Oct-2011 13:33:10 [---] Resuming computation 16-Oct-2011 13:33:42 [---] Benchmark results: 16-Oct-2011 13:33:42 [---] Number of CPUs: 10 16-Oct-2011 13:33:42 [---] 3333 floating point MIPS (Whetstone) per CPU 16-Oct-2011 13:33:42 [---] 8070 integer MIPS (Dhrystone) per CPU 16-Oct-2011 13:33:43 [---] Resuming computation Then I decided I was too lazy. (That's with data riding on a RAID 10 set, if that matters - for all I know, the benchmarking caches data in memory and so doesn't reflect disk I/O at all.) |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
When I was looking at this to crunch I had a concern over SSDs. Whilst they are a clear choice for processing heavy disc IO in terms of speed of processing, its at a long term price since SSDs are so dependent on a finite number of disc writes - compounded by the fact that an SSD will be accessed twice as much as a mechanical drive for each write due to the way SSDs work.
That made me cautious .... I ferreted around and then came across the solution - at least the solution for my situation. Diskeeper is without doubt the best solution for minimising fragmentation and disc writes out there. I have used it for years, magic stuff, true "fire and forget" software. So I had another peek at their site as I could not believe Diskeeper wouldnt be in the SSD world, and sure enough there sits Hyperfast, which is diskeeper on steroids designed for SSDs. The cost is only another £10 on top of the main Diskeeper product, no brainer. I installed it, and happily ran up to 8 of the WUs concurrently, together with a dual 5970 running full tilt on a demanding GPU WU using the other 4 cores dedicated to the GPUs. The SSD didnt even squeek, brilliant result, disc IO on the SSD has dropped through the floor, problem gone away, no screen lag whatsoever, and I can sleep at nights :) Am currently paused on these whilst I help out with a GFN Sieving problem at PrimeGrid, but will be back on them soon. Meanwhile, if anyone has similar thoughts or worries on SSDs, I cant recommend Hyperfast highly enough, and at £10 add on to main product, its not the end of the world if you end up not liking it - and I cant for the life of me see the latter happening, its really good stuff that does what it says on the tin, and more. You can in any case get it on 30 day trial, so that eliminates even that risk (trial is full product). Diskeeper - Hyperfast Regards Zy |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Dear Zydor,
thanks for the suggestion - this may clearly be an interesting option for other crunchers with SSD setups to look into. Best wishes Your Harvard CEP team |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hello ;
----------------------------------------I installed BOINC on a new computer, and I got errors : 23-Feb-2012 09:28:42 [World Community Grid] [error] File wcgrid_cep2_6.40_i686-pc-linux-gnu has wrong size: expected 1308060, got 926096 I think that "cep2" stands for Clean Energy Project. BTW I've no errors with other projects from worldcommunitygrid. [Edit 1 times, last edit by Former Member at Feb 23, 2012 3:10:30 PM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Dear aegir_fr,
it seems that something in the download of the application went wrong. Could you try to download CEP2 again? If this error is reproducible, it seems to indicate that maybe the CEP2 code repository at IBM/WCG is broken, in which case the IBM team should chime in. Best wishes Your Harvard CEP team |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Like RAM defragmenting, think SSD fragmenting is a load of cobblers, and the little coupled sale... you only want HF, but have to pay for DK to get at it... well I'm happy with free AusLogic. Anyway, on the cobblers, after writing the previous line, quick google and TomsHardwareStore has a few pennies to put in on this: http://www.tomshardware.com/news/diskeeper-ssd-defrag,6848.html ... snip "In fact, their numbers suggest that a HyperFast-enabled SSD will actually perform slower than one without the attached service.". The article mentions 2 alternate utilities worth considering. :D
----------------------------------------Whilst, as I wrote elsewhere, my new laptop with old tech SATA 5400RPM drive on W7-64 and Intel Rapid Storage Tech, does 98% efficiency when running 8 of these concurrent, initially with a staged start by hand after 1st or second checkpoint write [memory rated 1333 but per CPU-Z doing 667]. 99% or better when just on half the threads. Spinning drives are here to stay with ever denser storage... in fact IBM recently released a press note that they managed to get a reliable storage with just 12 atoms per bit on magnetic material... looking out for those 10-25TB pocket size drives in the proximate future. What on earth would we want with that [to fill it up of course ;-] --//-- edit: "half of the threads" [plural] --//-- [Edit 3 times, last edit by Former Member at Feb 24, 2012 4:35:11 PM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Dear aegir_fr, it seems that something in the download of the application went wrong. Could you try to download CEP2 again? If this error is reproducible, it seems to indicate that maybe the CEP2 code repository at IBM/WCG is broken, in which case the IBM team should chime in. Best wishes Your Harvard CEP team I cannot force CEP d/l ; the BOINC client asks some tasks to the scheduler ; and sometimes it got CEP2. I guess that I should disable all other worldcommunitygrid projects to force CEP2 d/l but I can't do it now because I've exceeded the d/l daily quota w/o sending results. BTW the error is reproductible, I had it yesterday when I set up this new client, and as you can see in the log I posted, I had it also this morning and this afternoon. Maybe it's a proxy issue ; I'll do some more tests to check that. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
hmmm, don't tell us you're using a v7 Alpha client? If the Result Status WU log on My Grid says error -200, then we know it's something to do with the download of the permanent science app files through the cloud [nothing wrong with the repository!].
--//-- |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
hmmm, don't tell us you're using a v7 Alpha client? If the Result Status WU log on My Grid says error -200, then we know it's something to do with the download of the permanent science app files through the cloud [nothing wrong with the repository!]. --//-- I use the client provided by the ubuntu 10.10 repository : Starting BOINC client version 6.10.58 for i686-pc-linux-gnu |
||
|
|
![]() |