| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: Suggestions / Feedback
Thread: BOINC credits vs WGC |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 109
|
|
| Author |
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Someone, Anyone?
----------------------------------------Friday 13, the rest of the world having a superstitious day, maybe WCG is afraid to go under by it's own success and likes to keep the hordes of cruncher at bay. 300 CPU years per day is a good number. ts05_ a043_ pca009_ 1-- 617 Valid 8/11/10 22:51:34 8/12/10 10:44:31 3.89 61.7 / 61.7 ts05_ a043_ pca009_ 0-- 617 Valid 8/11/10 22:51:31 8/13/10 15:44:25 4.59 124.0 / 61.7 < another miserable moment in my crunching live. Anything, Something should be easily fixable in the outlier rules. On next boot I'm going to hack the FPOPS down (not going to tell you how), half of something is more than half of nothing, in my world. Happy crunching... I'm going for a return to that top 1250 spot I once had before quad crunching commenced here in my corner of the world... today slipped to 1800.  Signed, sardonic me
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges: 
|
So what locks it down to 1.6 for benchmarking? I have no idea unless you can force the clock to 1.6 GHz always, which you cannot.Need plan C. With the economy mode I am afraid you have no chance since I remember having read that BOINC forces a normal priority during benchmarks. Edit: Added referenced quote ---------------------------------------- [Edit 1 times, last edit by JmBoullier at Aug 13, 2010 11:38:03 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Supposedly the newer clients are running the benchmark at the same idle priority to better reproduce the real crunching performance, but even the nice 19 kicks it up to 2.4, so it will be hard hacking the Dhrystone down to something like 6.900 which it is was when having the 32 bit client on windows. The current DDDT2 jobs have very short checkpoints, so it's kind of lossless and then find out what that will produce in claim and credit... which according calc is then translating to 19 per hour.
----------------------------------------
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Johnny Cool
Ace Cruncher USA Joined: Jul 28, 2005 Post Count: 8621 Status: Offline Project Badges: 
|
Someone, Anyone? Friday 13, the rest of the world having a superstitious day, maybe WCG is afraid to go under by it's own success and likes to keep the hordes of cruncher at bay. 300 CPU years per day is a good number. Signed, sardonic me Sek, maybe they think it's Friday the 12th.   I have probs getting "fair credit" for my 'Help Conquer Cancer' work units. My i7 870 is getting drubbed by older computers (even though I can crunch those work units *much faster*). I am getting more results in this project though.  ---------------------------------------- [Edit 1 times, last edit by Johnny Cool at Aug 13, 2010 5:46:06 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Well, I've licked the issue and found the sweetpoint. Not 12800 as the benchmark claims the device is able but 6850 **, essentially half of the regular Dhrystone portion of the test. This keeps it so that WCG is not deeming the ''outlier'' constantly slapped with the low claim, rather getting the mean of the quorum. I've scripted it now bashing it at a moments notice down. Particularly those riding with client version 6.4.5 and yes I see 5.4.xx and 5.8.xx still richly represented have this bad habit, particularly on Linux where these only benchmark 55% of their Windows equivalent.
----------------------------------------Guys, time to upgrade your client and ideally the 64 bit version if your OS is too! It may run, but broke it is never the less! ** Amazingly (not) this is practically spot on equal to the Windows Dhrystone test for this computer with a 64 bit client. edit: Yes, the evidence is mounting on my Result Status page... hit the sweetspot... no more outliers: ts05_ a055_ sr67b0_ 1-- 1292373 Valid 12-8-10 00:30:07 14-8-10 12:14:54 1.56 28.3 / 32.6 ts05_ a055_ sr78a0_ 1-- 1292373 Valid 12-8-10 00:30:07 14-8-10 11:53:25 1.56 28.3 / 25.6 ts05_ e033_ pe0000_ 1-- 1292373 Valid 11-8-10 21:56:04 14-8-10 10:59:51 15.97 289.7 / 301.7 ts05_ a069_ pqb009_ 0-- 1292373 Valid 12-8-10 01:07:26 14-8-10 10:48:48 4.75 86.2 / 86.2 ts05_ a054_ sr56b0_ 0-- 1292373 Pending Validation 12-8-10 00:28:22 14-8-10 10:36:52 1.58 28.6 / 0.0 ts05_ a054_ sr02b1_ 0-- 1292373 Valid 12-8-10 00:27:56 14-8-10 10:14:42 1.62 29.3 / 29.3 ts05_ a069_ pqa007_ 1-- 1292373 Valid 12-8-10 01:07:07 14-8-10 08:52:16 4.74 86.8 / 90.0
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Aug 14, 2010 2:20:42 PM] |
||
|
|
Dataman
Ace Cruncher Joined: Nov 16, 2004 Post Count: 4865 Status: Offline Project Badges: 
|
I especially love it (NOT!) when I get these:
----------------------------------------ts05_ b460_ pca002_ 0-- Dataman03 Valid 8/14/10 05:43:17 8/14/10 12:13:43 3.69 73.8 / 42.5 ts05_ b406_ pr78a0_ 2-- Dataman04 Valid 8/14/10 05:42:44 8/14/10 11:13:38 3.97 79.7 / 38.2 These are the worst but in sampling the first ~50 valids, only 3 were >= to claimed. I should have used the electricity to cool ME!  Win7-64, 6.10.58, i920   |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges:
|
Time to upgrade your client? Really?
----------------------------------------ts05_ d437_ pe0000_ 3-- 617 Valide 12/08/10 18:05:03 13/08/10 23:27:31 12,53 345,8 / 164,1<-- mine 6.10.17 ts05_ d437_ pe0000_ 2-- 617 Valide 12/08/10 04:55:57 12/08/10 19:58:13 14,10 164,1 / 164,1 <-- 6.10.56 And this one, both with 6.10.17: ts05_ a262_ pe0000_ 1-- 617 Valide 11/08/10 19:00:49 13/08/10 04:49:28 15,45 197,8 / 197,8 ts05_ a262_ pe0000_ 0-- 617 Valide 11/08/10 19:00:42 12/08/10 17:02:26 12,68 381,2 / 197,8 <-- mine An other interesting case. Why try to fry one's processor if it is for claiming peanuts? ts05_ d387_ pe0000_ 2-- 617 Valide 12/08/10 18:04:44 13/08/10 11:52:03 12,60 368,3 / 166,1 <-- mine 6.10.17 ts05_ d387_ pe0000_ 0-- 617 Valide 11/08/10 21:47:38 13/08/10 07:04:16 10,25 166,1 / 166,1 <-- fast wingman with an old 6.2.15 It seems that the benchmark of 32-bit Linux clients need some attention. Sigh... |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
By the client version number you infer 32 or 64 bit? I did say "and ideally the 64 bit version if your OS is too!"
----------------------------------------Actually the Linux 32 bit client benchmarks identical to the 64 bit Windows client both on Whet and Dhrystone [on my 64 bit Linux], so one could also argue that it's the 64 bit client that needs to have some sense knocked in, but really think the underlying issue could be the (Debian) compile that is circulated via the SPM (Synaptic) with the highly inflated integer portion. Anyway, done 4 more and still nicely the averages keep rolling in. My RAC is already showing the effect, so until... ... maybe homogeneous redundancy could be extended to match 32:32 and 64:64 ... Major :P
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 1 times, last edit by Sekerob at Aug 14, 2010 3:54:56 PM] |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3715 Status: Offline Project Badges:
|
By the client version number you infer 32 or 64 bit? No, it's from my second example where both copies were computed under version 6.10.17. Under the same system both benchmark routines should be identical and give close claims. We are far from this and that brings me memories of my very first installation of Ubuntu. It was in 32-bit mode and the benchmark was ridiculous. Much below what it was under XP 32-bit for the same machine at the same speed.And discrepancies in the other examples seem to be in the same range. In the third example my wingman is claiming 16.2 credits per hour with a machine probably clocked at least at 3.8 GHz, or probably more since there is the performance difference 32-vs-64 to compensate. That said, I agree that the 64-bit benchmark of 6.10.17 seems to be too high compared to the previous versions I have used. But not twice more. ---------------------------------------- [Edit 1 times, last edit by JmBoullier at Aug 15, 2010 2:37:11 AM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Almost doubtlessly a .sh install versus a Synaptic install and then 32 bit weighing up against a 64 bit client. Found the 32 bit .sh benchmarks dramatically lower on integer. The Whetstone/floating point seems to only marginally vary, which is apparently dictated by the fact that those registers are 32 bit in hardware. When double, it's the Dhrystone/Integer. 6800 for Windows 64 bit, 12850 for 64 bit linux and 6500 if the package is used and if I remember well the 32 bit Vista gave about 4400, which then makes it nearly triple on Linux with the 64 bit Synaptic kit. Ludacris... but then HCC does go double speed which does not translate in double credit per hour.
----------------------------------------Anyway, I've now had a whole bunch where the low claiming wingman was dismissed and my quad claim was declared canonical... consistent very close to historic :P (keeping an eye out for the next scheduled benchmark and if reverting will rerun the script to half the iops rating. Someone on the other side will probably be okay with those bonuses ;o)
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
|