| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 42
|
|
| Author |
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Yep, this is your misfortune of teaming up with 2 fast machines and really evidences the concept of quorum 3. What is agreeably wrong is that these guys should have been getting around 50/60 for the job...that would have been if under Woz...... This is where R@H has become impartial....cant care less as to what OS it was run on..... looking forward to a credit system revision.
----------------------------------------cheers
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
Electrolyte
Cruncher Joined: Jun 20, 2006 Post Count: 13 Status: Offline |
After a few experiments while compiling and optimising the BOINC client, I've decided to post this on the BOINC forum to see what they think of it. You can find the thread here: http://boinc.berkeley.edu/dev/forum_thread.php?id=1227
----------------------------------------![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Over on Rosetta@home, mmciastro has posted a benchmark comparison chart (that requires a lot of bandwidth, he posted a graphics file) that shows the points per hour for a lot of different chips and operating systems running BOINC. Linux usually only awards 55% as many points as Windows. http://boinc.bakerlab.org/rosetta/forum_thread.php?id=2415
----------------------------------------Lawrence [Edit 1 times, last edit by Former Member at Oct 15, 2006 3:33:55 AM] |
||
|
|
Electrolyte
Cruncher Joined: Jun 20, 2006 Post Count: 13 Status: Offline |
Over on Rosetta@home, mmciastro has posted a benchmark comparison chart (that requires a lot of bandwidth) that shows the points per hour for a lot of different chips and operating systems running BOINC. Linux is usually only awards 55% as many points as Windows. http://boinc.bakerlab.org/rosetta/forum_thread.php?id=2415 Lawrence Thanks for that. The chart pretty much proves Linux users aren't getting as much credit and therefore it needs to be sorted, otherwise it's going to drive Linux users away from BOINC. ![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Bandaid repair for this is simple.
Points awarded = Linux score x 1.82 As the deviation is a constant there is no reason this hasn't already been done. Cheers. ozylynx |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Well, that was disappointing. My first test gave an even worse result than the one you quote. Early days yet, though.
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Here's my latest theory resulting from discussion in another thread, but belongs here.
----------------------------------------My XP3000+ takes a little less than 6 hours, on average to complete a FAAH WU and claims 10.81 PPH for an average of 63.x PPWU My Celeron 1300 takes around 12 Hours to complete a FAAH WU and it claims 6.81 PPH at an average of nearly 80 PPWU The relationship between PPH and PPWU can be seen to be inconsistant, and both must be seen in context. Discussion of one will lead to the other in order to get a true view of what is happening. My observation from this, is that although the fastest machines claim more PPH than slower ones, they claim fewer PPWU due to completing the WU disproportionately faster. The result in a points quorum is that the fastest machines then have their effective PPH amplified by receiving the higher than claimed PPWU from the slower ones. I would like to out out a call for more datum so that this theory can be examined properly. What is needed? For this experiment please supply the type and speed of your CPU. A 'Cut & Paste' from some recently submitted and verified FAAH Work Unit results, BOINC only. These are available from 'My Grid-Device Manager-Results Status' Select the computer you have nominated in the left side drop down box and 'Valid' In the Results Status (right side) drop down window, then click 'Filter. From the reults shown, click on the name of a typical Work Unit and select the information shown in the pop-up window. Right click on your selection and select 'copy' from the dropdown box. You may then Paste the reults to this thread. It would help to underline the result from the nominated machine. Like this: faah0834_ bdb566_ mx1s6s_ 0C Valid 10/15/2006 11:02:38 10/16/2006 02:33:32 6.08 67 / 61 faah0834_ bdb566_ mx1s6s_ 0C Valid 10/15/2006 11:02:34 10/15/2006 19:13:11 5.68 61 / 61 faah0834_ bdb566_ mx1s6s_ 0C Valid 10/15/2006 10:54:55 10/16/2006 01:57:33 4.76 42 / 61 Thanks for you help. This is not intended to be a groan, it is designed to identify problems and perhaps lead to a viable solution to them. The more data collected the sooner theories can be confirmed or debunked. Either result is a step forward. Cheers. ozylynx ![]() New Thread started here [Edit 2 times, last edit by Former Member at Oct 16, 2006 5:48:34 AM] |
||
|
|
olympic
Senior Cruncher Joined: Jun 12, 2005 Post Count: 156 Status: Offline |
These FAAH results are from my Intel E6400 running at 3.4GHz. I'll delete everything except CPU time and point claimed/granted status since that's what is relevent to this topic.
----------------------------------------5.70.....33 / 63 5.47.....63 / 63 3.18.....64 / 63 18.66.....77 / 66 12.96.....60 / 66 3.31...... 66 / 66 3.22.....65 / 59 9.91.....59 / 59 5.21.....53 / 59 6.21.....47 / 52 6.71.....52 / 52 3.35.... 67 / 52 So it appears that on average, my machine is claiming more points per WU than the other machines. I think alot of this has to do with what is happening on the host machine when the benchmarks are run. If the user is playing a game or running any CPU intensive app the benchmarks will be slower than normal, thus affecting claimed credits for the following week. I've also witnessed wildly different benchmarks just by running them several times in a row. What we need is a more accurate and reliable benchmark that reflects a machines actual speed when crunching here on WCG. But there are so many variables that I don't know if that's even possible. ![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
These FAAH results are from my Intel E6400 running at 3.4GHz. I'll delete everything except CPU time and point claimed/granted status since that's what is relevent to this topic. 5.70.....33 / 63 5.47.....63 / 63 3.18.....64 / 63 18.66.....77 / 66 12.96.....60 / 66 3.31...... 66 / 66 3.22.....65 / 59 9.91.....59 / 59 5.21.....53 / 59 6.21.....47 / 52 6.71.....52 / 52 3.35.... 67 / 52 So it appears that on average, my machine is claiming more points per WU than the other machines. I think alot of this has to do with what is happening on the host machine when the benchmarks are run. If the user is playing a game or running any CPU intensive app the benchmarks will be slower than normal, thus affecting claimed credits for the following week. I've also witnessed wildly different benchmarks just by running them several times in a row. What we need is a more accurate and reliable benchmark that reflects a machines actual speed when crunching here on WCG. But there are so many variables that I don't know if that's even possible. Thanks olympic. I copied this to the new thread My response is there. Cheers ![]() |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I agree with the thread stater, here at WCG I get with my X2 4600+ (I don't use optimized clients) about 65-70 credits/6hours while on for instance Einstein I get 110-120 credits/6hours. It would be perfect if there would be a fixed amount of credit/WU like on Einstein and many other projects.
And I don't participate because of the points but to help the science. |
||
|
|
|