Message boards :
Number crunching :
credit comparison to other projects
Message board moderation
Author | Message |
---|---|
Send message Joined: 30 Aug 07 Posts: 2046 Credit: 26,480 RAC: 0 |
And now on to everybody's favorite subject :) Since we've moved over to the new app, I'm wondering how everyones credit rate compares to seti@home (which we'll be trying to use as a benchmark). update: comparison to other projects (like einstein@home) is also very welcome. we want to keep credit generation here similar to other boinc projects. |
Send message Joined: 6 Apr 08 Posts: 2018 Credit: 100,142,856 RAC: 0 |
|
Send message Joined: 7 Jun 08 Posts: 464 Credit: 56,639,936 RAC: 0 |
Here's my two cents, originally posted in a different thread. <edit to fix link> Sorry, wrong post! :-O <edit2> I'd make it a little higher than SAH, to be a burr under the saddle, but that's just me! :-D As I also said before, you're paying about what the Coop apps do rate wise, which is in the 1.5 to 2.5 times better than what the stocker can do ballpark. Alinator |
Send message Joined: 22 Mar 08 Posts: 90 Credit: 501,728 RAC: 0 |
And now on to everybody's favorite subject :) Why use seti as a benchmark? Run your own project! Personally, Ive seen my credits cut in half just since yesterday with the "new" app. and I expect them to go down more with the first full day of work being processed. A clear conscience is usually the sign of a bad memory |
Send message Joined: 28 Aug 07 Posts: 146 Credit: 10,703,601 RAC: 0 |
Well, as far as I'm concerned I'm quite satisfied with the credits, you shouldn't change anything about that! I get a credit level from 40-50 on my AMD X2 5200 under Win 32-Bit. Have not yet checked it under Linux 64-Bit but I'm sure it's not that much different. Leave it like it is and forget about the credit whiners. ;-D Anyway I think, studying the MilkyWay is much more worth than searching for aliens right now. ;-) Member of BOINC@Heidelberg and ATA! My BOINCstats |
Send message Joined: 12 Nov 07 Posts: 2425 Credit: 524,164 RAC: 0 |
Why use seti as a benchmark? Run your own project! Sums it up for me. Doesn't expecting the unexpected make the unexpected the expected? If it makes sense, DON'T do it. |
Send message Joined: 13 Jan 08 Posts: 19 Credit: 820,482 RAC: 0 |
As the WUs. seem to being purge faster them you can reach your task list how can you tell. A longer than ???? seconds would help. I agree why compare with seti claimed and granted rely on lowest returned. But I would like to see some returned credits here before trying to comparing. The only remaining WUs on the task list are errors. Task ID 56895678 Name nm_stripe82_1_25820_1228139976_0 Workunit 57113572 Created 1 Dec 2008 13:59:40 UTC Sent 1 Dec 2008 14:10:57 UTC Received 1 Dec 2008 17:08:26 UTC Server state Over Outcome Client error Client state Compute error Exit status -1 (0xffffffffffffffff) Computer ID 5448 Report deadline 4 Dec 2008 14:10:57 UTC CPU time 2007.233667 stderr out <core_client_version>6.4.1</core_client_version> <![CDATA[ <message> - exit code -1 (0xffffffff) </message> <stderr_txt> Unrecognized XML in parse_init_data_file: computation_deadline Skipping: 1228353057.000000 Skipping: /computation_deadline Unrecognized XML in GLOBAL_PREFS::parse_override: mod_time Skipping: /mod_time Unrecognized XML in GLOBAL_PREFS::parse_override: max_ncpus_pct Skipping: 100.000000 Skipping: /max_ncpus_pct APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 Unrecognized XML in parse_init_data_file: computation_deadline Skipping: 1228353057.000000 Skipping: /computation_deadline Unrecognized XML in GLOBAL_PREFS::parse_override: mod_time Skipping: /mod_time Unrecognized XML in GLOBAL_PREFS::parse_override: max_ncpus_pct Skipping: 100.000000 Skipping: /max_ncpus_pct Error reading into stream_integrals </stderr_txt> ]]> Validate state Invalid Claimed credit 8.19794151280182 Granted credit 0 application version 0.04 Task ID 56547667 Name nm_test27_4775_1228009632_0 Workunit 56776775 Created 30 Nov 2008 2:54:39 UTC Sent 30 Nov 2008 9:36:45 UTC Received 30 Nov 2008 15:41:06 UTC Server state Over Outcome Client error Client state Compute error Exit status -1 (0xffffffffffffffff) Computer ID 5448 Report deadline 3 Dec 2008 9:36:45 UTC CPU time 817.960043 stderr out <core_client_version>6.4.1</core_client_version> <![CDATA[ <message> - exit code -1 (0xffffffff) </message> <stderr_txt> Unrecognized XML in parse_init_data_file: computation_deadline Skipping: 1228163805.000000 Skipping: /computation_deadline Unrecognized XML in GLOBAL_PREFS::parse_override: mod_time Skipping: /mod_time Unrecognized XML in GLOBAL_PREFS::parse_override: max_ncpus_pct Skipping: 100.000000 Skipping: /max_ncpus_pct APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 APP: error writing checkpoint (closing checkpoint file) 0 Unrecognized XML in parse_init_data_file: computation_deadline Skipping: 1228163805.000000 Skipping: /computation_deadline Unrecognized XML in GLOBAL_PREFS::parse_override: mod_time Skipping: /mod_time Unrecognized XML in GLOBAL_PREFS::parse_override: max_ncpus_pct Skipping: 100.000000 Skipping: /max_ncpus_pct Error reading into stream_integrals </stderr_txt> ]]> Validate state Invalid Claimed credit 3.34071149889838 Granted credit 0 application version 0.04 Michael |
Send message Joined: 7 Jun 08 Posts: 464 Credit: 56,639,936 RAC: 0 |
As the WUs. seem to being purge faster them you can reach your task list how can you tell. A longer than ???? seconds would help. Just FYI about SAH: the only time there is a major difference in the granted for a task is if your wingman is running a deprecated or other wise broken CC or host. They could cut them off and eliminate that really annoying and credability busting problem, but won't for god only knows what reason (I don't buy the official line). ;-) AFA your problem goes, trying setting the apps to stay in memory when suspended for now. Alinator |
Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0 |
It's been a while since I've even participated at SETI, so I don't know if my sentiment matches Alinator's or not. However, based on anecdotal observations, I would tend to agree with what was said about the inflationary aspects of "the Coop" apps (optimized apps originally by Simon Zadra, now run by a different group at KSWN). The performance of the stock SETI application is significantly less than those applications. The figure reported at BOINC Combined Statistics is skewed upward vs. everyone running stock. How much of a skew is anyone's guess... Anyway, another problem at SETI is a severe AMD penalty for Astropulse if you run the stock Astropulse application. From what I have been able to determine, this performance penalty was noticed in beta before Astropulse was pushed out to the main project. One of my main issues with Cross-project Parity is that I feel you can't have cross-project values be "equivalent" if you can't even make intra-project values equivalent. With respect to that, the group at Einstein@Home has made a very valiant effort in bringing intra-project parity with their applications. I guess what I'm trying to say is that if you feel you MUST pick a project to align credits with, I'd align them with Einstein, which with the SSE2 application currently yields me about 25 cr/hr... The current 0.04 application here yields about 50-52 cr/hr...which is right about the same as the old non-optimized application. Personally, I think you should run your own project, but if you want a target reduction amount, I'd suggest cutting by a maximum of half the current level, but probably more along the lines of a 30-40% cut from the current level... |
Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0 |
Which "official line"? That they might be sponsors? That they might have an NTLM proxy? That the project does not want to cut off anyone submitting "valid science"? The 3.x clients give 0 credit for anyone who is "lucky" enough to get paired with them. 4.x clients seem to have a significantly higher risk of going into "stand-alone" mode and also asking for 0 credit. Might be enough to offset the optimized apps... Also, don't forget the issues with Astropulse too... |
Send message Joined: 21 Aug 08 Posts: 625 Credit: 558,425 RAC: 0 |
See my post in the windows_x86 thread. Kinda hard to keep apps in memory when you click Start -> Shutdown ;-) |
Send message Joined: 23 Sep 08 Posts: 3 Credit: 174,817 RAC: 0 |
All my credits comparison so far: http://www.editgrid.com/user/hickop/boinc |
Send message Joined: 27 Dec 07 Posts: 35 Credit: 1,432,926 RAC: 0 |
I keep them in my memory good when I do that, but the computers have a hard time doing the same..... :D |
Send message Joined: 4 Oct 08 Posts: 1734 Credit: 64,228,409 RAC: 0 |
And now on to everybody's favorite subject :) I have been running the new fast stock client for 24 hours now, and think the current set up for demand/WU results return/servers keeping up looks to be stable. The shorter WU (2.5 x longer than the Milksop client) gives about 10 CS, while the longer WU (4 x Milksop's client time) gives ~ 40 CS. If the stock clients only are compared, and the output comes from the newer Core 2 CPUs, then the output is about 45 - 50 CS per hour (roughly comparable to SETI and Einstein - I get 44 CS per hour from a Core 2 Quad). I am happy for the current set up to continue and the current credit levels to be where they are. The servers seem to be coping with demand (at least over the last 24 hours) and cross project parity is loosely maintained. You need to leave things as they are and settled for a week or two, and note the total credit score say tomorrow and next week. Then divide the rise in credit by the number of hosts crunching. This should give you the credit per hour actually done. Adjustments can then be made, should they prove necessary. |
Send message Joined: 13 Jan 08 Posts: 19 Credit: 820,482 RAC: 0 |
Just managed to get some data to compare between seti beta 6.03 and milky way 56890136 57109332 1 Dec 2008 12:39:54 UTC 1 Dec 2008 20:20:13 UTC Over Success Done 4,035.92 17.28 39.84 0.0042815516660389700489603361810938 credt per second 56888838 57108372 1 Dec 2008 12:20:41 UTC 1 Dec 2008 20:20:13 UTC Over Success Done 3,985.08 17.07 39.84 0.0042827035977720909227758542827036 credit per second seti beta 4918672 1520395 1 Dec 2008 9:04:01 UTC 1 Dec 2008 16:53:55 UTC Over Success Done 1,987.75 13.25 13.25 0.006665828197710979750974720160986 credits per second. 4918667 1520370 1 Dec 2008 9:04:01 UTC 1 Dec 2008 16:19:21 UTC Over Success Done 1,871.61 13.25 13.25 0.0070794663418126639631119731140569 credits er second. michael |
Send message Joined: 8 Nov 08 Posts: 178 Credit: 6,140,854 RAC: 0 |
On my Linux x86_64 system, 2.53GHz Core 2 Duo, the only project I run on there, ABC@Home, gives about 60 credits/hour. They also have good scale ups for 64-bit systems though, I think the ideal would be a little lower, maybe around 45-50 credits/hour. Last I checked (0.4 app), it was at 108 credits/hour, which was the cap implemented. |
Send message Joined: 22 Mar 08 Posts: 3 Credit: 57,884,314 RAC: 0 |
from first tests: hmm 15-20% better then opt. seti 64bit, but still testing |
Send message Joined: 10 Aug 08 Posts: 218 Credit: 41,846,854 RAC: 0 |
Personal opinion here since I refuse to run seti. Been running einstein for a number of years and I think they would be a favorable project for comparison. 50 credit per hour is a reasonable amount. That's 1/2 of what we were limited to before and while it might not get us down to seti's level it goes a long way to reducing the excess that made DA pay a visit here. Its your project and its how you determine what the value is for the work being done, NOT DA. 50 credits per hour seems to be a happy medium from the low projects and the higher paying projects. |
Send message Joined: 27 Aug 07 Posts: 647 Credit: 27,592,547 RAC: 0 |
... 50 credits per hour seems to be a happy medium from the low projects and the higher paying projects. 100% agreed!! ;) Lovely greetings, Cori |
Send message Joined: 22 Mar 08 Posts: 90 Credit: 501,728 RAC: 0 |
|
©2024 Astroinformatics Group