Holy Crap!

Message boards : Graphics cards (GPUs) : Holy Crap!

Author	Message
The Gas Giant Send message Joined: 20 Sep 08 Posts: 54 Credit: 607,157 RAC: 0 Level Scientific publications	Message 5521 - Posted: 12 Jan 2009 \| 10:37:59 UTC Last modified: 12 Jan 2009 \| 10:39:33 UTC
	I just completed a Windows 6.55 wu that had a time per step of 181ms! This is up from 136ms for my previous worst case and 103ms typically on my (9600GT). Overall time was 25 hrs for a credit of 2933. This is compared to 23 hrs for 3232 credits. This sucks the big one!
	ID: 5521 \| Rating: 0 \| rate: / Reply Quote

X1900AIW Send message Joined: 12 Sep 08 Posts: 74 Credit: 23,566,124 RAC: 0 Level Scientific publications	Message 5536 - Posted: 12 Jan 2009 \| 17:50:04 UTC - in response to Message 5521.
	You have to compare same WUs, I got meanwhile only four valid 2932-Point-WUs (210428, 208326, 201276, 188467), they need +90-92% more calculating time in relation to the 3232-WUs (52ms vs. 27ms with GTX 260/216, 57ms vs. 30ms with GTX 260/192, see my posting). Using 6.4.5/6.5.0. Do you really think it´s a matter of the BOINC-version ?
	ID: 5536 \| Rating: 0 \| rate: / Reply Quote

The Gas Giant Send message Joined: 20 Sep 08 Posts: 54 Credit: 607,157 RAC: 0 Level Scientific publications	Message 5537 - Posted: 12 Jan 2009 \| 18:20:31 UTC
	What I'm getting at is the large difference in credits per hour. It makes people want to abort the poor paying wu's.....
	ID: 5537 \| Rating: 0 \| rate: / Reply Quote

X1900AIW Send message Joined: 12 Sep 08 Posts: 74 Credit: 23,566,124 RAC: 0 Level Scientific publications	Message 5539 - Posted: 12 Jan 2009 \| 18:54:36 UTC - in response to Message 5537.
	I don´t matter about "selecting" WUs, take them as they come along, analyzing my task-list (1) I see a well-balanced mix of WUs, can´t complain about clearing the "rubbish." Up to now, I think the problem is observed by the admins, the project deserves sympathy and participation, some troubles you can find in any other project. Perhaps aborting WUs should be faced with reducing the limit of max WUs per day, could be a efficient method to slow down point-chasers.
	ID: 5539 \| Rating: 0 \| rate: / Reply Quote

Jayargh Send message Joined: 21 Dec 07 Posts: 47 Credit: 5,252,135 RAC: 0 Level Scientific publications	Message 5541 - Posted: 12 Jan 2009 \| 19:30:31 UTC - in response to Message 5539. Last modified: 12 Jan 2009 \| 19:32:31 UTC
	I don´t matter about "selecting" WUs, take them as they come along, analyzing my task-list (1) I see a well-balanced mix of WUs, can´t complain about clearing the "rubbish." Up to now, I think the problem is observed by the admins, the project deserves sympathy and participation, some troubles you can find in any other project. Perhaps aborting WUs should be faced with reducing the limit of max WUs per day, could be a efficient method to slow down point-chasers. Would also be an efficient way to stop users with 2 or 3 cards in their host from getting enough work to use all they have....yes that would be a smart move! Especially now with a GTX295 if you have 2 of those you will run out of work with the current 15 per day limit! They need to up the limit not lower it!
	ID: 5541 \| Rating: 0 \| rate: / Reply Quote

localizer Send message Joined: 17 Apr 08 Posts: 113 Credit: 1,656,514,857 RAC: 0 Level Scientific publications	Message 5543 - Posted: 12 Jan 2009 \| 19:39:51 UTC - in response to Message 5541. Last modified: 12 Jan 2009 \| 19:41:26 UTC
	............. and if the project needs the science, the science needs the point chasers. Limit is already too low for some hosts. How do you think your proposal would fit with the 4 GPU machine the project team has built and the GTX295 based machine is considering building? There will always be aborts.
	ID: 5543 \| Rating: 0 \| rate: / Reply Quote

Jayargh Send message Joined: 21 Dec 07 Posts: 47 Credit: 5,252,135 RAC: 0 Level Scientific publications	Message 5545 - Posted: 12 Jan 2009 \| 19:51:37 UTC
	and if you are talking about reducing by 1 the task limit for every abort it will do no good.....there is an easy workaround for that...if you set your host to nnw then download and abort while the d/l is in progress it becomes a d/l error and not an abort.....why try to con volute the problem by punishing people for wanting equal credit for equal time? Seems the time you would have admin spending doing all this could be time spent actually solving the problem ;)
	ID: 5545 \| Rating: 0 \| rate: / Reply Quote

ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications	Message 5553 - Posted: 12 Jan 2009 \| 22:02:49 UTC
	The last time someone posted that the daily quota was not enough for him GDF raised it almost immediately.. so I really think they're not going to reduce it now. And they're aware of the credit issue and are trying to correct it, see here. And now that the scheduler issue is finally resolved they have more time to deal with other problems. And the first post in this thread is actually very valuable, especially if the wall clock time confirms the long run time. They need such feedback to better adjust the credits. However, I wouldn't call it "sucks the big one!" .. compared to the 1888-credit WUs these ones are almost a good fit ;) MrS ____________ Scanning for our furry friends since Jan 2002
	ID: 5553 \| Rating: 0 \| rate: / Reply Quote

JAMC Send message Joined: 16 Nov 08 Posts: 28 Credit: 12,688,454 RAC: 0 Level Scientific publications	Message 5613 - Posted: 14 Jan 2009 \| 16:01:29 UTC Last modified: 14 Jan 2009 \| 16:05:12 UTC
	damn... http://www.gpugrid.net/workunit.php?wuid=161268 A 5 hour WU on a GTX260 for 2933 credits- almost twice the 'normal' WU CPU time-WTF??! And my wingman aborted- rightfully so! And I noticed one of the uploading files was 12.78MB... I had not noticed any files this big uploading before
	ID: 5613 \| Rating: 0 \| rate: / Reply Quote

mike047 Send message Joined: 21 Dec 08 Posts: 47 Credit: 7,330,049 RAC: 0 Level Scientific publications	Message 5617 - Posted: 14 Jan 2009 \| 16:53:57 UTC - in response to Message 5613.
	damn... http://www.gpugrid.net/workunit.php?wuid=161268 A 5 hour WU on a GTX260 for 2933 credits- almost twice the 'normal' WU CPU time-WTF??! And my wingman aborted- rightfully so! And I noticed one of the uploading files was 12.78MB... I had not noticed any files this big uploading before I have seen files this big uploaded on a consistent basis. Can't say about the work time though, I don't have any "big" cards:(. ____________ mike
	ID: 5617 \| Rating: 0 \| rate: / Reply Quote

ExtraTerrestrial Apes Volunteer moderator Volunteer tester Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level Scientific publications	Message 5622 - Posted: 14 Jan 2009 \| 19:24:50 UTC - in response to Message 5613.
	damn... http://www.gpugrid.net/workunit.php?wuid=161268 This WU: 26160 s, 2932 credits -> 404 cr/h Old WU: 23800 s, 3232 credits -> 489 cr/h Old WU: 17200 s, 1888 credits -> 395 cr/h Seems like the new one is based on the same flop estimate as the 1888-credit WUs. MrS ____________ Scanning for our furry friends since Jan 2002
	ID: 5622 \| Rating: 0 \| rate: / Reply Quote

Post to thread

Message boards : Graphics cards (GPUs) : Holy Crap!

	About	Science	Volunteers	Performance	Forum	Join us	Donate