Message boards :
Number crunching :
VLAR's to Kepler cards.
Message board moderation
Author | Message |
---|---|
Wiggo Send message Joined: 24 Jan 00 Posts: 34845 Credit: 261,360,520 RAC: 489 |
I've just got to say that I'm really not impressed with these on my GTX 660's even reducing the number of concurrent w/u's from 3 down to 2. Ridiculously long run times, my 2500K can do these quicker. Cheers. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13746 Credit: 208,696,464 RAC: 304 |
I've just got to say that I'm really not impressed with these on my GTX 660's even reducing the number of concurrent w/u's from 3 down to 2. Is that on CUDA 32, 42, 50 or all of the above? Grant Darwin NT |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Vlars takes 4 to 5X to crunch on a 670/690 than a normal WU. That´s why i ask to return to the old way, No Vlars to Nvidias... |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13746 Credit: 208,696,464 RAC: 304 |
Vlars takes 4 to 5X to crunch on a 670/690 than a normal WU. Do you get 4-5X as much credit? Grant Darwin NT |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Vlars takes 4 to 5X to crunch on a 670/690 than a normal WU. Of course not! |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13746 Credit: 208,696,464 RAC: 304 |
Vlars takes 4 to 5X to crunch on a 670/690 than a normal WU. So it's not worth it. Even more so from what i've seen posted about the reduced amount of credit people are getting now for normal run time WUs. Grant Darwin NT |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13746 Credit: 208,696,464 RAC: 304 |
Even more so from what i've seen posted about the reduced amount of credit people are getting now for normal run time WUs. Just had a quick look back though my Valid results. From the looks of it a completed unit is getting about half the credit it used to. But it's taking twice as long (or more) to do. So that works out to about a 75% reduction in credit per hour by my very rough barely concious reckoning. Grant Darwin NT |
Wiggo Send message Joined: 24 Jan 00 Posts: 34845 Credit: 261,360,520 RAC: 489 |
I've just got to say that I'm really not impressed with these on my GTX 660's even reducing the number of concurrent w/u's from 3 down to 2. They are cuda50 but I think that I'll just start aborting them and go back to doing 3 at a time. Cheers. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34845 Credit: 261,360,520 RAC: 489 |
OK, that does it they're all being aborted here now. While running 2 VLAR's on a card my memory controller load drops down to 35-40% but the GPU load jumps to 99% and the power consumption goes up to 90%. Running 3 normal w/u's the memory controller runs between 55-75%, the GPU load between 80-90% and the power consumption goes down to 75-80%. Cheers. |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
IMHO VLARs is still work and science. Better than no work at all. With each crime and every kindness we birth our future. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34845 Credit: 261,360,520 RAC: 489 |
IMHO VLARs is still work and science. Everyone is entitled to their own opinion whether it is good or not but IMO anyone wanting to run them must be a masochist. Cheers. |
atlov Send message Joined: 11 Aug 12 Posts: 35 Credit: 32,718,664 RAC: 34 |
I can confirm Wiggo's observations. Also, the estimated crunch times for VLARs are obscure: 15:33 estimated for cuda32 and my GTX 660 needs about 58mins (1 WU/GPU). GPU load is at 90-99%, while for non-VLAR it is at ~61%. |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
VLAR is sicience to, but why not just put as an option to the user, then we choose if we whant or not be a masochist. Something like: <VLAR on GPU>1|0<VLAR on GPU> will avoid all this worries. |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Maybe but it would be more realistic in comparison against ATI/AMD cards. With each crime and every kindness we birth our future. |
juan BFP Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 |
Compare ATI vs Nvidia is something like compare coke vs pepsi, each one have it´s personal flavor favorite. On other hand try no run Vlars on Nvidias is like try to mix water and oil, even if you could do, they not mix well with the past of the time... |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Here's a VLAR on an AMD; 16fe12ab.28387.14939.6.11.120.vlar_0 Run time 2,382.83 CPU time 130.53 That's about the same runtime for an Astropulse on that card. Offer the same amount of credits for a VLAR as an AP, and I'll bet no one would complain about VLARs again. Just a thought... ;-) |
Mike Send message Joined: 17 Feb 01 Posts: 34258 Credit: 79,922,639 RAC: 80 |
Here's a VLAR on an AMD; 16fe12ab.28387.14939.6.11.120.vlar_0 If you run a high blanked AP task it takes 3 times longer but you claim the same amount of credits. With each crime and every kindness we birth our future. |
Wiggo Send message Joined: 24 Jan 00 Posts: 34845 Credit: 261,360,520 RAC: 489 |
Here's a VLAR on an AMD; 16fe12ab.28387.14939.6.11.120.vlar_0 I'll still complain. ;-) Cheers. |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 |
Here's a VLAR on an AMD; 16fe12ab.28387.14939.6.11.120.vlar_0 The difference being, you would know how long a VLAR would be before you chose to run it. Along with all the associated warts, as in screen lag. Here is what I would do to to a Blanked AP if I knew it before hand; ap_03jn10aa_B6_P0_00018_20130523_23446.wu_1 percent blanked: 93.55 And, you would get about the same credit for that task. |
W-K 666 Send message Joined: 18 May 99 Posts: 19074 Credit: 40,757,560 RAC: 67 |
Here's a VLAR on an AMD; 16fe12ab.28387.14939.6.11.120.vlar_0 I think Eric knows that when the crunchers stop complaining he has a problem. |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.