VLAR change

Message boards : Technical News : VLAR change

To post messages, you must log in.

1 · 2 · 3 · Next

AuthorMessage
Jeff Cobb
Volunteer moderator
Project administrator
Project developer
Project scientist

Send message
Joined: 1 Mar 99
Posts: 115
Credit: 40,367
RAC: 0
United States
Message 1015702 - Posted: 14 Jul 2010, 23:06:04 UTC

We are beta testing a change whereby VLAR WUs are not scheduled onto GPUs. We hope to move this to the main project next week.


ID: 1015702 · Report as offensive
ClaggyProject Donor
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4623
Credit: 46,348,375
RAC: 2,975
United Kingdom
Message 1015704 - Posted: 14 Jul 2010, 23:13:18 UTC - in response to Message 1015702.  

Thanks for the update Jeff,

Claggy

ID: 1015704 · Report as offensive
Profile Zeus Fab3r
Avatar

Send message
Joined: 17 Jan 01
Posts: 649
Credit: 174,593,099
RAC: 65,757
Serbia
Message 1015711 - Posted: 14 Jul 2010, 23:27:08 UTC

This is indeed a good news, thanks Jeff.



Who the hell is General Failure and why is he reading my harddisk?¿

ID: 1015711 · Report as offensive
Profile soft^spirit
Avatar

Send message
Joined: 18 May 99
Posts: 6438
Credit: 31,836,647
RAC: 6,628
United States
Message 1015721 - Posted: 15 Jul 2010, 0:22:16 UTC - in response to Message 1015702.  

thank you thank you thank you for testing it first!!! :D



Janice

ID: 1015721 · Report as offensive
Cruncher-American

Send message
Joined: 25 Mar 02
Posts: 1310
Credit: 176,037,583
RAC: 108,765
United States
Message 1015729 - Posted: 15 Jul 2010, 1:12:18 UTC - in response to Message 1015702.  

We are beta testing a change whereby VLAR WUs are not scheduled onto GPUs. We hope to move this to the main project next week.


Excellent! Thank you so much!
Good job.

ID: 1015729 · Report as offensive
Profile Dirk Sadowski
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7066
Credit: 101,428,469
RAC: 80,331
Germany
Message 1015746 - Posted: 15 Jul 2010, 2:37:52 UTC - in response to Message 1015702.  
Last modified: 15 Jul 2010, 2:44:47 UTC

We are beta testing a change whereby VLAR WUs are not scheduled onto GPUs. We hope to move this to the main project next week.


Thank you very much! *thumb up*


BTW. Which BOINC version at least is needed for this function? Or is the BOINC version not important?


[EDIT: This means 0.12x+ ARs only to GPUs?]

ID: 1015746 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 1015747 - Posted: 15 Jul 2010, 2:45:59 UTC - in response to Message 1015746.  
Last modified: 15 Jul 2010, 2:47:44 UTC

...
Which BOINC version at least is needed for this function? Or is the BOINC version not important?

It's a BOINC server code change. BOINC core client doesn't matter.

Edit: Yes, the criteria is at 0.12 AR.
                                                                 Joe

ID: 1015747 · Report as offensive
Profile Geek@Play
Volunteer tester
Avatar

Send message
Joined: 31 Jul 01
Posts: 2467
Credit: 86,146,931
RAC: 0
United States
Message 1015754 - Posted: 15 Jul 2010, 3:42:20 UTC
Last modified: 15 Jul 2010, 3:43:25 UTC

Excellent!!


Boinc....Boinc....Boinc....Boinc....

ID: 1015754 · Report as offensive
Wedge009
Volunteer tester
Avatar

Send message
Joined: 3 Apr 99
Posts: 448
Credit: 241,790,461
RAC: 90,529
Australia
Message 1015812 - Posted: 15 Jul 2010, 11:46:56 UTC

Thanks for this.

Just a question about the 0.12 upper AR limit, though. I had a bunch of WUs a short while back that were about 0.18 AR, and they were quite slow on the GPU as well. Well, I suppose even if the limit is hard-coded, it's still better than the current state of allowing VLAR WUs through to the GPU.


Soli Deo Gloria

ID: 1015812 · Report as offensive
PhonAcq

Send message
Joined: 14 Apr 01
Posts: 1629
Credit: 28,527,540
RAC: 3,358
United States
Message 1015815 - Posted: 15 Jul 2010, 12:02:54 UTC - in response to Message 1015721.  

thank you thank you thank you for testing it first!!! :D


Yes, please test the heck out of it.

ID: 1015815 · Report as offensive
PhonAcq

Send message
Joined: 14 Apr 01
Posts: 1629
Credit: 28,527,540
RAC: 3,358
United States
Message 1015816 - Posted: 15 Jul 2010, 12:08:01 UTC

Of course, this change will mean that we non-GPUers' will suffer, won't it? Actually the question is whether we non-GPU hosts will get a fair credit for dealing with the more difficult wu's.

Conversely, will there be enough wu's produced to keep the GPUers' happy, because the VLAR bottleneck will essentially vanish?

And, finally, does this change have any implication for the AP crunchers?

ID: 1015816 · Report as offensive
Profile Bill Walker
Avatar

Send message
Joined: 4 Sep 99
Posts: 3868
Credit: 2,697,159
RAC: 4
Canada
Message 1015820 - Posted: 15 Jul 2010, 12:16:55 UTC - in response to Message 1015816.  

Of course, this change will mean that we non-GPUers' will suffer, won't it?


Doens't it mean that the VLARs will only go to CPUs? That should mean a larger pool of WUs for us non-GPU crunchers to chase.


ID: 1015820 · Report as offensive
Profile Link
Avatar

Send message
Joined: 18 Sep 03
Posts: 805
Credit: 1,678,562
RAC: 22
Germany
Message 1015821 - Posted: 15 Jul 2010, 12:17:29 UTC - in response to Message 1015816.  

Of course, this change will mean that we non-GPUers' will suffer, won't it? Actually the question is whether we non-GPU hosts will get a fair credit for dealing with the more difficult wu's.

Actually after my calculations VLARs are the best paid WUs on CPU. Here example for my AthlonXP 2000+:

- VLAR (< 0.013): about 350 CPU-seconds / 1Cr
- AR ~ 0.38: about 365 CPU-seconds / 1Cr
- AR ~ 0.4x: about 470 CPU-seconds / 1Cr
- VHAR: about 500-700 CPU-seconds / 1Cr
.

ID: 1015821 · Report as offensive
Profile Bill Walker
Avatar

Send message
Joined: 4 Sep 99
Posts: 3868
Credit: 2,697,159
RAC: 4
Canada
Message 1015828 - Posted: 15 Jul 2010, 12:30:38 UTC - in response to Message 1015821.  
Last modified: 15 Jul 2010, 12:30:54 UTC

Woohoo! Send me your VLARS please.



ID: 1015828 · Report as offensive
Profile Fred J. Verster
Volunteer tester
Avatar

Send message
Joined: 21 Apr 04
Posts: 3252
Credit: 31,903,643
RAC: 0
Netherlands
Message 1015862 - Posted: 15 Jul 2010, 14:56:10 UTC - in response to Message 1015828.  

Actually after my calculations VLARs are the best paid WUs on CPU. Here example for my AthlonXP 2000+:

- VLAR (< 0.013): about 350 CPU-seconds / 1Cr
- AR ~ 0.38: about 365 CPU-seconds / 1Cr
- AR ~ 0.4x: about 470 CPU-seconds / 1Cr
- VHAR: about 500-700 CPU-seconds / 1Cr



Are these figures, really all from your AthlonXP 2000+ , on my Q6600's, it's
vice versa.
OTOH, I never crunch VHAR's on CPU.

Ehh, sorry I didn't pay attention to the credits per seconds, not really interrested in credits, only processing time, credits
have a low priority, in my crunching ability/scheme.


And you've little influence on the Angle Range of WU's sended to you, maybe this will change if the new AR-Function-Scheduler, is in place and works as it supposed to.
Then you would be able to ask only work for CPU (<0.013 AR) or GPU (>0.013 AR).

ID: 1015862 · Report as offensive
Profile hiamps
Volunteer tester
Avatar

Send message
Joined: 23 May 99
Posts: 4292
Credit: 72,971,319
RAC: 0
United States
Message 1015873 - Posted: 15 Jul 2010, 15:34:21 UTC - in response to Message 1015816.  

Of course, this change will mean that we non-GPUers' will suffer, won't it? Actually the question is whether we non-GPU hosts will get a fair credit for dealing with the more difficult wu's.

Conversely, will there be enough wu's produced to keep the GPUers' happy, because the VLAR bottleneck will essentially vanish?

And, finally, does this change have any implication for the AP crunchers?

Seems most of us with GPU's also have CPU's in our machines...Why would you do all the hard work?
Official Abuser of Boinc Buttons...
And no good credit hound!

ID: 1015873 · Report as offensive
PhonAcq

Send message
Joined: 14 Apr 01
Posts: 1629
Credit: 28,527,540
RAC: 3,358
United States
Message 1015875 - Posted: 15 Jul 2010, 15:40:17 UTC - in response to Message 1015873.  

Obviously, I'm addressing the relative amount of work for the host in total.

ID: 1015875 · Report as offensive
PhonAcq

Send message
Joined: 14 Apr 01
Posts: 1629
Credit: 28,527,540
RAC: 3,358
United States
Message 1015878 - Posted: 15 Jul 2010, 15:45:14 UTC - in response to Message 1015820.  

Of course, this change will mean that we non-GPUers' will suffer, won't it?


Doens't it mean that the VLARs will only go to CPUs? That should mean a larger pool of WUs for us non-GPU crunchers to chase.


Not understanding this reply. There are cpu only and gpu+cpu crunchers. If the VLAR's are blocked from the gpu's, then the non-VLAR wu's will be sopped up by the gpu's, leaving the cpu's with a strongly biased ratio of VLAR's.

I suppose it's a sh.ty job, but someone has to do it.

ID: 1015878 · Report as offensive
Profile Dirk Sadowski
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7066
Credit: 101,428,469
RAC: 80,331
Germany
Message 1015879 - Posted: 15 Jul 2010, 15:48:27 UTC

BTW.

Nice to know that the admins have an opened ear for suggestions of the members for to make better the project (for to get the max performance of all hosts for accelerated science)..

Who said this won't work? X-D

O.K., now I can switch on again my PC.. :o)


ID: 1015879 · Report as offensive
Profile Dirk Sadowski
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7066
Credit: 101,428,469
RAC: 80,331
Germany
Message 1015885 - Posted: 15 Jul 2010, 15:53:32 UTC

Current I can't post some examples of my E7600, because the new Cr.-System isn't finished balanced.

But, AFAIK, the VLAR WUs get more Cr./hour on CPU, than normal AR WUs.

So CPU only (and of course CPU & GPU) hosts will have increased RAC. This isn't well?


ID: 1015885 · Report as offensive
1 · 2 · 3 · Next

Message boards : Technical News : VLAR change


 
©2016 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.