Catalyst 13.8 Beta

Message boards : Number crunching : Catalyst 13.8 Beta
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile cov_route
Avatar

Send message
Joined: 13 Sep 12
Posts: 342
Credit: 10,270,618
RAC: 0
Canada
Message 1400180 - Posted: 7 Aug 2013, 15:46:42 UTC

I ran some benchmarks on Cat 13.8 beta.



Similar/slightly improved vis-a-vis 13.4.

One thing that might have improved is consistency in run times. The run-time/credit ratio is qualitatively more consistent (but that might be a coincidence of the wu's getting sent right now).

Here is a chart of the elapsed vs credit for the MB GPU wu's on my acct page right now:



The points outside the two angle-range clusters are 30-spikes and -9 overflows.

Better consistency might explain why the APR on my account page is finally starting to calculate a value for MB v7. My understanding is that if the observed run times vary too much the APR algorithm declines to calculate a value.
ID: 1400180 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1400205 - Posted: 7 Aug 2013, 16:31:10 UTC - in response to Message 1400180.  
Last modified: 7 Aug 2013, 16:33:57 UTC

Better consistency might explain why the APR on my account page is finally starting to calculate a value for MB v7. My understanding is that if the observed run times vary too much the APR algorithm declines to calculate a value.

No, I reported to the boinc_dev list last night that the AMD APR problem was still there, and Eric has diagnosed, and Fixed it:

>> Turns out this is a known validator bug, that I thought I had checked in
>> a fix for about six months ago. If a host claims an unreasonable amount
>> of credit, its APR is never updated. Because its APR is never updated, it
>> never claims a reasonable amount of credit. Because it never claims a
>> reasonable amount of credit, its APR is never updated. etc.

http://lists.ssl.berkeley.edu/pipermail/boinc_dev/2013-August/020262.html

Claggy
ID: 1400205 · Report as offensive
Profile cov_route
Avatar

Send message
Joined: 13 Sep 12
Posts: 342
Credit: 10,270,618
RAC: 0
Canada
Message 1400217 - Posted: 7 Aug 2013, 16:45:12 UTC - in response to Message 1400205.  

Thanks for staying with that issue Claggy.
ID: 1400217 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1400229 - Posted: 7 Aug 2013, 16:58:07 UTC - in response to Message 1400217.  

Thanks for staying with that issue Claggy.

Doesn't help the hosts's like Guy's that have a too High APR already, so all tasks end in Maximum Time Exceeded, he'd have to run Fred's reschedule program and use it to protect against -177 errors, and hope what it does is enough:

http://setiathome.berkeley.edu/forum_thread.php?id=72473&postid=1400201

Claggy
ID: 1400229 · Report as offensive
Profile cov_route
Avatar

Send message
Joined: 13 Sep 12
Posts: 342
Credit: 10,270,618
RAC: 0
Canada
Message 1400266 - Posted: 7 Aug 2013, 17:55:34 UTC

Happy that APR is finally working. On the down side, now I get to see what a wretched number it is for V7...in the low 30's. Eeeesh that looks like a CPU value.
ID: 1400266 · Report as offensive

Message boards : Number crunching : Catalyst 13.8 Beta


 
©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.