Deprecated: Function get_magic_quotes_gpc() is deprecated in /disks/centurion/b/carolyn/b/home/boincadm/projects/beta/html/inc/util.inc on line 663
excessive invalids/errors

excessive invalids/errors

Message boards : SETI@home Enhanced : excessive invalids/errors
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile David S
Volunteer tester
Avatar

Send message
Joined: 10 Sep 13
Posts: 1187
Credit: 2,791,507
RAC: 0
United States
Message 48295 - Posted: 5 Dec 2013, 14:44:53 UTC

Is there a thread at Beta for reporting hosts with high numbers of invalids and errors?

I happened to notice http://setiweb.ssl.berkeley.edu/beta/show_host_detail.php?hostid=64599 with a lot. (The same box at Seti main only has one task, an AP that will time out in under 12 hours from now.)

David
signature sent back to alpha testing
ID: 48295 · Report as offensive
Profile David S
Volunteer tester
Avatar

Send message
Joined: 10 Sep 13
Posts: 1187
Credit: 2,791,507
RAC: 0
United States
Message 48296 - Posted: 5 Dec 2013, 14:46:47 UTC - in response to Message 48295.  

Is there a thread at Beta for reporting hosts with high numbers of invalids and errors?

I happened to notice http://setiweb.ssl.berkeley.edu/beta/show_host_detail.php?hostid=64599 with a lot. (The same box at Seti main only has one task, an AP that will time out in under 12 hours from now.)

It could be that this guy's problem fits some other thread. The WU that caught my attention is, for him, an openCL on Intel graphics.

David
signature sent back to alpha testing
ID: 48296 · Report as offensive
Josef W. Segur
Volunteer tester

Send message
Joined: 14 Oct 05
Posts: 1137
Credit: 1,848,733
RAC: 0
United States
Message 48303 - Posted: 5 Dec 2013, 17:17:13 UTC

As we're doing Beta testing, hosts which reveal a problem are perhaps more valuable than those which don't.

host 64599 does seem very prone to creating false Autocorr signals with high peak powers, and is running Intel's OpenCL 1.1 drivers. I don't know if all hosts running those drivers are suffering the same, though, nor whether Raistmer will be able to diagnose and find a workaround. Potentially, a release to the main project might have to require Intel's OpenCL 1.2 drivers.
                                                                   Joe
ID: 48303 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 48310 - Posted: 6 Dec 2013, 8:01:38 UTC - in response to Message 48303.  

I would say that host requires reboot first. And only if it will not help, then driver upgrade. I had such state on own PC once, with OpenCL 1.2 drivers installed.
I'm afraid not much to diagnose here besides some similarity in reported overflows to teach app to recognise wrong apps and throw clean error instead of false overflow. User intervention required (well, maybe worth to reboot such host programmatically from app, but I'm afraid some users could upset ;) )
ID: 48310 · Report as offensive
GonoszTopi
Volunteer tester

Send message
Joined: 21 Jan 14
Posts: 2
Credit: 294,575
RAC: 0
Hungary
Message 48926 - Posted: 23 Jan 2014, 17:06:52 UTC - in response to Message 48310.  

I have the same issue, with OpenCL 1.1 and an Intel(R) HD Graphics 4000 card.

I would say that host requires reboot first.

That does not help at all.

Potentially, a release to the main project might have to require Intel's OpenCL 1.2 drivers.

I agree. As you can see here, all S@H tasks produce lots of autocorr and finally
SETI@Home Informational message -9 result_overflow
NOTE: The number of results detected equals the storage space allocated.

The host mentioned in the above post is now using OpenCL 1.2 drivers on the same GPU and gets valid results.

I also noticed that AP workunits also fail (seems to, at the time of this writing neither of my AP workunits are marked as invalid, but they will be, for sure)
ID: 48926 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 48930 - Posted: 23 Jan 2014, 19:47:57 UTC - in response to Message 48926.  

Well, current Intel drivers support OpenCL 1.2 so no probs with that.
ID: 48930 · Report as offensive
GonoszTopi
Volunteer tester

Send message
Joined: 21 Jan 14
Posts: 2
Credit: 294,575
RAC: 0
Hungary
Message 48956 - Posted: 25 Jan 2014, 0:29:00 UTC - in response to Message 48930.  

I just wanted to suggest that the server should not give out any intel_gpu work units if OpenCL < 1.2
ID: 48956 · Report as offensive

Message boards : SETI@home Enhanced : excessive invalids/errors


 
©2022 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.