Deprecated: Function get_magic_quotes_gpc() is deprecated in /srv/BOINC/live-webcode/html/inc/util.inc on line 640
general question about Albert

WARNING: This website is obsolete! Please follow this link to get to the new Albert@Home website!

general question about Albert

Message boards : Problems and Bug Reports : general question about Albert
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile Stephan Goll

Send message
Joined: 13 Dec 05
Posts: 19
Credit: 1,874,367
RAC: 0
Message 111426 - Posted: 28 Nov 2011, 22:09:34 UTC

Well, Bernd clearly stated that the primary goal is to test and develop OpenCL components for e@h. With this objective in mind: does it make sense for non-ATI-crunchers to participate in Albert? CUDA and CPU results can be used to validate OpenCL results, but then ...? I'm missing a little information about this. I'm getting WUs for two of my systems but the question remains: does Albert need / want non-ATI-Crunchers?
Stephan
ID: 111426 · Report as offensive     Reply Quote
Profile Bernd Machenschalk
Volunteer moderator
Project administrator
Project developer
Avatar

Send message
Joined: 15 Oct 04
Posts: 1956
Credit: 6,218,130
RAC: 0
Message 111430 - Posted: 29 Nov 2011, 9:29:45 UTC
Last modified: 29 Nov 2011, 9:35:49 UTC

I think I posted one or the other information already somewhere else, but to summarize:

If we wouldn't need non-ATI / OpenCL results, we would have disabled the respective application versions (as we had in the past).

Were it only for developing new versions of the existing BRP4 application we could as well have tested it on the main project. But it's not.

The OpenCL stuff requires not only new application binaries, but (for Einstein@Home) a couple of modifications to the server code, and currently special workunits, too. Apart from validating the results we also need to test how these changes work with the existing application verions.

However, currently we're getting by far the most results back from Windows CUDA apps (twice as many as from all other app versions together). We certainly need a few of these, but probably not that much.

BM
ID: 111430 · Report as offensive     Reply Quote
Profile tullio

Send message
Joined: 22 Jan 05
Posts: 796
Credit: 137,342
RAC: 0
Message 111462 - Posted: 1 Dec 2011, 15:53:01 UTC
Last modified: 1 Dec 2011, 15:54:03 UTC

I have the impression that a binary radio pulsar search on Albert@home is slower that a similar search on Einstein@home on my Opteron 1210 CPU running Linux at 1.8 GHz. But are those tasks identical or is there a difference?. I am glad to say I can see graphics on both.
Tullio
ID: 111462 · Report as offensive     Reply Quote
Profile Bernd Machenschalk
Volunteer moderator
Project administrator
Project developer
Avatar

Send message
Joined: 15 Oct 04
Posts: 1956
Credit: 6,218,130
RAC: 0
Message 111466 - Posted: 1 Dec 2011, 20:48:28 UTC

This is an effect of the "special workunits" I mentioned. Indeed these are a bit different from the BRP4 ones on Einstein.

BM
ID: 111466 · Report as offensive     Reply Quote
dunx

Send message
Joined: 14 Nov 11
Posts: 1
Credit: 43,000
RAC: 0
Message 111481 - Posted: 3 Dec 2011, 15:37:52 UTC

I'm busy trying to buy a decent ATI card, but money is tight this time of year...

Will resume CUDA/CPU work if it's helping.

dunx
ID: 111481 · Report as offensive     Reply Quote
Profile [SETI.USA]Tank_Master
Avatar

Send message
Joined: 22 Jan 05
Posts: 17
Credit: 301,403
RAC: 0
Message 111482 - Posted: 3 Dec 2011, 19:59:12 UTC

wait for the 7000 series, should be announced on monday. This new series will have much reduced power consumptions, and improved performance over the 6000 series. Or, you can see if you can find the new entry level ATI 6930.
ID: 111482 · Report as offensive     Reply Quote

Message boards : Problems and Bug Reports : general question about Albert



This material is based upon work supported by the National Science Foundation (NSF) under Grant PHY-0555655 and by the Max Planck Gesellschaft (MPG). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the investigators and do not necessarily reflect the views of the NSF or the MPG.

Copyright © 2024 Bruce Allen for the LIGO Scientific Collaboration