New Einstein GPU App: Binary Radio Pulsar Search (Parkes PMPS XT) BRP6

[FourOh]

This new app is for AMD/Nvidia GPU only as of 2/16/15; admin indicates they have no intention of adding a BRP6 CPU app.

Appears to pay about the same as BRPS (Perseus Arm Survey) but with higher credit & longer run time. Work distribution is limited now but this is intended to become the primary GPU app for the project.

At any rate, it may be time to wrap up any hours goals on BRP4 or BRP5... just in case :D

http://einstein.phys.uwm.edu/forum_thread.php?id=11161
http://einstein.phys.uwm.edu/apps.php

[denim]

I saw that, but have yet to get any. Was lucky enough to get a couple of the Gravitational Wave search S6Bucket Follow-up #1.

[FourOh]

I saw that, but have yet to get any. Was lucky enough to get a couple of the Gravitational Wave search S6Bucket Follow-up #1.

I got my 5000 hours on S6Bucket Follow Up a few weeks ago before their upload issues started affecting work distribution. According to the status page, there is plenty of S6Bucket work available with anticipated completion about 40 days out.

[denim]

I got my 5000 hours on S6Bucket Follow Up a few weeks ago before their upload issues started affecting work distribution. According to the status page, there is plenty of S6Bucket work available with anticipated completion about 40 days out.

Some reason, I am only seeing that WU on my deathly slow Linux box with a P4 2.0ghz.

[FourOh]

I'm turning my AMD GPUs over to Einstein for a while, and found from their user stats and message boards that the Parkes XT Beta work runs faster and grants the same credit as the normal app. So if you're running Parkes, select beta versions in your project preferences for 15-30% MOAR CREDITS :D

Another thing... this app seems even more bus-reliant than earlier GPU apps. PCIe 2.0 appears to be a severe limitation on fast cards.

Edit: OK I did some research and my Tahiti cards (7970 and 280x) were getting WAY slower results than they should. I uninstalled AMD Omega drivers and reverted to Catalyst 13.12 + APP SDK v2.9 for the 7970 and Catalyst 14.9 + APP SDK v2.9.1 for the 280x. So far the work units are processing MUCH faster, putting my results in line with users in the top of the stats pages.

[FourOh]

Einstein is issuing BETA Parkes PMPS XT work units for Nvidia cards right now. They pay about 15% better (finish 15% faster, rather) than regular units.

[denim]

They started to come in on my lappy with an Ati card. Long run times though.