They took BOINC version 6.10.36 out as stable and put 6.10.18 back.
There is a beta 6.10.37
Yeah noticed that but it seems to be ok here? Slightly panicky about high-priority scheduling but it’s not the end of the world…kinda stuck with it anyway, AQUA@Home needs version > 6.10.18 to fix multicore scheduling issues. Betcha its something to do with the 25% processor usage code, I switched that off here.
there’s a reported issue with a TON of cuda work being requested under 6.10.36, enough to empty servers of all CUDA work in a couple of instances with Einstein@Home. Due to the potential loads and project issues, the 6.10.36 was pulled. Combine this with some local preference problems and we have a release that has been pulled back.
It does highlight in a way that some many of the alpha testers are no longer active, myself guilty of not finding the time 
A video of from the alpha mailing list of one of the issues is here http://www.youtube.com/watch?v=LPO6GyjJb3M&feature=channel
DT.
Aaaaah well that’s why it doesn’t matter to me, my 8800 cards are not good enough for CUDA any more. I should go and buy a 200 series. Except my 8800GTX Ultras cost over £700 for the pair, still eat modern games at high res for breakfast, and therefore CUDA can eat my shorts.
:Pimp: They are still good enough for around 5K PPD each GPU folding , and the team could do with some help.
Not having any problem with it, so I’ll keep 6.10.36 until the next comes out. I have noticed an unusual amount of SETI WUs… perhaps this is why. A bit hard to say though as most of them are very small units.
When I have everything settled in a few days I’ll take a look at that :nod: 2 x 8800GTX, 1 x 8800GTS and a 8600GT (Linux) should help then?
edit try this, reluctant to switch off SLI as my machine has issues if I keep enabling/disabling it, but this looks ok no? Using less than 10% of CPU so BOINC is unflustered too. Unless you switch on the pretty window, which manages to use more than 60% of processor 
Slightly OT I know guys, my apologies.
As far as I know you can fold in SLI mode but can only use one GPU - both my Windows boxes are Quads and have 2 GPUs Folding but I find that I have to limit BOINC to 3 cores or the Folding output is severely curtailed .
GPU folding under Linux is a bit interesting , there is no Linux GPU client - you need to use the Windows client under WINE - things are a lot simpler now, there is an easy to follow Ubuntu based how to on the folding forums ,gives all the essential steps and easily modified for a desktop system , if you have done any SETI CUDA crunching under Linux the set up is very similar - if you’ve got one working the other will as well. 64 bit Linux Quads love the SMP client and the GPU client plays nicely with it.
Yeah currently working fine in SLI mode, but cannot “see” a second GPU which I guess is understandable given that SLI mode is active. Doesn’t bother BOINC (it sees 2 GPUs), however I haven’t crunched BOINC CUDA in a while so that maybe would have issues too.
8600GT has been slapped in a Windows XP box because I am lazy, and is Folding away 
8800GTS (probably the fastest because of the G92 core clock speed) will have to wait until the box is back in running mode…
Thanks for the help 
I myself never had an issue 6.10.36, and ran since beta. Kept it on the 9100e, yet running 6.10.37 on the 9550… Those issues would not affect me either.
Have read around the net to find out more, and it appears the 36 and 37 releases have an old bug in them, one of the symptoms of this bug is the BOINC scheduler does not deal well with variable length units, so runs many units to partway through and then parks them, to start others. Net result lots of parked units, virtual memory use, and scheduling problems.
This has hit the farm with a vengeance leaving me no alternative but to set all projects to NNW and let the caches clear out. I have several projects here, some uniprocessor, some multiprocessor, some quick units, some which take days and days even on four cores simultaneously.
When each machine clears out completely I’ll set it to download Docking only, and restart the other projects whenever the bug is fixed. So output will decrease on everything but Docking, until such a time as the issue is fixed. When will that be? dunno ask Rom Walton at U C Berkeley he is dealing with it.
SETI, SIMAP and Einstein have been stopped here anyway, on the grounds that they always have plenty of volunteers and clear their work. Rosetta has been stopped for the massive amount of data it ships, Milkyway for the crazy credit system, RNA@Home because it cancels workunits willy-nilly (may restart its a Beta project). Artificial Intelligence, Predictor and Malaria control are dead projects, and I really can’t see the point in burning megawatts of electricity to investigate climate change (although I appreciate the irony) so CPDN went on the dead list a long time ago. This leaves me with Aqua, Docking, QMC, Quantumfire, LHC, uFluids, Seti-beta, RALPH, and WCG, most of which have the variable unit length problem, and in the case of the last three I don’t run full time anyway.
Short term Docking@Home win, long term nobody loses
I’ve got a server with over 100units now sat at about 90% complete !!! I did the same and set NNW. Mine only has Docking + SIMAP though, so a touch easier to manage.
DT.
Not sure how to say this…
I LOVE THIS BUG!
Docking docking docking ~smiles~
Sorry your problems make me happy:D
Beta 6.10.43 is now available
Yep, and been running it since it came out… So far no issues detected.