I immediately thought it was a stress issue on the PC so ran memory tests and checked fans etc. Nothing helped there. I noticed that at the point the PC hung the PC produced a prolonged a single long tone from the soundcard. A little bit like a scream! Nothing in the event log. The first work unit ran to 43% before problems, the second work unit got to 38%. I don't know if there is a change in the type of calculation around those points?

Having checked the work unit it looks like the other users returned valid results, so this is probably PC related.

Work Unit:
BETA_E200367_974_A.24.C19H12N2S3.192.0.set1d06

For reference:
Shuttle K45 KPC
Intel 945 Chipset
C2D E4300
2 GB RAM
250Gb WD 7200rpm SATA Drive
Soundblaster XiFi XtremeMusic Card

I think it is a bit more of a problem than that - I have been running up to 4 betas at a time and getting crashes running W7 64 on an OC i7 920 with 12Gb ram with constant stability when running any other WU's (except CEP obviously) concurrently. Add beta CEP into the mix and off it goes - but I had the system running for several hours before a crash, but when it came it was very messy. The beta crashes happened when there were other threads crunching other WU types (C4CW, HFCC etc) and these were variously damaged/reset to zero or some just restarted from where they last saved - which WU's reacted which way to the crash seemed somewhat arbitrary - not a good outlook for wider usage at the moment I would have thought?

Thanks for the advice, but I know this isn't a BIOS beep. It's being emitted from an add in PCI Soundblaster Xtreme Music soundcard. It's slightly better than the average card, and I occassionally use this Win XP machine for music composition (requiring ASIO drivers).

Is it possible that a memory address gets touched by the client which is reserved for the soundcard? Or what about interrupts these days... are they important.

Anyway I aborted the BETA's giving the errors and am now back to crunching other projects on this machine. Consider the issue closed and I'll get back to crunching.

Thanks for all the comments.

CEP2 has a huge amount of disk IO and can cause problems on some machines if too many WU's are run simultaneously. My i7 920/Ubuntu machine runs 8 simultaneous CEP2's with no problem at all other than a 4-5% loss in CPU efficiency. But my 980x/Win7 rig crashed hard during the last BETA test (in Sept) whenever I ran more than 4 CEP2's at a time. This time around, I moved the BOINC folders onto a spare 40GB OCZ Vertex 2 SSD I had and I could crunch 12 WU's at a time with no problems at 99% CPU efficiency.

My wifes machine has the OS (Win7) on an SSD and 2 mechanical drives for storage and backup, one of which BOINC is installed on. I never pushed it to the max but it could run 5 or 6 CEP2's at a time without issue.

If you want to run this project and have a multi-core machine with more than 1 hard drive, my suggestion is to move your BOINC folders to a drive that does not contain your OS. Also, you must stagger the startup of the WU's. If you slam the system with 4,8,12, etc WU's at the same instant it will not like it and likely freeze.