Hi all.
I'm new to WCG, so please bear with me if I tend to ramble a bit. :)

Here at CAH we've been running WCG for about 4 weeks now, mostly devoted to FAAH, running under BOINC. Apart from a few old Pentium 2 and Pentium 3's, and a scattering of Athlons, CAH is almost exclusively Pentium 4 driven (about 180 devices)

I've noted a "potential" issue with running BOINC and the new HPF2.
When a BOINC client attempts to run a benchmark, which occurs at least once a week, it pauses the running processes, performs the benchmark, sends off the result, and resumes computation. This works with no problems under FAAH.
However, when the BOINC client is processing a HPF2 result, it attempts to pause the process and is unable to do so. It is clearly marked in the messages log in the BOINC client that it was unable to pause the process. This results in a "Computation Error", which is reported, and a new work unit is requested.
Of course, this has the effect of wasting any processing time that was already put into the workunit, as well as creating overhead by having to send the workunit back out to someone else to complete.
I've since noticed this behaviour on at least 6 of the grid PCs here, in the two hours I've been at work today.

Is this a reported issue ? I will probably suspend working on HPF2 while I see whether others have the same issue..

Also, for the benefit of other Pentium 4 users, there is a confirmed 50% throughput benefit, if your PC has enough RAM, of using Hyperthreading when using BOINC.
Example: my 2.8GHz P4 took 7-8 hours to complete a FAAH workunit without hyperthreading on. With Hyperthreading, the PC completes 2 workunits in ~10 hours.

Thanks for listening,
Randall Hose, ICT Manager, Colac Area Health