wow, these WU use a lot of memory or at least a lot of virtual memory. I had 8 WU crap out because of out of virtual memory.

CPU times for this B.pe type is odd. Seems longer than type A.

TypeA:
ts05_ e143_ ps0000_ 0-- Valid 23/07/10 19:47:54 25/07/10 23:45:06 10.95 203.0 / 181.5
ts05_ e130_ ps0000_ 1-- Valid 23/07/10 19:47:37 24/07/10 13:12:26 11.06 207.0 / 205.9
ts05_ b133_ ps0000_ 1-- Valid 22/07/10 23:09:28 23/07/10 23:54:55 11.00 205.8 / 187.3
ts05_ a328_ ps0000_ 0-- Valid 22/07/10 23:00:49 25/07/10 11:55:10 11.05 206.8 / 139.9

TypeB.pe:
ts05_ e386_ pe0000_ 0-- Valid 30/07/10 22:07:28 31/07/10 16:48:05 15.99 298.6 / 194.1
ts05_ e381_ pe0000_ 0-- Pending Validation 30/07/10 22:07:10 31/07/10 14:09:26 15.94 297.6 / 0.0
ts05_ e385_ pe0000_ 1-- Pending Validation 30/07/10 22:07:27 31/07/10 16:18:17 16.12 301.0 / 0.0

this runs on stock q6600 w/ 4gb ram and win7x64

At 16h per task on a Q6600, a mere 21 tasks would get a cruncher to Brown. However if 4 ran at the same time you would need 1540MB free. 2GB on XP SP2 or Linux would probably suffice, but on Vista or W7 at least 3GB would be needed - unless you like drive thrashing and failures. Some people might want to mix and match (crunch more than one task type and keep an eye on the system so it does not just run DDDT2 WU’s).
During the last scattering of showers I picked up some tasks. It took me to alter two profiles and leave it overnight – I really just wanted to see if this approach would be successful, as it beats pinging the server to death for requests. Well it works, so it might be an idea to create a profile for this project and group computers into that profile (the ones with sufficient resources):
Unfortunately some Pe tasks ended up on some of my lesser systems, which means they will take longer to complete. On an E4400 one task took about 19h 25min. Fine for badges, but perhaps not for the project. Whoever has this PV task is in for a very long wait,
ts05_ e092_ pe0000_ 0--
I guess some people were not happy picking them up at all,
ts05_ f046_ pe0000_ 0-- 617 User Aborted 30/07/10 23:55:37 01/08/10 00:15:04 0.00 0.0 / 0.0
ts05_ f046_ pe0000_ 2-- - In Progress 01/08/10 00:15:12 03/08/10 19:27:12 0.00 0.0 / 0.0
ts05_ f046_ pe0000_ 1-- 617 Pending Validation 30/07/10 23:55:45 01/08/10 12:10:40 15.97 253.5 / 0.0 (my task on a Q6600, and 3GB RAM fortunately)
Also came across one wingman with a pe error,
ts05_ f047_ pe0000_ 1-- 617 Error 30/07/10 23:55:48 01/08/10 00:46:48 0.00 0.0 / 0.0 (my task is in PV)

I personally was one of the people who had type B error out on them. This is because when I checked task manager it says the WU was using around 35k memory per WU. So I figure it was fine to let it all run together and not have to monitor them. What I didn't know was that it was also using a TON of virtual memory. Since a of my machines are 16 cores (8*2 for hyperthreading), that means when they run 16 DDDT, I was screwed. I had 1 machine that just shut down (happened 2x -- lost a lot of runtime) and I had another machine that error out on 8 WU.
Now I have to monitor them to make sure that no more then 7 DDDT jobs are running at once and balance them out with HCC

I could be wrong but I guess that you will need something like 16 GB of RAM running under a Windows 64 or any Windows Server (and I don't know about Linux enough to post an opinion) to recognize it amount of RAM and use effectivelly it with comfortable gaps, and in this way avoid any risc to error out DDDT2 WU's.

Additionally, consider that it is just to run WCG WU's...if you have other applications running at same time, probably will need something more. It is case by case experience when you have hard applications running at sametime using Windows.