Thanks for these test info's DM.

I've got it 6.10.6 running on the Vista 32 bit quad. Opposed to the 6.10.4 disaster, this one ran, but took precautionary measures and it still looked at the wrong data-dir location which I told it to move during the install to a new location i.e. not fit for beta public consumption. Stopped client, copied the data from old to new and a backup store, whilst internet disconnected. After restarting client still looked in wrong place, so deleted that and only then did it start crunching with the data in new location. Somewhere a conflict I presume in the registry, which is bad. For safety added the data-dir path to the cc_config.xml.

Only doing CPU crunching with GPU explicitly disabled through said cc_config.xml, but still seeing initial lines pertaining to NVidia and ATI being searched. Anyway, it's been up now for the best part of 48 hours and all continues to validate.

For those interested seeing a report on the ups and downs in the rDCF, add this line to <log_flags> section.

<!-- Next line activates logging of rDCF adjustments after a Task completion or benchmarks - Active since Aug.'09-->
<dcf_debug>1</dcf_debug>

Quickly tells you if a job ran quicker or slower than predicted.

The line above it is the way to insert comments to the cc_config.xml. Found out through a short Q&A last week.

ttyl

edit: yes the log messages that the flag generates in example:

23/09/2009 15.22.39 WCG Computation for task HFCC_t1_02237448_TrkB_0001_0 finished
23/09/2009 15.22.39 WCG [dcf] DCF: 1.642016->1.624209, raw_ratio 1.463945, adj_ratio 0.891553

Only 1 line per finished result, minimal log filler :D

GPUGrid's new app (6.71) seems to be working now on Win7 64-bit & GTX260 cards. 6.10.6 continues to run WCG just fine.

I reinstalled 6.10.6 and reset GPUGrid and it is now requesting work properly.

Win7 64-bit, BOINC 6.10.6, nVidia 190.62, CUDA 2.3

This is a strange interpretation of "upward compatibility".

Users should no longer see repeated requests to our servers for gpus apps. We have the back-off for those set to one week now. You will now only occasionally see that message (when the backoff expires or when you manually force an update).

Hmm, one of the nice things with WIKI's is, you can look on old copies of the page, and the oldest copy of "Client configurations" from 14 April 2008 is informative:

<ncpus>
Act as if there were N CPUs: run N tasks at once. This is for debugging, i.e. to simulate 2 CPUs on a machine that has only 1. Don't use it to limit the number of CPUs used by BOINC; use general preferences instead.

Also worth reading is the info about using zero for disabling <log_flags>, and selected flags under <options>. <ncpus> recides under <options>, and while listed, it does not show zero as an option that will disable the flag...


Well, even v6.8.xx and v6.10.xx includes a bug-fix so <ncpus> finally works as it should, according to the documentation, some users will still be surprised by this change. The wast majority of affected users will be CUDA-users that ran v6.4.x previously, a few developers and alpha-testers, and some users that's been using an option even there isn't any good reason for their usage.

Atleast it's an easily-fixed problem for the few affected users.