BOINC 5.* OK, but 6.6.* NOK
15 Oct 2010 8:25:10 UTC
Topic 195384
(moderation:
Hi all,
funny thing, just upgraded from v5 to v6.6* (I know it is an old version) and now I get
Quote:
2010-10-15 08:18:41.7237 [PID=17098] outdated client version 60641 < min core version 61000
2010-10-15 08:18:41.7237 [PID=17098] [version] no app version available: APP#11 (einsteinbinary_ABP2) PLATFORM#1 (i686-pc-linux-gnu) min_version 0
2010-10-15 08:18:41.7252 [PID=17098] Not sending work because client is outdated
2010-10-15 08:18:41.7359 [PID=17098] [debug] [HOST#1967541] MSG(high) No work sent
2010-10-15 08:18:41.7359 [PID=17098] [debug] [HOST#1967541] MSG(high) (your BOINC client is old - please install current version)
Using v5 I got work w/o problems. Can somebody explain this? Just curious.
Michael
BOINC 5.* OK, but 6.6.* NOK
)
It seems that you got this message as a response to a scheduler request that asked ONLY for GPU work. And I suspect the message indeed only refers to CUDA workunits.
The initial CUDA support of BOINC wasn't so great so E@H will send out CUDA tasks only to newer (> 6.10.something I guess) clients. Older clients like the one you used before (5.*) don't even support CUDA at all so they never asked for CUDA tasks, that's the reason you didn't see this error message before.
I assume you will still get CPU tasks with your current BOINC client, tho.
CU
HB
RE: It seems that you got
)
This sounds reasonable. So I will wait till next request for *C*PU work.
Michael
Team Linux Users Everywhere
Hi, no chance. 6.6.*
)
Hi,
no chance. 6.6.* client always requests work for GPU first. It follows "communication deferred", and a request for GPU work again...
Stuck in a loop....
OK, I'am upgrading to 6.10.* already.
Greetings,
Michael
Team Linux Users Everywhere
You really want to have that
)
You really want to have that Nvidia 8200 (125Mb) detected? You can't use it here or at many other projects anyway. If any? perhaps DnetC or Collatz, if they have a Linux app for the GPU?
Else just disable its detection as it's not going to be used anyway.
The next cc_config.xml file in your BOINC Data directory will do the trick:
1
Save, restart BOINC, gone is inadequate GPU detection and you can stay at 6.6, or wait a couple of days and upgarde to 6.12 ... ;-)
RE: Hi, no chance. 6.6.*
)
I just saw the latest scheduler log entry indicating you got a new CPU task yesterday evening with BOINC 6.10.58 :-)
CU
HBE
RE: You really want to have
)
Now that Michael has upgraded to 6.10.58, this is no longer necessary. With BOINC 6.10.58 client, E@H will not try to assign work to a GPU with insufficient video RAM (it's quite possible that this was the reason for excluding 6.6.x clients from getting CUDA work) .
CU
HBE
RE: You really want to have
)
I am using a NVIDIA 8200 at DNETC. WUs take approx 14 hours to complete :-)
Michael