I forwarded the info on the "resource limit exceeded" problem to the devs.
The ABP1 apps are configured to require at most 1.1 Peta Floating Point Operations .... which is ok if BOINC is using this figure against the CPU benchmarks. If BOINC (maybe only in newer versions) figures these are CUDA floating point ops, then this could be a problem...some cards have 800GFlops and more throughput.
In the meantime until the devs have responded: if you are courageous, you can try the following (experts only, and only if you get this nasty -177 error!!!:
* Stop BOINC
* edit the client_state.xml file :
**** find tags 1100000000000000.000000
**** add two more zeros after the 11
* save client_state.xml
* restart BOINC.
Thanks for testing! The way BOINC works here it will be a bit difficult to set reasonable resource limits for WUs that can be processed by either COU or GPU tasks, but at least the early aborts can be fixed now.
what is the hidden point of use expensive graphic cards and get boring 20% of perfomance increase?
for example milkyway@home get up to 60x (6000%) !! performance increase by using GPU.
what is the hidden point of use expensive graphic cards and get boring 20% of perfomance increase?
for example milkyway@home get up to 60x (6000%) !! performance increase by using GPU.
"Maximum elapsed time
)
"Maximum elapsed time exceeded"
Yep, I've just had two in a row do the same thing after five hours, don't think my 8800gtx is up to it possibly!
:O(
how much % improvement to CPU
)
how much % improvement to CPU only.
seems a bit slow on improved clients.
3.10 was from 19 aug. gettign almost 2 months old ?!!!
and no questions from the developer about bugs that get reported.
Hi! I forwarded the info
)
Hi!
I forwarded the info on the "resource limit exceeded" problem to the devs.
The ABP1 apps are configured to require at most 1.1 Peta Floating Point Operations .... which is ok if BOINC is using this figure against the CPU benchmarks. If BOINC (maybe only in newer versions) figures these are CUDA floating point ops, then this could be a problem...some cards have 800GFlops and more throughput.
In the meantime until the devs have responded: if you are courageous, you can try the following (experts only, and only if you get this nasty -177 error!!!:
* Stop BOINC
* edit the client_state.xml file :
**** find tags 1100000000000000.000000
**** add two more zeros after the 11
* save client_state.xml
* restart BOINC.
Please report if this helped.
Thx
Bikeman
@Bikeman Ok It works
)
@Bikeman
Ok It works !!!!!!!!!!
RE: @Bikeman Ok It works
)
Thanks for testing! The way BOINC works here it will be a bit difficult to set reasonable resource limits for WUs that can be processed by either COU or GPU tasks, but at least the early aborts can be fixed now.
CU
Bikeman
what is the hidden point of
)
what is the hidden point of use expensive graphic cards and get boring 20% of perfomance increase?
for example milkyway@home get up to 60x (6000%) !! performance increase by using GPU.
RE: what is the hidden
)
It's not "hidden", it's public...
Oliver
Einstein@Home Project
RE: It's not "hidden",
)
i mean why so small advantage of using gpu?
RE: i mean why so small
)
Well, I tried to explain it: this is a first test using a trivial approach. We will tune the application for maximum performance as we go...
Oliver
Einstein@Home Project
RE: We will tune the
)
ok. i will wait for "fermi" and change my current nvidia to ati )