ATI host is 'checkpointing' for hours, what could cause such behavior?
10 Oct 2012 20:05:06 UTC
Topic 196563
(moderation:
Looking at the results of my validated work, I've come
cross this BRP4 result.
The number and times of the checkpoints are a bit overdone.....?
ATI host is 'checkpointing' for hours, what could cause such beh
)
There's a typo in your address (ede, must be edu).
Correct link to results makes that http://einsteinathome.org/workunit/134033594
Stale ocz uses both his CPUs and GPU here at Einstein. If he didn't preserve one CPU core for his GPU, then that will account for the long run time.
RE: There's a typo in your
)
EDITTED
Sorry for the typo :-)
I run Einstein on GPU and LHC/SETI on CPU, but if 1 E@h WU runs at 1 GPU and I'm using 2 (ATI HD5870) GPUs, 2 x 0.5= 1CPU core is freed.
Doing 2 per GPU, (4x 0.5CPU)=2 CPU-cores freed.
But only on ATI GPUs.
(I've to take my GTX470 apart, probably full of dust, it stopped, thermal shutdown)!
Hope to get this cleaned, whithout missing deadlines, otherwise I apollogize
to my wingmen!
After cleaning all parts of
)
After cleaning all parts of this host and reseat the CPU cooler and fan,
the GPU and PSU, it now works OK. Eight hours delay ;-)
(About ~50 grams of fatty dust blown out of radiators/coolers blades).
It's now doing 4 Gravitanionel Wave S6 WUs on the CPU (Q6600)
and 2 BRP4CUDA on a GTX470.
RE: After cleaning all
)
Ah, the Dreaded Dust Bunnies. They rove over the globe looking for warm nests to leave their fuzz.
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
RE: RE: After cleaning
)
Yep, from time to time, you've to clean out dust, but also CPU-cooler + fan.
I had to take the GTX470 apart, which was easy and only the fan was dirty.
The PSU was the worst and took it out, removed 2 fans and cleaned it.
Compressed air works very good, but can freeze a component!
I was looking at my results and found a partissipant with ~10 hosts and
quite a lot of errors and aborted work. Here is the url:
Owner.
It's not only counter productive, also uses resources which is a waste.
I hope he/she visits the boards and checks her/his hosts...