RPi freezes after a while and some tasks go invalid
28 Dec 2015 2:54:17 UTC
Topic 198355
(moderation:
Hi again,
I have my RPi crunching away on Einstein and WUProp. Since going over to Einstein, my RPi won't go out of black screen and won't respond besides the light blinking from network activity sometimes. This happens 1-2 times a day and I have to hard reboot it.
Another thing is that half my tasks become invalid.
Any idea or help will be greatly appreciated.
Thanks.
RPi freezes after a while and some tasks go invalid
)
Validate errors (result contains obvious garbage) are usually a sign of hardware operating too close to (or beyond) its limits.
Are you overclocking?
How good is the cooling?
How good is the power supply?
Try using a room fan to improve cooling to see if that makes any difference.
Cheers,
Gary.
- OC'ed to 1000 Mhz. -
)
- OC'ed to 1000 Mhz.
- Open case with ambient around 24 C. Core temps never go above 69 C.
- It came with a package. 5 volts; 2000mAH.
It runs all 4 cores 75% of the time.
With the stated facts, it
)
With the stated facts, it seems very likely that you have the clock set higher than the part will perform the required tasks at the existing conditions perfectly correctly. (I am an echo of Gary here, but with the added assurance given by your new details)
If it were my system, as a first triage step I'd drop the clock substantially, and run for long enough to confirm or deny that there was a drastic reduction in the invalid rate.
If the answer is yes, you know that your major available variables are clock rate, cooling, and supply voltage, and can have another try at finding a combination that works and provides performance you can accept.
Returned it to stock clocks.
)
Returned it to stock clocks. I'll report back once a fresh set of wu's run and finish.
What is the stock
)
What is the stock frequency?
As you were getting some valid results at 1000 MHz, a drop to say 950 MHz may very well have been sufficient to cure the problem.
Cheers,
Gary.
I put it back to 900 MHz for
)
I put it back to 900 MHz for safe measures and temps are now 62 C.
If the change from 1000 to
)
If the change from 1000 to 900 MHz does not make a dramatic difference in your invalid rate then there is something more subtle going on. My guess is that it will. The lower temperature will help the suspected problem as well, but not by nearly so much as the 10% clock rate decrease.
I've been having the same
)
I've been having the same problem, but was getting away with it for a while--until we installed a new furnace. Now it's not as cool where my RPi2 is sitting and it freezes up every few days. I was going to restore the clock speed to the recommended setting for Raspberry Pi 2, but am hoping to avoid putting a heat sink on it.
From what I understand, heat
)
From what I understand, heat sinks don't function really well on RPi's.
It looks like frequency may
)
It looks like frequency may have been the problem. The first 4 results returned after the reduction to 900 MHz have all validated.
These latest results show an increase in elapsed time of about an hour or so, which is to be expected. Makes these results rather obvious to spot :-).
If you could arrange for better cooling, you may be able to run without error at 1000 MHz since you did get some valid results at that speed.
Cheers,
Gary.