With just BOINC, it frequently reaches 100 C, at which some program I have not identified automatically forces it into sleep mode.
100°C is not a healthy temperature for a CPU, even a laptop CPU. For a lot of laptops, that's actually the temperature where the CPU itself shuts down to prevent damage to the silicon. I'd highly recommend checking your laptop's heatsink and fan for dust (if you can do this without having to take it apart too much). You can also try to use ThrottleStop to lower the CPU voltage, but if you do be sure to do some stress testing to ensure it's still functioning properly. I lowered my old laptop's CPU temperature by ~25°C this way without affecting stability.
That's not something I can do myself, but since I've recently bought a new desktop and an installation service for it, I'll ask if the installer can handle it.
Belated thanks for your reply and thanks too for the new app. I've now completed my first task with the Mac version and have had an inconclusive validation against a Linux machine also running v0.23. That machine's result is marked "Validate error".
My result's now been validated another Linux one (non-GPU this time, if it's relevant), so I've no reason to suspect a continuation of the validation problem at this stage. Apologies for the noise.
If one result gets a "validation error", the other result of the quorum is always marked "inconclusive" until a third result comes in. This does not indicate a problem.
With app version 23 the cross-platform validation issues surely were reduced (Windows-Linux from ~0.5% to ~0.02%). But although validating better now, the Mac OS App still sticks out ("invalids" down from ~30% w. version 22 to ~10% with 23).
Holy quantization, Batman! Those checkpoints sure are far apart!
I have Boinc set up to suspend itself when I'm using the machine, which I do intermittently. When it resumes, does it start again from the last checkpoint, or pick up where it left off? It looks to me like it's more likely the former, in which case greater gaps between checkpoints means more work lost.
Holy quantization, Batman! Those checkpoints sure are far apart!
I have Boinc set up to suspend itself when I'm using the machine, which I do intermittently. When it resumes, does it start again from the last checkpoint, or pick up where it left off? It looks to me like it's more likely the former, in which case greater gaps between checkpoints means more work lost.
It depends whether you have "Leave applications in memory when suspended?" active (Computing preferences, top section, around the middle).
As noted, the major impact is on the amount of swap file space consumed - I've never had a problem with it, but if your other computing activities are memory-intensive, you may want to experiment with caution.
With applications in memory, they resume from where they left off: with applications not in memory, and hence after any computer reboot, they restart from the last checkpoint.
I would be happy if you can reduce the GPU memory to run 4 in parallel effective, only can run 2 with 1GB only. Also with some cpu usage and much faster in calculation against running with cpu the credits must be increase by some factor x.
Holy quantization, Batman! Those checkpoints sure are far apart!
I have Boinc set up to suspend itself when I'm using the machine, which I do intermittently. When it resumes, does it start again from the last checkpoint, or pick up where it left off? It looks to me like it's more likely the former, in which case greater gaps between checkpoints means more work lost.
It depends whether you have "Leave applications in memory when suspended?" active (Computing preferences, top section, around the middle).
As noted, the major impact is on the amount of swap file space consumed - I've never had a problem with it, but if your other computing activities are memory-intensive, you may want to experiment with caution.
With applications in memory, they resume from where they left off: with applications not in memory, and hence after any computer reboot, they restart from the last checkpoint.
GPU tasks are always resumed from the last checkpoint (I know Darren doesn't have any GPU's, and that Fermi LAT gamma-ray pulsar search is CPU only):
Change Log for Boinc 6.6.37:
- client: when suspending a GPU job, always remove it from memory, even if it hasn't checkpointed. Otherwise we'll typically run another GPU job right away, and it will bomb out or revert to CPU mode because it can't allocate video RAM
I would be happy if you can reduce the GPU memory to run 4 in parallel effective, only can run 2 with 1GB only. Also with some cpu usage and much faster in calculation against running with cpu the credits must be increase by some factor x.
I don't see a way to reduce the memory consumption of the app significantly without sacrificing speed too much :-(
I haven't looked at the code of course, but since the only thing that matters here is GPU memory consumption, would it be possible to say, do calculations first on one half of an array, then the second half, instead of uploading it all at once?
RE: With just BOINC, it
)
100°C is not a healthy temperature for a CPU, even a laptop CPU. For a lot of laptops, that's actually the temperature where the CPU itself shuts down to prevent damage to the silicon. I'd highly recommend checking your laptop's heatsink and fan for dust (if you can do this without having to take it apart too much). You can also try to use ThrottleStop to lower the CPU voltage, but if you do be sure to do some stress testing to ensure it's still functioning properly. I lowered my old laptop's CPU temperature by ~25°C this way without affecting stability.
That's not something I can do
)
That's not something I can do myself, but since I've recently bought a new desktop and an installation service for it, I'll ask if the installer can handle it.
I wrote:Belated thanks for
)
My result's now been validated another Linux one (non-GPU this time, if it's relevant), so I've no reason to suspect a continuation of the validation problem at this stage. Apologies for the noise.
NG
NG
If one result gets a
)
If one result gets a "validation error", the other result of the quorum is always marked "inconclusive" until a third result comes in. This does not indicate a problem.
With app version 23 the cross-platform validation issues surely were reduced (Windows-Linux from ~0.5% to ~0.02%). But although validating better now, the Mac OS App still sticks out ("invalids" down from ~30% w. version 22 to ~10% with 23).
BM
BM
Holy quantization, Batman!
)
Holy quantization, Batman! Those checkpoints sure are far apart!
I have Boinc set up to suspend itself when I'm using the machine, which I do intermittently. When it resumes, does it start again from the last checkpoint, or pick up where it left off? It looks to me like it's more likely the former, in which case greater gaps between checkpoints means more work lost.
RE: Holy quantization,
)
It depends whether you have "Leave applications in memory when suspended?" active (Computing preferences, top section, around the middle).
As noted, the major impact is on the amount of swap file space consumed - I've never had a problem with it, but if your other computing activities are memory-intensive, you may want to experiment with caution.
With applications in memory, they resume from where they left off: with applications not in memory, and hence after any computer reboot, they restart from the last checkpoint.
I would be happy if you can
)
I would be happy if you can reduce the GPU memory to run 4 in parallel effective, only can run 2 with 1GB only. Also with some cpu usage and much faster in calculation against running with cpu the credits must be increase by some factor x.
RE: RE: Holy
)
GPU tasks are always resumed from the last checkpoint (I know Darren doesn't have any GPU's, and that Fermi LAT gamma-ray pulsar search is CPU only):
Change Log for Boinc 6.6.37:
- client: when suspending a GPU job, always remove it from memory, even if it hasn't checkpointed. Otherwise we'll typically run another GPU job right away, and it will bomb out or revert to CPU mode because it can't allocate video RAM
Claggy
RE: I would be happy if you
)
I don't see a way to reduce the memory consumption of the app significantly without sacrificing speed too much :-(
HB
I haven't looked at the code
)
I haven't looked at the code of course, but since the only thing that matters here is GPU memory consumption, would it be possible to say, do calculations first on one half of an array, then the second half, instead of uploading it all at once?