Thanks for your answer : do you mean that "for a given application (in this case O3) GPU usage will vary from one task to another one" ? or your answer is more globally for one application ? can it vary depending on the OS of the application ? (in my case macOS)
Because my experience was (hardly) "no GPU usage" (monitoring in detail) during all execution time of the O3 tasks...
So, as I explained above, I only use GRP which is actually using GPU on my machine - I prefer to let GPU task for projects / applications that actually use the GPU, I feel like it is a waste of resources if I don't :)
More globally, a group of tasks is coded to provide a result, sometimes the next batch of those tasks, they could have a different version number or just be a new batch that's tweaked from the last batch, could be coded to do it slightly differently so they get the expected results from them. Coding is an art and yes alot of people can do it but not alot of people know how to get the most out of your hardware while doing it, the guy that helped Seti way back when is now helping Einstein but I don't know how much he's been able to accomplish so far or what apps he's working on.
I have a friend who used to code for ATM machines and she would have to get up in the middle of the night all the time to remotely fix some lines of code that weren't doing the right thing. They paid her a fortune but when the ATM's went down there could be 250K of them down!! She would often complain about other coders and how sloppy they were just to get a simple job done and how she would have to lock out several thousand lines of code and replace them with a few hundred lines to do the same thing only much faster.
I tried again "All-Sky Gravitational Wave search on O3 v1.07 (GW-opencl-ati-2)" on the same iMac using "AMD Radeon Pro 5700 XT Compute Engine (16368MB)" and the problem is exactly the same : you can see when I switch from the Primegrid task that was using 100% GPU resource, GPU usage suddenly dropped to less than 5% as soon as I switch to the O3 task
it was using approx 30% of one CPU core + 5,4% of the GPU when I took the screenshot. The task has been running more than 20mn now and it is exactly the same.
I will let that task end but there is no point wasting GPU cycles to do nothing, I've read recent advertising for the project on the forum and I saw macOS apps version had been updated end of last year, so i hoped for an improvement of this problem, but there is none :/
I suspect this to be a macOS problem, I observed problems with graphics RAM for a long time on my Mac system. Even with the FGRP tasks it would use up all RAM and slow down, now with that tasks it still worked when using only Boinc, but when I had to do other work I had to do a restart otherwise my whole system was way too slow to be productive. With the gravity wave tasks using much more memory and more intensive this is put into an extreme. My RX580 used to spin up the fans for a few seconds drawing good power, then all memory was used up, fans slowed down, computation slowed down. For some reason macOS can't free graphics memory on those cards. Not even when quitting the Boinc, only a restart works.
There are no of such issues on Windows. Except if I do heavy work with graphic memory intensive programms on windows while running 3 GW tasks the same time on my card with only 8GB graphics memory. Then the same symptoms can be seen: GPU utilisation goes down, power draw goes down, while RAM is used up completely and the system slows down. But unlike on macOS with Windows as soon as I quit a program or snooze Boinc the memory is freed up immediately and computation runs again with max performance.
Except that for me this is not a "general problem for macOS", I have no issue when I run any other GPU boinc tasks (primegrid uses 100% GPU, amicable uses a bit less if I remember well, etc), there is plenty of RAM on the mac (40 GB) and in that GPU (16 GB), and I never experience what you describe with "all the system slowing down" with the general usage with boinc. Sometimes if I watch videos or movies on the mac while crunching a GPU app I can experience some lag (depending on the boinc project app running) so I will suspend it temporarily, but is is rare (and it is not "all the system slowing down", only in case of parallel usage of that GPU boinc + video, which can even be painless with other boinc apps).
It is a a problem with the coding of the "All-Sky Gravitational Wave search on O3 v1.07 (GW-opencl-ati-2)" application for macOS, at least with my version of GPU "AMD Radeon Pro 5700 XT Compute Engine (16368MB)".
Definitely not a Mac OS problem. All of my All-sky GW GPU based tasks exhibit this incredibly annoying behavior. When a task first starts it sits with the GPU idle for a long time, then at the halfway point it sits idle again for a long time, then at 99% it will sit for the last long period of no GPU use.
None of this would bother me much except it really isn't loading up the CPU either. I've tried running multiple tasks on the GPU in an attempt to spread them out but they invariably will bunch together and still leave you with long gaps of low to no GPU usage. I changed the job priorities (Set them all to HIGH) in an attempt to process them a little faster but I didn't see any change.
Odd thing is my old i7-4790 CPU @ 3.60GHz appears to crunch through these pauses faster than my Threadripper 2950X 16-Core Processor
[AF>Le_Pommier wrote:
)
More globally, a group of tasks is coded to provide a result, sometimes the next batch of those tasks, they could have a different version number or just be a new batch that's tweaked from the last batch, could be coded to do it slightly differently so they get the expected results from them. Coding is an art and yes alot of people can do it but not alot of people know how to get the most out of your hardware while doing it, the guy that helped Seti way back when is now helping Einstein but I don't know how much he's been able to accomplish so far or what apps he's working on.
I have a friend who used to code for ATM machines and she would have to get up in the middle of the night all the time to remotely fix some lines of code that weren't doing the right thing. They paid her a fortune but when the ATM's went down there could be 250K of them down!! She would often complain about other coders and how sloppy they were just to get a simple job done and how she would have to lock out several thousand lines of code and replace them with a few hundred lines to do the same thing only much faster.
[AF>Le_Pommier wrote:
)
you kept referring to GRP tasks as S@H.
S@H is a common acronym for Seti@Home. Maybe you meant to say E@H?
_________________________________________________________________________
Not me, but tiger did
)
Not me, but tiger did :D
Anyway, it seems O3 is not for my machine... maybe I'll give it another try in a distant future...
Sorry. My bad. Not enough
)
Sorry. My bad. Not enough coffee and too many term papers to grade.
"I was born in a small town, and I live in a small town." - John Mellencamp
Yep. My bad.
)
Yep. My bad. Sorry.
"I was born in a small town, and I live in a small town." - John Mellencamp
Hi I tried again "All-Sky
)
Hi
I tried again "All-Sky Gravitational Wave search on O3 v1.07 (GW-opencl-ati-2)" on the same iMac using "AMD Radeon Pro 5700 XT Compute Engine (16368MB)" and the problem is exactly the same : you can see when I switch from the Primegrid task that was using 100% GPU resource, GPU usage suddenly dropped to less than 5% as soon as I switch to the O3 task
it was using approx 30% of one CPU core + 5,4% of the GPU when I took the screenshot. The task has been running more than 20mn now and it is exactly the same.
I will let that task end but there is no point wasting GPU cycles to do nothing, I've read recent advertising for the project on the forum and I saw macOS apps version had been updated end of last year, so i hoped for an improvement of this problem, but there is none :/
I see the exact same problem
)
I see the exact same problem with a Vega64.
I suspect this to be an macOS
)
I suspect this to be a macOS problem, I observed problems with graphics RAM for a long time on my Mac system. Even with the FGRP tasks it would use up all RAM and slow down, now with that tasks it still worked when using only Boinc, but when I had to do other work I had to do a restart otherwise my whole system was way too slow to be productive. With the gravity wave tasks using much more memory and more intensive this is put into an extreme. My RX580 used to spin up the fans for a few seconds drawing good power, then all memory was used up, fans slowed down, computation slowed down. For some reason macOS can't free graphics memory on those cards. Not even when quitting the Boinc, only a restart works.
There are no of such issues on Windows. Except if I do heavy work with graphic memory intensive programms on windows while running 3 GW tasks the same time on my card with only 8GB graphics memory. Then the same symptoms can be seen: GPU utilisation goes down, power draw goes down, while RAM is used up completely and the system slows down. But unlike on macOS with Windows as soon as I quit a program or snooze Boinc the memory is freed up immediately and computation runs again with max performance.
Except that for me this is
)
Except that for me this is not a "general problem for macOS", I have no issue when I run any other GPU boinc tasks (primegrid uses 100% GPU, amicable uses a bit less if I remember well, etc), there is plenty of RAM on the mac (40 GB) and in that GPU (16 GB), and I never experience what you describe with "all the system slowing down" with the general usage with boinc. Sometimes if I watch videos or movies on the mac while crunching a GPU app I can experience some lag (depending on the boinc project app running) so I will suspend it temporarily, but is is rare (and it is not "all the system slowing down", only in case of parallel usage of that GPU boinc + video, which can even be painless with other boinc apps).
It is a a problem with the coding of the "All-Sky Gravitational Wave search on O3 v1.07 (GW-opencl-ati-2)" application for macOS, at least with my version of GPU "AMD Radeon Pro 5700 XT Compute Engine (16368MB)".
Definitely not a Mac OS
)
Definitely not a Mac OS problem. All of my All-sky GW GPU based tasks exhibit this incredibly annoying behavior. When a task first starts it sits with the GPU idle for a long time, then at the halfway point it sits idle again for a long time, then at 99% it will sit for the last long period of no GPU use.
None of this would bother me much except it really isn't loading up the CPU either. I've tried running multiple tasks on the GPU in an attempt to spread them out but they invariably will bunch together and still leave you with long gaps of low to no GPU usage. I changed the job priorities (Set them all to HIGH) in an attempt to process them a little faster but I didn't see any change.
Odd thing is my old i7-4790 CPU @ 3.60GHz appears to crunch through these pauses faster than my Threadripper 2950X 16-Core Processor