I don't think the latest change to prevent this behavior made it into .36, I think you need .37 or later for it ...
This is an issue that has been a serious problem for some time... there have been I think 3 fixes tried so far though I have more hopes for the 4th and latest in that it is finally looking at things at the project level and then prioritizing the already running task(s) ...
I don't think the latest change to prevent this behavior made it into .36, I think you need .37 or later for it ...
Nope, they didn't make it into .37 unfortunately. Looks like we are waiting for .38 to come out.
I see the dev version is up to 6.10.43. Has this bug been fixed yet?
Because of the way the 6.10.18 scheduler behaves with settings to try and maintain a 4 day queue my million:1 resource share split between milkeyway and collataz/gpugrid is ignored in favor of trying to keep 4 days of GPU work on hand, vs my desired behavior of keeping the max number of MW work units on hand and only running the other cuda projects if I run out of MW.
Edit: What I'm looking for from the newest client is the .36 feature that allows setting resource share to zero to define a backup only project.
They certainly made another attempt at fixing it, so 6.10.43 is definitely worth checking out. This particular build also fixes a bug that was introduced at some point that made the Options button no longer work. Dunno how long that bug was around for, but..
Edit: What I'm looking for from the newest client is the .36 feature that allows setting resource share to zero to define a backup only project.
That's in 6.10.43, but it needs an update of server software on the project as well to work. So far only Seti has this update, with a slight possibility of DrugDiscovery@Home having it as well, since Jack updated the code yesterday.
Hello all I'm new to the system :)
I also have problems with client errors, seems my intel machine has yet to error, but my AMD pcs (965 phenom and 1ghz Athlon) seem to have alot, the 1ghz has yet to finish a wu, keeps erroring but runs other projects fine.
Is this just a AMD bug?
Hello all I'm new to the system :)
I also have problems with client errors, seems my intel machine has yet to error, but my AMD pcs (965 phenom and 1ghz Athlon) seem to have alot, the 1ghz has yet to finish a wu, keeps erroring but runs other projects fine.
Is this just a AMD bug?
OK ... all sorts of stuff seem to happening with these two machines - and notably their wingmen are also frequently going down in flames on these WU's. Predominantly trouble with locating etc the input files. When I get the chance ( be few hours yet ) I'll give the devs a note on this .... server side problem perhaps.
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
... my AMD pcs (965 phenom and 1ghz Athlon) seem to have alot, the 1ghz has yet to finish a wu, keeps erroring but runs other projects fine.
Are you overclocking those two machines? The errors, though different on each machine, are very much reminiscent of what you can see if the overclocking is too aggressive or if the cpu cooling is inadequate. And, yes, it is possible for other project's tasks to complete OK whilst E@H tasks cannot, once again due to overclocking.
This is not a problem that is specific to AMD machines.
30 started tasks, none completed!
)
I don't think the latest change to prevent this behavior made it into .36, I think you need .37 or later for it ...
This is an issue that has been a serious problem for some time... there have been I think 3 fixes tried so far though I have more hopes for the 4th and latest in that it is finally looking at things at the project level and then prioritizing the already running task(s) ...
RE: I don't think the
)
Nope, they didn't make it into .37 unfortunately. Looks like we are waiting for .38 to come out.
BOINC blog
RE: RE: I don't think the
)
I see the dev version is up to 6.10.43. Has this bug been fixed yet?
Because of the way the 6.10.18 scheduler behaves with settings to try and maintain a 4 day queue my million:1 resource share split between milkeyway and collataz/gpugrid is ignored in favor of trying to keep 4 days of GPU work on hand, vs my desired behavior of keeping the max number of MW work units on hand and only running the other cuda projects if I run out of MW.
Edit: What I'm looking for from the newest client is the .36 feature that allows setting resource share to zero to define a backup only project.
They certainly made another
)
They certainly made another attempt at fixing it, so 6.10.43 is definitely worth checking out. This particular build also fixes a bug that was introduced at some point that made the Options button no longer work. Dunno how long that bug was around for, but..
RE: Edit: What I'm looking
)
That's in 6.10.43, but it needs an update of server software on the project as well to work. So far only Seti has this update, with a slight possibility of DrugDiscovery@Home having it as well, since Jack updated the code yesterday.
Hello all I'm new to the
)
Hello all I'm new to the system :)
I also have problems with client errors, seems my intel machine has yet to error, but my AMD pcs (965 phenom and 1ghz Athlon) seem to have alot, the 1ghz has yet to finish a wu, keeps erroring but runs other projects fine.
Is this just a AMD bug?
RE: Hello all I'm new to
)
OK ... all sorts of stuff seem to happening with these two machines - and notably their wingmen are also frequently going down in flames on these WU's. Predominantly trouble with locating etc the input files. When I get the chance ( be few hours yet ) I'll give the devs a note on this .... server side problem perhaps.
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
Thanks for looking into the
)
Thanks for looking into the problem for me, I'm using version 6.10.18
if theres any other specs you need let me know.
thanks again
RE: Thanks for looking into
)
Well the main thing is that you haven't hidden the computers within your account, so I can go and inspect matters. ;-)
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: ... my AMD pcs (965
)
Are you overclocking those two machines? The errors, though different on each machine, are very much reminiscent of what you can see if the overclocking is too aggressive or if the cpu cooling is inadequate. And, yes, it is possible for other project's tasks to complete OK whilst E@H tasks cannot, once again due to overclocking.
This is not a problem that is specific to AMD machines.
Cheers,
Gary.