Hello. I write in german, why my english is not the best. Ich habe eine neue Nvidia Grafikkarte RTX2070. Die ersten WU's liefen sehr gut "LATeah0104W, LATeah1025L". Die aktuellen WU's "LATeah2002L" und andere speichern kein Zwischenergebnis und fangen immer bei "0" wieder an. Auch die Nutzung der Grafikkarte ist sehr gering "60-70W statt 140W" und statt nur 10min über 50min Rechenzeit. Schauen Sie bitte mal nach. Danke.
Hello. I write in german, why my english is not the best. Ich habe eine neue Nvidia Grafikkarte RTX2070. Die ersten WU's liefen sehr gut "LATeah0104W, LATeah1025L". Die aktuellen WU's "LATeah2002L" und andere speichern kein Zwischenergebnis und fangen immer bei "0" wieder an. Auch die Nutzung der Grafikkarte ist sehr gering "60-70W statt 140W" und statt nur 10min über 50min Rechenzeit. Schauen Sie bitte mal nach. Danke.
Derzeit werden verschiedene Typen von Aufgaben an die Rechner verschickt, die sich auch in der Laufzeit etwas unterscheiden. Diese Typen verschickt der Server für einen bis mehrere Tage, bis sie abgearbeitet sind. Manche laufen auch auf den RTX Karten, andere leider nicht. Am besten keine Aufgaben für die Karte anfordern, wenn sie nicht laufen, dann vielleicht einige Tage später nochmal probieren. Eine endgültige technische Lösung ist noch nicht angekündigt.
Inzwischen kann man aber CPU Aufgaben auf dem Rechner laufen lassen, dazu unter Webseite - Konto - Preferences - Projekt - Applications die anderen Anwendungen auswählen. :-)
The currently configured maximum daily quota is 32*8=128 per core. You probably reported a lot of computing errors lowering your daily quota (this is to prevent misconfigured hosts to trash arbitrary numbers of tasks). Every successful result returned should double your quota.
I just doubled the quota for GPU apps (32*16) for the remaining time of the current "short" tasks (LATeah2003L), will set it back when we start issuing tasks that do run (roughly) as long as the BRP4G do.
Bernd, does your version of the server code allow you to set a 'maximum tasks in progress' for the "Gamma-ray pulsar binary search #1 on GPUs" app, during this current run of short tasks? If so, that might be a better tool to use.
Using a daily quota risks a huge demand for tasks as each computer passes its day-end, followed by a loss of project power in the evenings.
The problem is aggravated by this project's reliance on DCF for estimating runtimes. If there are users who run this app exclusively, they will be requesting up to four times the normal number of tasks, which will put a strain on your work generators and database. Then, when normal-length tasks are issued again, those machines may find themselves in deadline trouble.
In hindsight, it might have been better if <rsc_fpops_est> could have been reduced to match the new running time, but it might be too late for that now.
Edit - looks like the 'maximum tasks in progress' control should be available to you, although the syntax has changed over the years.
<max_wus_in_progress> A </max_wus_in_progress>
[ <max_wus_in_progress rsc="cpu"> B </max_wus_in_progress> ]
[ <max_wus_in_progress rsc="gpu"> C </max_wus_in_progress> ]
How long are the 10x shorter tasks expected to last?
How long is a piece of string? :-) Each data file (in this case LATeah2003L.dat) usually provides tasks for a few days which may well be the case this time also.
I'm not sure how you get 10x. There are some tasks for the previous data file (LATeah2002L.dat) in your tasks list that show about 890 secs. Your 2003L tasks currently show around 230 secs although some very early ones were a bit shorter. A fair speedup to quote is more like 4x and not 10x.
Don't get too carried away - the party will probably be over soon enough :-).
String theory is finally proven.........by a pulsar.......expected to find this over at Cern
Well Gary you and I are about to start year #15 here so maybe they are just making sure we stay here another year.......not that I will catch you by just using one of my GeForce cards but my light bill keeps telling me the other ones need to stay in hibernation
Hi Samson,
Good to see you still hanging around here :-).
String Theory Mechanic wrote:
... Well Gary you and I are about to start year #15 here ...
Did you have to put it quite like that?? :-). Woulda sounded much better as, "soon be completing #14" -- or better still you could have said, "we've only completed 13 years so we're still just spring chickens ...". When you get to our age you should always round down ....
Hi Samson,
Good to see you still hanging around here :-).
String Theory Mechanic wrote:
... Well Gary you and I are about to start year #15 here ...
Did you have to put it quite like that?? :-). Woulda sounded much better as, "soon be completing #14" -- or better still you could have said, "we've only completed 13 years so we're still just spring chickens ...". When you get to our age you should always round down ....
Well Gary imagine if we were only 15 years old now ........then we could still dream that will could travel to Mars or farther out across the Milky Way when we master traveling at the speed of light through a worm hole or bending space-time!
Like that summer when I was 11 years old and thought I would get to fly to the Moon some day back in 1969
Now I can only hope to see us land humans on Mars and return......or maybe they can figure out how to make us live another 500 years or so
Yep, now work is coming in
)
Yep, now work is coming in (and out:) at a solid rate. Back in business! Keep up the great work Bernd and the Einstein@Home team!
@Bernd, Any idea how many
)
Sorry wrong thread. Was going to ask about the tuning run.
BOINC blog
Hello. I write in german, why
)
Hello. I write in german, why my english is not the best. Ich habe eine neue Nvidia Grafikkarte RTX2070. Die ersten WU's liefen sehr gut "LATeah0104W, LATeah1025L". Die aktuellen WU's "LATeah2002L" und andere speichern kein Zwischenergebnis und fangen immer bei "0" wieder an. Auch die Nutzung der Grafikkarte ist sehr gering "60-70W statt 140W" und statt nur 10min über 50min Rechenzeit. Schauen Sie bitte mal nach. Danke.
Roland Glaubitz
)
Hallo, das Problem ist bekannt und betrifft tatsächlich alle RTX Karten von Nvidia, also nur diese neuen Karten. Archae86 hat viele Infos dazu gesammelt: https://einsteinathome.org/de/content/pascal-again-available-turing-may-be-coming-soon
Derzeit werden verschiedene Typen von Aufgaben an die Rechner verschickt, die sich auch in der Laufzeit etwas unterscheiden. Diese Typen verschickt der Server für einen bis mehrere Tage, bis sie abgearbeitet sind. Manche laufen auch auf den RTX Karten, andere leider nicht. Am besten keine Aufgaben für die Karte anfordern, wenn sie nicht laufen, dann vielleicht einige Tage später nochmal probieren. Eine endgültige technische Lösung ist noch nicht angekündigt.
Inzwischen kann man aber CPU Aufgaben auf dem Rechner laufen lassen, dazu unter Webseite - Konto - Preferences - Projekt - Applications die anderen Anwendungen auswählen. :-)
Bernd Machenschalk wrote:The
)
Bernd, does your version of the server code allow you to set a 'maximum tasks in progress' for the "Gamma-ray pulsar binary search #1 on GPUs" app, during this current run of short tasks? If so, that might be a better tool to use.
Using a daily quota risks a huge demand for tasks as each computer passes its day-end, followed by a loss of project power in the evenings.
The problem is aggravated by this project's reliance on DCF for estimating runtimes. If there are users who run this app exclusively, they will be requesting up to four times the normal number of tasks, which will put a strain on your work generators and database. Then, when normal-length tasks are issued again, those machines may find themselves in deadline trouble.
In hindsight, it might have been better if <rsc_fpops_est> could have been reduced to match the new running time, but it might be too late for that now.
Edit - looks like the 'maximum tasks in progress' control should be available to you, although the syntax has changed over the years.
https://boinc.berkeley.edu/trac/wiki/ProjectOptions?version=87 (ten years old!) has
whereas https://boinc.berkeley.edu/trac/wiki/ProjectOptions (current version) has
How long are the 10x shorter
)
How long are the 10x shorter tasks expected to last? But more importantly, thanks for the early Christmas present!
I'm racking up points like cRaZZZy! 
10esseeTony wrote:How long
)
How long is a piece of string? :-) Each data file (in this case LATeah2003L.dat) usually provides tasks for a few days which may well be the case this time also.
I'm not sure how you get 10x. There are some tasks for the previous data file (LATeah2002L.dat) in your tasks list that show about 890 secs. Your 2003L tasks currently show around 230 secs although some very early ones were a bit shorter. A fair speedup to quote is more like 4x and not 10x.
Don't get too carried away - the party will probably be over soon enough :-).
Cheers,
Gary.
Gary Roberts wrote: How long
)
String theory is finally proven.........by a pulsar.......expected to find this over at Cern
Well Gary you and I are about to start year #15 here so maybe they are just making sure we stay here another year.......not that I will catch you by just using one of my GeForce cards but my light bill keeps telling me the other ones need to stay in hibernation
Hi Samson, Good to see you
)
Hi Samson,
Good to see you still hanging around here :-).
Did you have to put it quite like that?? :-). Woulda sounded much better as, "soon be completing #14" -- or better still you could have said, "we've only completed 13 years so we're still just spring chickens ...". When you get to our age you should always round down ....
Cheers,
Gary.
Gary Roberts wrote:Hi
)
Well Gary imagine if we were only 15 years old now ........then we could still dream that will could travel to Mars or farther out across the Milky Way when we master traveling at the speed of light through a worm hole or bending space-time!
Like that summer when I was 11 years old and thought I would get to fly to the Moon some day back in 1969
Now I can only hope to see us land humans on Mars and return......or maybe they can figure out how to make us live another 500 years or so