The setup is quite different. We split the original WUs in half and processed these one after the other, because that reduced the memory consumption. For the new WUs this isn't necessary (and actually more difficult). Also too the "recalc" stage at the end (mostly done on the CPU) is much shorter compared to the total runtime. I'm not entirely sure whether the new WUs run slower or faster on a specific machine, by how much and what properties influence that. On our machines the runtimes of the old an new workunits are comparable. So far I don't have enough results for a statistical evaluation.
Is the part at the end about the run times beeing "comparable" still true ?
Just wondering ...
So the new low frequency gravity wave tasks take 3x longer to run on my 4070 super and give 40% of the credit… seems like pretty clear disincentive to run these vs BRP7 currently.
So the new low frequency gravity wave tasks take 3x longer to run on my 4070 super and give 40% of the credit… seems like pretty clear disincentive to run these vs BRP7 currently.
Same here. The new O3AS WUs are taking 2-3 times longer and giving anywhere from 10% t0 40% (1000 - 4000) of the credit compared to the old WUs. So even if the new WU credit was 20k we'd have less credit/hour than before.
I'm crunching tasks from BRP7/MeerKAT, but my workstation stopped receiving those tasks and started receiving tasks from the All-Sky Gravitational Wave search on O3. However, in the project preferences, I selected to receive only BRP7. Any idea what might be happening?
Don't forget, prior to January 16, 2024, these work units only gave 5000 credits.
I guess the New year's gift is over. :-)
That’s still a 20% reduction in credit per task (5k to 4k) and 2-300% increase in task run time, while also eliminating the ability to stagger tasks, further driving down throughput.
I’m not concerned about the numbers themselves, just use them as indicators of what the project prioritizes; seems pretty clear the gravity wave data seems to be taking a backseat now vs earlier this year where they clearly prioritized it via granting double credit.
The new LF task will have lower credit and takes longer time to complete. Some crunchers who are interested in credit (some don't care about credit) will start to deselect O3AS task and instead select BRP7. Only time will tell if such switch will happen.
Bernd Machenschalk
)
Is the part at the end about the run times beeing "comparable" still true ?
Just wondering ...
sfv
So the new low frequency
)
So the new low frequency gravity wave tasks take 3x longer to run on my 4070 super and give 40% of the credit… seems like pretty clear disincentive to run these vs BRP7 currently.
br3achd wrote: So the new
)
Same here. The new O3AS WUs are taking 2-3 times longer and giving anywhere from 10% t0 40% (1000 - 4000) of the credit compared to the old WUs. So even if the new WU credit was 20k we'd have less credit/hour than before.
For those who care about
)
For those who care about credit otherwise can ignore this post: I think there should be some sort of a baseline to determine "equal pay for equal work". Folding@Home uses i5-750-[at]-2.67GHz as a baseline to determine the credit https://foldingathome.org/faqs/statistics-teams-usernames/how-do-you-decide-how-much-credit-a-work-unit-is-worth-how-do-you-determine-how-many-points-a-work-unit-is-worth/ Not saying that this the right method as long as something resembling an approach towards the principle of "equal pay for equal work".
Don't forget, prior
)
Don't forget, prior to January 16, 2024, these work units only gave 5000 credits.
I guess the New year's gift is over. :-)
+1
)
+1
I'm crunching tasks from
)
I'm crunching tasks from BRP7/MeerKAT, but my workstation stopped receiving those tasks and started receiving tasks from the All-Sky Gravitational Wave search on O3. However, in the project preferences, I selected to receive only BRP7. Any idea what might be happening?
Richard M wrote: Don't
)
That’s still a 20% reduction in credit per task (5k to 4k) and 2-300% increase in task run time, while also eliminating the ability to stagger tasks, further driving down throughput.
I’m not concerned about the numbers themselves, just use them as indicators of what the project prioritizes; seems pretty clear the gravity wave data seems to be taking a backseat now vs earlier this year where they clearly prioritized it via granting double credit.
to recap, The reason for the
)
to recap, The reason for the bump in credit from 5000 → 10000 is explained in the post https://einsteinathome.org/content/new-improved-gravitational-wave-app-happy-new-year-2024-special
The new LF task will have lower credit and takes longer time to complete. Some crunchers who are interested in credit (some don't care about credit) will start to deselect O3AS task and instead select BRP7. Only time will tell if such switch will happen.
We'll certainly tune up
)
We'll certainly tune up credit, just need to get the search running technically at all first.
BM