Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Generalized Fermat Prime Search :
Do You Feel Lucky? Jumped from 2+ days to 6+ days remaining. Why?
| Author |
Message |
|---|
|
|
|
MacBook Pro 16" M1 Max 64GB RAM running Ventura 13.2.1.
BOINC veteran but new to PrimeGrid.
DYFL? genefer_extreme_68937150_2 (executable genefer22g_macarm64_22.11.04) running on 0.000635 CPUs & Apple M1 Max GPU. The other 9 CPU cores are loaded with Einstein tasks, nothing else using the GPU.
I shut down then restarted less than 15 minutes later. The task jumped from less than 2 days remaining to over 6 days. In 5 hours it's only come down to 5 days 14 hours.
The task is at 47% completion in 30 hours, on target for the original 2+ days. Calculating the % progress since the restart also results in the original 2+ days.
I understand that check-pointing isn't an exact science for most BOINC tasks. BOINC is set to checkpoint at most every 60 seconds.
Any explanation of what went amok?
TIA,
Sam
| |
|
Yves Gallot Volunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 820
ID: 164101
Credit: 305,989,513
RAC: 1,728
|
|
Remaining time is always incorrect with Boinc. It evaluates complex statistics and generates a wrong estimate.
The simple formula elapsed = progress * total and then remaining = elapsed / progress - elapsed is accurate.
I don't know why Boinc doesn't use this relation and if it will be fixed one of these days.
Note that check-pointing and progress are two independent features. One checkpoint every 60 seconds for a long task is too much. During a checkpoint, CPU and GPU are synchronized, GPU memory is read and written to the disk. That slows the system down. 600 seconds is sufficient. | |
|
 mikey
Send message
Joined: 17 Mar 09
Posts: 1788
ID: 37043
Credit: 793,475,008
RAC: 1,211,045
                     
|
MacBook Pro 16" M1 Max 64GB RAM running Ventura 13.2.1.
BOINC veteran but new to PrimeGrid.
DYFL? genefer_extreme_68937150_2 (executable genefer22g_macarm64_22.11.04) running on 0.000635 CPUs & Apple M1 Max GPU. The other 9 CPU cores are loaded with Einstein tasks, nothing else using the GPU.
I shut down then restarted less than 15 minutes later. The task jumped from less than 2 days remaining to over 6 days. In 5 hours it's only come down to 5 days 14 hours.
The task is at 47% completion in 30 hours, on target for the original 2+ days. Calculating the % progress since the restart also results in the original 2+ days.
TIA,
Sam
Most projects advise us to wait until at least 10 tasks, of the same kind, have been processed to get a good handle on the actual time it takes to run a task. Prime Grid tasks are no different in other projects long tasks in that they tend to over estimate rather than under estimate the actual time it will take to finish a task when we get new for us tasks. | |
|
pschoeferVolunteer developer
Volunteer tester
Send message
Joined: 20 Sep 05
Posts: 686
ID: 845
Credit: 2,910,184,413
RAC: 199,509
                           
|
Remaining time is always incorrect with Boinc. It evaluates complex statistics and generates a wrong estimate.
The simple formula elapsed = progress * total and then remaining = elapsed / progress - elapsed is accurate.
I don't know why Boinc doesn't use this relation and if it will be fixed one of these days.
BOINC does not use this relation by default, because some applications have progress bars that are inaccurate or do not update very often, or have no progress bar at all. However, both the project and the user (via app_config.xml) already can tell the client to use the simple linear interpolation (fraction_done_exact). PrimeGrid actually does this, but it looks like it does not work after a restart.
____________
  | |
|
|
|
Note that check-pointing and progress are two independent features. One checkpoint every 60 seconds for a long task is too much. During a checkpoint, CPU and GPU are synchronized, GPU memory is read and written to the disk. That slows the system down. 600 seconds is sufficient.
I didn't know that, thanks. I changed the checkpoint to 600. PrimeGrid is the only project (that I know of) using the M1's GPU. | |
|
|
|
BOINC does not use this relation by default, because some applications have progress bars that are inaccurate or do not update very often, or have no progress bar at all. However, both the project and the user (via app_config.xml) already can tell the client to use the simple linear interpolation (fraction_done_exact). PrimeGrid actually does this, but it looks like it does not work after a restart.
Thanks, now it makes perfect sense. BOINC's feeble attempts to update this accurately say it's down to 2 days when it only has about 12 hours to go.
I may do an app_config.xml just so this is more linear. | |
|
Post to thread
Message boards :
Generalized Fermat Prime Search :
Do You Feel Lucky? Jumped from 2+ days to 6+ days remaining. Why? |
