Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Generalized Fermat Prime Search :
GFN-17-MEGA at CPU B-Limit
| Author |
Message |
|---|
 Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13570
ID: 53948
Credit: 249,621,575
RAC: 130,561
                           
|
|
We appear to have reached the b-limit for the x87 transform on 17-MEGA. I've disabled the GPU apps for 17-MEGA, and you will no longer be able to get CPU tasks for 17-MEGA.
If you have 17-MEGA CPU tasks in progress, you might want to consider aborting them if they're near the beginning. If they're near the end I'd probably let them complete. It's impossible to predict which tasks will succeed and which will fail.
More details can be found here: https://www.primegrid.com/forum_thread.php?id=8557, although discussion should occur in this thread.
____________
My lucky number is 75898524288+1 | |
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 858
ID: 301928
Credit: 503,804,973
RAC: 264,743
                
|
We're still theoretically about 6 million away from the B limit, but the limit is fuzzy and some candidates will produce much higher (or lower) error values than surrounding candidates.
There is no fixed limit which could be calculated easily. B/MaxErr graph look like set of triangles (or like a triangle wave where triangles have same amplitude, but "DC offset" rising slowly as B grows). First, only few tasks (top corners of triangles) will touch the limit line, but almost all of tasks still could be completed. Over time, as B and "DC offset" grows, more and more tasks will be above limit, and error will eventually be noticed by "limit test" or something like this.
| |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13570
ID: 53948
Credit: 249,621,575
RAC: 130,561
|
|
Besides the original error, we've now seen the same task error out on a Kaby Lake, an Ivy Bridge, and the same model Sandy Bridge as the original. It's safe to say we've hit the limit.
____________
My lucky number is 75898524288+1 | |
|
mikey
Send message
Joined: 17 Mar 09
Posts: 1252
ID: 37043
Credit: 523,220,863
RAC: 142,324
            
|
We appear to have reached the b-limit for the x87 transform on 17-MEGA. I've disabled the GPU apps for 17-MEGA, and you will no longer be able to get CPU tasks for 17-MEGA.
If you have 17-MEGA CPU tasks in progress, you might want to consider aborting them if they're near the beginning. If they're near the end I'd probably let them complete. It's impossible to predict which tasks will succeed and which will fail.
More details can be found here: https://www.primegrid.com/forum_thread.php?id=8557, although discussion should occur in this thread.
So are you going to remove the cpu checkbox on the Primegrid project preferences page at some point?
Generalized Fermat Prime Search n=17, b>=42,597,774 (GFN-17-Mega or Genefer 131072 Mega)
b131072+1 (or b217+1), b>=42,597,774
Deadline: 4 days (up to 30 days)
Recent average CPU time: 12:05:54
Recent average GPU time: 0:24:07
CPU
OpenCL1,2,7,10,11
OpenCL1,6,7 | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13570
ID: 53948
Credit: 249,621,575
RAC: 130,561
|
So are you going to remove the cpu checkbox on the Primegrid project preferences page at some point?
Yes, but that's not a trivial change. It will have to wait until later today.
____________
My lucky number is 75898524288+1 | |
|
Message boards :
Generalized Fermat Prime Search :
GFN-17-MEGA at CPU B-Limit |
