Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Generalized Fermat Prime Search :
GPU processing of GFN-15
| Author |
Message |
|---|
|
|
|
Since these are so small, and I have a mega GPU crunching them (RTX 3090), has anyone figured out about how simultaneous tasks you can run without causing major slow down? I was doing 2, now I am doing 4, they run in just over a minute time.
I am thinking on trying 8, since they go so fast. The GPU is hardly taxed. Load is about 75% and temps are the same as the room temp. It is not using much memory at all.
Had anyone tried more than 8?
____________
My lucky numbers are 121*2^4553899-1 and 3756801695685*2^666669±1
My movie https://vimeo.com/manage/videos/502242 | |
|
 Nick 
Send message
Joined: 11 Jul 11
Posts: 2298
ID: 105020
Credit: 8,430,342,272
RAC: 6,831,360
                            
|
Since these are so small, and I have a mega GPU crunching them (RTX 3090), has anyone figured out about how simultaneous tasks you can run without causing major slow down? I was doing 2, now I am doing 4, they run in just over a minute time.
I am thinking on trying 8, since they go so fast. The GPU is hardly taxed. Load is about 75% and temps are the same as the room temp. It is not using much memory at all.
Had anyone tried more than 8?
Hey man I noticed you are running 2 x SOB at once on your excellent new computer. You need to only have 1 task running - you will get at least 1.5 times better throughput. I remember that 1 task took about 10 hours. I can confirm this in about a week's time when I move from PSP to SOB. | |
|
|
|
|
A while back I tried 16 on my 1660Ti and the times were close to double running 8 threads (so basically the same number of tasks per hour). At 16 tasks at a time it was about 480 seconds per task. However, I quickly hit my 10k hour goals and moved on to other GFN tasks, so I haven't tried running the tasks recently.
____________
Werinbert is not prime... or PRPnet keeps telling me so.
Badge score: 2x3 + 5x4 + 5x5 + 5x7 + 1x9 + 3x10 = 125 | |
|
|
|
|
Current at 8. They are processing faster than 4, CPU time has not changed and heat of card has not changed. Wonder if I should try 16?
____________
My lucky numbers are 121*2^4553899-1 and 3756801695685*2^666669±1
My movie https://vimeo.com/manage/videos/502242 | |
|
tngSend message
Joined: 29 Aug 10
Posts: 486
ID: 66603
Credit: 47,484,342,501
RAC: 27,426,127
                                                   
|
Current at 8. They are processing faster than 4, CPU time has not changed and heat of card has not changed. Wonder if I should try 16?
My approach is to increase until I find the point of diminishing returns or something breaks. Don't think I've done that with GFN, but have with PPS-Sieve and manual sieving. With PPS-Sieve throughput decreased past a certain number; manual sieving developed problems with too many simultaneous tasks.
____________
 | |
|
Message boards :
Generalized Fermat Prime Search :
GPU processing of GFN-15 |
