Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Number crunching :
DYFL v4.01
| Author |
Message |
|---|
xii5kuSend message
Joined: 17 Dec 16
Posts: 96
ID: 476505
Credit: 1,630,752,495
RAC: 2,229,401
                  
|
|
The home page https://www.primegrid.com/ says "CPU^(MT-R F) GPU" at DYFL.
Yet according to https://www.primegrid.com/apps.php,
– there is no cpuGFNEXTREME,
– whereas OCLcudaGFNEXTREME and openclatiGFNEXTREME are now at v4.01, hence should have fast proofs enabled now. Therefore the home page should show GPU^F instead of just GPU.
Just a temporary glitch? | |
|
xii5kuSend message
Joined: 17 Dec 16
Posts: 96
ID: 476505
Credit: 1,630,752,495
RAC: 2,229,401
|
|
PS,
genefer22's data allocation for DYFL workunits would be >81 MBytes according to a quick spot check. That looks like something at which a Ryzen 7 5800X3D or a Milan-X would be useful at, with 96 MBytes last-level cache per core complex. | |
|
 Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14044
ID: 53948
Credit: 482,306,103
RAC: 564,923
                            
|
The home page https://www.primegrid.com/ says "CPU^(MT-R F) GPU" at DYFL.
Yet according to https://www.primegrid.com/apps.php,
– there is no cpuGFNEXTREME,
– whereas OCLcudaGFNEXTREME and openclatiGFNEXTREME are now at v4.01, hence should have fast proofs enabled now. Therefore the home page should show GPU^F instead of just GPU.
Just a temporary glitch?
My mistake. The apps page is correct. There is no DYFL CPU app. I will correct this later in the day.
____________
My lucky number is 75898524288+1 | |
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3256
ID: 130544
Credit: 2,459,260,966
RAC: 4,370,694
                   
|
|
& there's us getting all excited. | |
|
|
|
|
but the cpu application for DYFL is under testing?... ) | |
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14044
ID: 53948
Credit: 482,306,103
RAC: 564,923
|
but the cpu application for DYFL is under testing?... )
No.
It will be too slow, and we're not going to enable it.
____________
My lucky number is 75898524288+1 | |
|
|
|
but the cpu application for DYFL is under testing?... )
No.
It will be too slow, and we're not going to enable it.
Is it the same case for GFN 22 CPU? Will it stay enabled?
____________
My lucky number is 6219*2^3374198+1
 | |
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 1051
ID: 301928
Credit: 563,881,725
RAC: 768
     
|
but the cpu application for DYFL is under testing?... )
No.
It will be too slow, and we're not going to enable it.
Is it the same case for GFN 22 CPU? Will it stay enabled?
GFN-22 works with smaller numbers, fast algorithm exist for this range. GPN-22 CPU will stay enabled almost until the end of project (until it reaches DYFL).
| |
|
|
|
|
Can someone provide sample run times before and after V4.01 for DYFL for a GTX 10xxx series card? | |
|
|
|
|
Like Stream says, CPU apps can take advantage of other, more efficient "transforms" when the value of b is low. This is the case for regular (non-DYFL) GFN-22 where b is currently under 250'000. So, when you take into account that GFN-22 candidate numbers are roughly twice as long (in digit count) as GFN-21 candidates, CPUs do well on GFN-22 regular. /JeppeSN | |
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 843
ID: 164101
Credit: 306,554,426
RAC: 5,458
|
Like Stream says, CPU apps can take advantage of other, more efficient "transforms" when the value of b is low. This is the case for regular (non-DYFL) GFN-22 where b is currently under 250'000. So, when you take into account that GFN-22 candidate numbers are roughly twice as long (in digit count) as GFN-21 candidates, CPUs do well on GFN-22 regular. /JeppeSN
These tables will illustrate this:
Ryzen 9 7950X, 8 threads, 2 tasks
GFN sec credit credit/sec
21 80000/2 161000 4.0
22 145000/2 735000 10.1 RTX 3070 Ti
GFN sec credit credit/sec
21 20000 161000 8.0
22 71000 735000 10.3
DYFL 79000 954000 12.1
GPU performance of DYFL is better because of long job credit bonus. On the other hand, the likelihood of success is greater for GFN-21 then the choice depends on your goal.
CPU performance of GFN-22 is much better because of a more efficient transform. However, 12M or 22M digits is a matter of choice. | |
|
Post to thread
Message boards :
Number crunching :
DYFL v4.01 |
