mfaktc: a CUDA program for Mersenne prefactoring

[axn]

Originally Posted by xx005fs

Because TF uses integer operations, which is much faster on turing than on pascal.

To be more specific, integer mutliply thruput is much higher on Turing compared to pascal.

[RobertKazan]

Originally Posted by xx005fs

Because TF uses integer operations, which is much faster on turing than on pascal.

Thank you for information

[clowns789]

Originally Posted by lolapus

If anyone can compile mfaktc for cuda version 11.2 I would appreciate it.

I would as well! I'm considering a dual boot into Arch Linux to use it on my 3060 Ti in the meantime.

[DrobinsonPE]

I borrowed a GTX 660 from my nephew to benchmark it.

TF exp=103529149 from 74 to 75.

Windows 10:

Mfaktc-0.21.win_cuda100-2047 errored out and would not run even though the GTX 660 is a cuda 3.0 compute card and it should be supported.

Mfaktc-0.21.win_cuda80-2047 works and produces about 184 GHz-d/day.

Linux:

Mfaktc-0.21.linux64.cuda10.1 works and produces about 195 GHz-d/day.

Interesting note:

On Linux, when mfaktc first started up the .ini file was configured with GPUSieveSize=2047. Mfaktc ran at about 195 GHz-d/day but the computer was almost unusable because mfaktc was using both the GPU and CPU at almost 100%. I changed to GPUSieveSize=64 and mfkatc was then only using the GPU and ran at about 192 GHz-d/day. I incrementally increased to GPUSieveSize=256 where it ran at about 195 GHz-d/day and just started to use the CPU a little bit.

[rebirther]

Originally Posted by TheJudger

Finally got in touch with a RTX 3090:

mfaktc v0.22-pre8 (64bit built)
[...]
CUDA version info
  binary compiled for CUDA  11.10
  CUDA runtime version      11.10
  CUDA driver version       11.10

CUDA device info
  name                      GeForce RTX 3090
  compute capability        8.6
  max threads per block     1024
  max shared memory per MP  102400 byte
  number of multiprocessors 82
  clock rate (CUDA cores)   1755MHz
  memory clock rate:        9751MHz
  memory bus width:         384 bit
[...]
Starting trial factoring M66362159 from 2^74 to 2^75 (57.65 GHz-days)
 k_min =  142321062303420
 k_max =  284642124610180
Using GPU kernel "barrett76_mul32_gs"
Date    Time | class   Pct |   time     ETA | GHz-d/day    Sieve     Wait
Nov 01 00:21 |    0   0.1% |  0.794  12m41s |   6535.09    82485    n.a.%
Nov 01 00:21 |    4   0.2% |  0.788  12m35s |   6584.85    82485    n.a.%
Nov 01 00:21 |    9   0.3% |  0.776  12m23s |   6686.67    82485    n.a.%
[...]
Nov 01 00:35 | 4617 100.0% |  0.849   0m00s |   6111.73    82485    n.a.%
no factor for M66362159 from 2^74 to 2^75 [mfaktc 0.22-pre8 barrett76_mul32_gs CUDA 11.10 arch 8.0] B29A657C
tf(): total time spent: 13m 28.241s

GPU clock was 1650-1680 MHz once the GPU heated up, at startup it was a bit higher.
Power consumption during the run was about 340-345 Watt as reported by nvidia-smi.

Oliver

@TheJudger: Can you provide us the mfaktc v0.22-pre8 cuda11 for windows, would help a lot of us to bring up more computing power.

[ramgeis]

Originally Posted by rebirther

@TheJudger: Can you provide us the mfaktc v0.22-pre8 cuda11 for windows, would help a lot of us to bring up more computing power.

[LaurV]

Oliver is a busy guy, he doesn't read these threads very often. You can PM him.

[TheJudger]

Hi there,

there are absolutely no specific changes in my current devel version which was done for Ampere generation (e.g. Geforce 30xx series). In fact 0.21 runs fine on Ampere when using CUDA 11.x. I have had many issues when creating binaries for Windows, from one day to another day without changing the code I wasn't able a binary which works (guess windows update caused the change)... others had similiar issues while others can compile... I have no clue what the root cause is. Maybe it is the code, maybe it is the why the Makefile calls the compiler... I'm anoyed as hell about this (while I know it is possible that I caused the issue by myself...)!

Oliver

[ZFR]

Is there a 0.22 version available of mfaktc? The link in the opening post has only up to 0.21.

[James Heinrich]

Originally Posted by ZFR

Is there a 0.22 version available of mfaktc? The link in the opening post has only up to 0.21.

Not yet. The developer (TheJudger, aka Oliver) has been working on v0.22 for a long time now, but as yet there is no public release.

[ZFR]

Originally Posted by James Heinrich

Not yet. The developer (TheJudger, aka Oliver) has been working on v0.22 for a long time now, but as yet there is no public release.

Gotcha, thanks.