mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware > GPU Computing

Reply
 
Thread Tools
Old 2020-12-07, 03:34   #3444
axn
 
axn's Avatar
 
Jun 2003

23·607 Posts
Default

Quote:
Originally Posted by aheeffer View Post
Ignoring the CudaError, the program goes through all the test TF's and they all fail.
That's too bad. So the error is indicative of genuine problem.
axn is offline   Reply With Quote
Old 2020-12-07, 16:30   #3445
aheeffer
 
Aug 2020

2516 Posts
Default RTX 3060 ti

I gave up on Windows, installed Ubuntu 18.04, the CUDA toolkit and drivers, updated the Makefile and could finally get my RTX 3060 ti running. These are the figures, basic settings (1800 MHz core and 700 Mhz memory) and no optimizations, 334M exponent:

Code:
Using GPU kernel "barrett87_mul32_gs"
Date    Time | class   Pct |   time     ETA | GHz-d/day    Sieve     Wait
Dec 07 17:24 |    0   0.1% |  1.448  23m09s |   2849.02    82485    n.a.%
Dec 07 17:24 |    4   0.2% |  1.421  22m41s |   2903.16    82485    n.a.%
Dec 07 17:24 |    7   0.3% |  1.419  22m38s |   2907.25    82485    n.a.%
Dec 07 17:24 |   12   0.4% |  1.421  22m38s |   2903.16    82485    n.a.%
Dec 07 17:24 |   15   0.5% |  1.418  22m34s |   2909.30    82485    n.a.%
Dec 07 17:24 |   19   0.6% |  1.420  22m35s |   2905.20    82485    n.a.%
And I am having a beer now.
aheeffer is offline   Reply With Quote
Old 2020-12-07, 17:07   #3446
James Heinrich
 
James Heinrich's Avatar
 
"James Heinrich"
May 2004
ex-Northern Ontario

2·3·5·109 Posts
Default

When you get a TF run completed, could you please submit benchmark data (especially including the actual boost clock it runs at while mfaktc is running)
https://www.mersenne.ca/mfaktc.php#benchmark
James Heinrich is offline   Reply With Quote
Old 2020-12-17, 03:26   #3447
Dylan14
 
Dylan14's Avatar
 
"Dylan"
Mar 2017

2×281 Posts
Default

The pkgbuild for mfaktc has been updated. A new CUDA version, 11.2 came out which requires a new driver which has not been fully released (it is in beta). Hence the dependencies have been updated.
For anyone running the beta driver (presently 460.27.04), here is a mfaktc executable with CUDA 11.2 support on Linux.
Attached Files
File Type: zip mfaktc.zip (475.1 KB, 32 views)
Dylan14 is offline   Reply With Quote
Old 2020-12-25, 18:18   #3448
RobertKazan
 
Dec 2020

3 Posts
Default

I hve two cards GTX 1650 Super and GTX 1070 ti.
Why GTX 1650 super about 40 percent faster than GTX 1070 Ti in mfaktc but 2x slower than GTX 1070 ti in CUDA Lucas.
What is the reason for such a noticeable difference in performance in mfaktc, although the number of multiprocessors is noticeably lower than in GTX 1070 TI?
RobertKazan is offline   Reply With Quote
Old 2020-12-25, 18:39   #3449
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

52×11×17 Posts
Default

What is the ratio of double-precision to single-precision computation in each card? This is often written DP and SP in reviews or specs.

Nvidia dropped the DP/SP ratio quite often going from generation to generation, which for us means trial-factoring work gets faster but LL work less so.
VBCurtis is offline   Reply With Quote
Old 2020-12-25, 19:26   #3450
RobertKazan
 
Dec 2020

3 Posts
Default

Quote:
Originally Posted by VBCurtis View Post
What is the ratio of double-precision to single-precision computation in each card? This is often written DP and SP in reviews or specs.

Nvidia dropped the DP/SP ratio quite often going from generation to generation, which for us means trial-factoring work gets faster but LL work less so.
GTX 1650 Super SP 4416/DP 138 GFLOPS Cuda cores 1280
GTX 1070TI SP 8186/DP 256 GFLOPS Cuda cores 2432
As can we see GTX 1070 Ti much quickly SP and DP also?
Why GTX 1650 Super 1,4x faster then GTX 1070Ti in trial-factoring?

Last fiddled with by RobertKazan on 2020-12-25 at 19:28
RobertKazan is offline   Reply With Quote
Old 2020-12-25, 20:39   #3451
lolapus
 
Jun 2020

1 Posts
Default

I have been trying for the past few hours messing around with my 3090 trying everything but I cannot get it to work. If anyone can compile mfaktc for cuda version 11.2 I would appreciate it.

I don't know if this will show up before my other reply but I am trying to get a working program for cuda 11.2 on windows not linux.
lolapus is offline   Reply With Quote
Old 2020-12-25, 20:59   #3452
moebius
 
moebius's Avatar
 
Jul 2009
Germany

22316 Posts
Default

Quote:
Originally Posted by RobertKazan View Post
Why GTX 1650 Super 1,4x faster then GTX 1070Ti in trial-factoring?
Maybe mfaktc can use half precision.The GTX 1070 TI only has 127 GLOPS FP16 and GDDR5 RAM. The GTX 1650 Super has 8832 GLOPS FP16 and GDDR6.

Last fiddled with by moebius on 2020-12-25 at 21:31
moebius is offline   Reply With Quote
Old 2020-12-25, 23:47   #3453
xx005fs
 
"Eric"
Jan 2018
USA

22·53 Posts
Default

Quote:
Originally Posted by RobertKazan View Post
GTX 1650 Super SP 4416/DP 138 GFLOPS Cuda cores 1280
GTX 1070TI SP 8186/DP 256 GFLOPS Cuda cores 2432
As can we see GTX 1070 Ti much quickly SP and DP also?
Why GTX 1650 Super 1,4x faster then GTX 1070Ti in trial-factoring?

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

Last fiddled with by xx005fs on 2020-12-25 at 23:48
xx005fs is offline   Reply With Quote
Old 2020-12-26, 01:13   #3454
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

467510 Posts
Default

Quote:
Originally Posted by xx005fs View Post
Because TF uses integer operations, which is much faster on turing than on pascal.
Aha! Mea Culpa- sorry for the mistaken info on SP.
VBCurtis is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
mfakto: an OpenCL program for Mersenne prefactoring Bdot GPU Computing 1668 2020-12-22 15:38
The P-1 factoring CUDA program firejuggler GPU Computing 753 2020-12-12 18:07
gr-mfaktc: a CUDA program for generalized repunits prefactoring MrRepunit GPU Computing 32 2020-11-11 19:56
mfaktc 0.21 - CUDA runtime wrong keisentraut Software 2 2020-08-18 07:03
World's second-dumbest CUDA program fivemack Programming 112 2015-02-12 22:51

All times are UTC. The time now is 22:49.

Thu Mar 4 22:49:06 UTC 2021 up 91 days, 19 hrs, 0 users, load averages: 2.22, 1.73, 1.56

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.