gpuOwL: an OpenCL program for Mersenne primality testing - Page 245

Xyzzy · 2021-01-12, 16:07

Code:

2021-01-12 06:32:56 Intel(R) UHD Graphics 610-0 OpenCL compilation in 6.50 s
2021-01-12 06:33:20 Intel(R) UHD Graphics 610-0 77936867 OK        0 loaded: blockSize 400, 0000000000000003
2021-01-12 06:34:24 Intel(R) UHD Graphics 610-0 77936867 OK      800   0.00%; 53884 us/it; ETA 48d 14:32; 1579c241dc63eca6 (check 21.50s)
2021-01-12 06:43:02 Intel(R) UHD Graphics 610-0 77936867 OK    10000   0.01%; 53947 us/it; ETA 48d 15:45; fc4f135f7cf4ad29 (check 21.46s)
2021-01-12 06:52:20 Intel(R) UHD Graphics 610-0 77936867 OK    20000   0.03%; 53590 us/it; ETA 48d 07:53; 3cd1bd9d5e09cbc5 (check 21.44s)
2021-01-12 07:02:27 Intel(R) UHD Graphics 610-0 77936867 OK    30000   0.04%; 58429 us/it; ETA 52d 16:28; c4e0ff35e3290d98 (check 22.70s)
2021-01-12 07:12:15 Intel(R) UHD Graphics 610-0 77936867 OK    40000   0.05%; 56653 us/it; ETA 51d 01:52; dffe1b1b0d748128 (check 22.27s)
2021-01-12 07:22:05 Intel(R) UHD Graphics 610-0 77936867 OK    50000   0.06%; 56585 us/it; ETA 51d 00:14; 52e286945371ed29 (check 23.69s)
2021-01-12 07:32:33 Intel(R) UHD Graphics 610-0 77936867 OK    60000   0.08%; 60573 us/it; ETA 54d 14:20; 0945da4dc08bdd95 (check 22.21s)
2021-01-12 07:42:47 Intel(R) UHD Graphics 610-0 77936867 OK    70000   0.09%; 59272 us/it; ETA 53d 10:02; 7131fa4eb77f4bb2 (check 21.81s)
2021-01-12 07:52:06 Intel(R) UHD Graphics 610-0 77936867 OK    80000   0.10%; 53773 us/it; ETA 48d 10:57; 8d76071d27ee4221 (check 21.37s)

Xyzzy · 2021-01-14, 15:39

3070

Code:

2021-01-14 09:04:27 GeForce RTX 3070-0 OpenCL compilation in 0.02 s
2021-01-14 09:04:32 GeForce RTX 3070-0 77936867 OK      800   0.00%; 2769 us/it; ETA 2d 11:56; 1579c241dc63eca6 (check 1.14s)
2021-01-14 09:09:09 GeForce RTX 3070-0 77936867 OK   100000   0.13%; 2785 us/it; ETA 2d 12:14; 6d7296b9e2830f50 (check 1.15s)
2021-01-14 09:13:49 GeForce RTX 3070-0 77936867 OK   200000   0.26%; 2787 us/it; ETA 2d 12:11; f0b04b45b0855bd2 (check 1.15s)
2021-01-14 09:18:29 GeForce RTX 3070-0 77936867 OK   300000   0.38%; 2787 us/it; ETA 2d 12:06; 990aa099aad5bf9c (check 1.15s)
2021-01-14 09:23:09 GeForce RTX 3070-0 77936867 OK   400000   0.51%; 2788 us/it; ETA 2d 12:02; c03f94396a5aa29e (check 1.15s)
2021-01-14 09:27:49 GeForce RTX 3070-0 77936867 OK   500000   0.64%; 2787 us/it; ETA 2d 11:57; 591eecd8448042ad (check 1.15s)

Xyzzy · 2021-01-14, 21:32

https://www.mersenneforum.org/showpo...&postcount=119

Xyzzy · 2021-01-30, 00:35

2080 Ti

Code:

2021-01-29 17:10:19 GeForce RTX 2080 Ti-0 OpenCL compilation in 1.66 s
2021-01-29 17:10:20 GeForce RTX 2080 Ti-0 77936867 OK        0 loaded: blockSize 400, 0000000000000003
2021-01-29 17:10:23 GeForce RTX 2080 Ti-0 77936867 OK      800   0.00%; 2051 us/it; ETA 1d 20:24; 1579c241dc63eca6 (check 0.86s)
2021-01-29 17:17:10 GeForce RTX 2080 Ti-0 77936867 OK   200000   0.26%; 2040 us/it; ETA 1d 20:03; f0b04b45b0855bd2 (check 0.87s)

SethTro · 2021-01-30, 13:57

Is it possible to run the stage 2 (of P-1/PM1) after factors where found in stage 1?

Viliam Furik · 2021-01-30, 14:36

Quote:

Originally Posted by SethTro

Is it possible to run the stage 2 (of P-1/PM1) after factors where found in stage 1?

In gpuOwl, most certainly not. In Prime95, yes, AFAIK.

Zenzoma · 2021-01-30, 17:21

Quote:

Originally Posted by ewmayer

MattL, how many watts-at-wall (above system baseline) is the system pulling with the R7 Pro running?

It's using 240W above system baseline running PRP with gpuowl-v7.2-21-g28dbf88.

Xyzzy · 2021-02-12, 13:53

3070

Code:

2021-02-11 16:59:30 GeForce RTX 3070-0 OpenCL compilation in 1.38 s
2021-02-11 16:59:32 GeForce RTX 3070-0 77936867 OK        0 loaded: blockSize 400, 0000000000000003
2021-02-11 16:59:35 GeForce RTX 3070-0 77936867 OK      800   0.00%; 2836 us/it; ETA 2d 13:23; 1579c241dc63eca6 (check 1.17s)
2021-02-11 17:00:31 GeForce RTX 3070-0 77936867 OK    20000   0.03%; 2878 us/it; ETA 2d 14:17; 3cd1bd9d5e09cbc5 (check 1.16s)
2021-02-11 17:01:29 GeForce RTX 3070-0 77936867 OK    40000   0.05%; 2827 us/it; ETA 2d 13:10; dffe1b1b0d748128 (check 1.17s)
2021-02-11 17:02:27 GeForce RTX 3070-0 77936867 OK    60000   0.08%; 2837 us/it; ETA 2d 13:22; 0945da4dc08bdd95 (check 1.16s)
2021-02-11 17:03:25 GeForce RTX 3070-0 77936867 OK    80000   0.10%; 2822 us/it; ETA 2d 13:02; 8d76071d27ee4221 (check 1.16s)
2021-02-11 17:04:22 GeForce RTX 3070-0 77936867 OK   100000   0.13%; 2813 us/it; ETA 2d 12:49; 6d7296b9e2830f50 (check 1.16s)
2021-02-11 17:05:20 GeForce RTX 3070-0 77936867 OK   120000   0.15%; 2813 us/it; ETA 2d 12:49; 79ae5dad855057ad (check 1.16s)
2021-02-11 17:06:17 GeForce RTX 3070-0 77936867 OK   140000   0.18%; 2822 us/it; ETA 2d 12:59; e1db15f897271496 (check 1.16s)
2021-02-11 17:07:15 GeForce RTX 3070-0 77936867 OK   160000   0.21%; 2832 us/it; ETA 2d 13:11; 25b7b6206fc6f085 (check 1.18s)
2021-02-11 17:08:13 GeForce RTX 3070-0 77936867 OK   180000   0.23%; 2837 us/it; ETA 2d 13:16; 6bee5d054f770861 (check 1.16s)
2021-02-11 17:09:12 GeForce RTX 3070-0 77936867 OK   200000   0.26%; 2875 us/it; ETA 2d 14:05; f0b04b45b0855bd2 (check 1.16s)

kriesel · 2021-03-22, 01:09

So many benchmarks in this thread. Many are much less useful or comparable than they could be, because they omit:

Gpuowl version
Operating system (in detail)
Driver name and versions
Gpu power settings
Gpu operating temperature
Statements of any over or under clocking or powering
Probably more relevant variables

Driver VERSION changes have been seen to contribute as much as 5% change.

kriesel · 2021-03-30, 14:03

Two great versions, head to head: P-1 speed, 103M, V6.11-380 vs v7.2-53

XFX Radeon VII, gpu ram clock 1150 MHz, Windows 10 X64 build 19041.867, Celeron G1840 cpu

V6.11-380-g79ea0cc
config.txt: -user kriesel -cpu asr2/radeonvii4 -d 4 -use NO_ASM -maxAlloc 15000 -cleanup
B=1000000,B2=30000000;PFactor=AID,1,2,103218151,-1,76,2
P-1 start stage 1, 2021-03-21 11:17:00
P-1 begin stage 2, 2021-03-21 11:37:15, gpu stage 1 time 0:20:15
P-1 begin stage 2 GCD, 2021-03-21 12:02:56, gpu stage 2 time 25:41; gpu both stages time 0:45:56
P-1 finish last stage 2 GCD, 2021-03-21 12:03:50, total elapsed time 0:46:50
This program version can perform 24hr/0:45:56 = 31.35 wavefront P-1 to B1=1M, B2=30M per day on this gpu. (Timing on gpu without final gcd duration is used for computing throughput, since the gcd occurs on the cpu in parallel with the beginning of the next worktodo entry running on the gpu, and in this case the gpu is limiting on throughput.)

V7.2-53-ge27846f -maxalloc 15G
config.txt: -user kriesel -cpu asr2/radeonvii4 -d 4 -maxAlloc 15G -proof 9 -use NO_ASM -autoverify 10
B1=1000000,B2=30000000;PRP=AID,1,2,103276501,-1,76,2
P-1 start stage 1, 2021-03-25 17:34:31
P-1 begin stage 2, 2021-03-25 18:01:12, gpu stage 1 time 0:26:41
P-1 begin last stage 2 GCD, 2021-03-25 21:31:20, gpu stage 2 time 3:30:18, while no PRP progress occurs; gpu both stages time 3:56:59
P-1 finish last stage 2 GCD, 2021-03-25 21:32:17, total elapsed time 3:57:46
This program version can perform 24h/3:56:59 = 6.08 wavefront P-1 to B1=1M, B2=30M per day on this gpu.
Even ignoring the stage 1 time, if the PRP progress occurring with stage 1 is useful, the drop in P-1 productivity is considerable.
(24hr/3:30:18 = 7.26/day of stage 2 time, 23.2% of the V6.11-380 rate of both stages.
Is stage 2 being slowed by using too much gpu ram, causing paging? retest with -maxalloc 14G)

After system restart and v7.2-53 changed to -maxalloc 14G, run v7.2-53 again:
B1=1000000,B2=30000000;PRP=AID,1,2,103281509,-1,76,2
start program 2021-03-26 07:20:37 (don't count startup time which is incurred once)
P-1 stage 1 start 2021-03-26 07:20:55
P-1 stage 1 end gpu compute, start stage 2 2021-03-26 07:46:35; stage 1 gpu time 0:25:40
P-1 stage 2 end gpu compute 2021-03-26 08:21:30, stage 2 gpu time 0:34:55, combined gpu time 1:00:35
P-1 stage 2 final gcd done & result written 2021-03-26 08:22:35 total elapsed time 1:01:40
(24hr/1:0035 = 23.77 wavefront P-1 to B1=1M, B2=30M per day on this gpu as configured;
needed lower -maxalloc than v6.11-380 for possibly full performance; 23.77/31.35=75.82% of v6.11-380 throughput, even after optimistically neglecting the V7.2 stage 1 cost)

V7.2-53 apparently does not support B2 extension after initial stage 2 completion, despite https://mersenneforum.org/showpost.p...8&postcount=61.
Test method:
B1=1000000,B2=30000000;PRP=AID,1,2,103276501,-1,76,2 in worktodo
Run until those bounds are complete, results file contains the P-1 result line, halt program.
Observe worktodo entry remains B1=1000000,B2=30000000;PRP=AID,1,2,103276501,-1,76,2, gpuowl did not change it to ...,76,0.
Modify to B1=1000000,B2=40000000;PRP=AID,1,2,103276501,-1,76,2, restart program.
Program continues PRP only, without P-1 stage 2 from 30M to 40M or from 1M to 40M.
Perhaps the extension only works during the time window before B2=30000000 is reached.

kriesel · 2021-04-03, 22:15

Quote:

Originally Posted by ewmayer

E-mail exchange between me and George in last 24 hours which may be of use to gpuowl users preparing to switch to the latest build which supports the verified-proof mechanism:

GW: Yes, new gpuowl builds will create proofs for old PRP assignments.

You can switch to the new gpuowl mid-run. You won't get a proof, but the next assignment will.

If you switch early enough in the run, you may get a proof.

Code:

Power 2^power %max for transition to proof generation version (~point of no return/proof)
10     1024    0.0976
9       512    0.1953
8       256    0.3906
7       128    0.7812
6        64    1.5625

I have recently gone through the exercise of migrating old benchmarking partial runs of many different fft lengths to proof-capable versions, with some additional benchmarking for newer versions along the way.
Occasionally I see it's getting close to a cutoff percentage, and save a copy of the exponent folder before continuing to benchmark non-proof versions for comparison purposes. A saved copy is also useful sometimes because gpuowl file versions migrate forwards within limits, but don't migrate backwards, and the fastest version is not always the newest proof-capable version or the same file version number.

2021-01-30, 13:57	#2689
SethTro "Seth" Apr 2019 269₁₀ Posts	P-1 Is it possible to run the stage 2 (of P-1/PM1) after factors where found in stage 1?

2021-03-22, 01:09	#2693
kriesel "TF79LL86GIMPS96gpu17" Mar 2017 US midwest 7×719 Posts	So many benchmarks in this thread. Many are much less useful or comparable than they could be, because they omit: Gpuowl version Operating system (in detail) Driver name and versions Gpu power settings Gpu operating temperature Statements of any over or under clocking or powering Probably more relevant variables Driver VERSION changes have been seen to contribute as much as 5% change.

2021-03-30, 14:03	#2694
kriesel "TF79LL86GIMPS96gpu17" Mar 2017 US midwest 5033₁₀ Posts	p-1 v7 vs v6 Two great versions, head to head: P-1 speed, 103M, V6.11-380 vs v7.2-53 XFX Radeon VII, gpu ram clock 1150 MHz, Windows 10 X64 build 19041.867, Celeron G1840 cpu V6.11-380-g79ea0cc config.txt: -user kriesel -cpu asr2/radeonvii4 -d 4 -use NO_ASM -maxAlloc 15000 -cleanup B=1000000,B2=30000000;PFactor=AID,1,2,103218151,-1,76,2 P-1 start stage 1, 2021-03-21 11:17:00 P-1 begin stage 2, 2021-03-21 11:37:15, gpu stage 1 time 0:20:15 P-1 begin stage 2 GCD, 2021-03-21 12:02:56, gpu stage 2 time 25:41; gpu both stages time 0:45:56 P-1 finish last stage 2 GCD, 2021-03-21 12:03:50, total elapsed time 0:46:50 This program version can perform 24hr/0:45:56 = 31.35 wavefront P-1 to B1=1M, B2=30M per day on this gpu. (Timing on gpu without final gcd duration is used for computing throughput, since the gcd occurs on the cpu in parallel with the beginning of the next worktodo entry running on the gpu, and in this case the gpu is limiting on throughput.) V7.2-53-ge27846f -maxalloc 15G config.txt: -user kriesel -cpu asr2/radeonvii4 -d 4 -maxAlloc 15G -proof 9 -use NO_ASM -autoverify 10 B1=1000000,B2=30000000;PRP=AID,1,2,103276501,-1,76,2 P-1 start stage 1, 2021-03-25 17:34:31 P-1 begin stage 2, 2021-03-25 18:01:12, gpu stage 1 time 0:26:41 P-1 begin last stage 2 GCD, 2021-03-25 21:31:20, gpu stage 2 time 3:30:18, while no PRP progress occurs; gpu both stages time 3:56:59 P-1 finish last stage 2 GCD, 2021-03-25 21:32:17, total elapsed time 3:57:46 This program version can perform 24h/3:56:59 = 6.08 wavefront P-1 to B1=1M, B2=30M per day on this gpu. Even ignoring the stage 1 time, if the PRP progress occurring with stage 1 is useful, the drop in P-1 productivity is considerable. (24hr/3:30:18 = 7.26/day of stage 2 time, 23.2% of the V6.11-380 rate of both stages. Is stage 2 being slowed by using too much gpu ram, causing paging? retest with -maxalloc 14G) After system restart and v7.2-53 changed to -maxalloc 14G, run v7.2-53 again: B1=1000000,B2=30000000;PRP=AID,1,2,103281509,-1,76,2 start program 2021-03-26 07:20:37 (don't count startup time which is incurred once) P-1 stage 1 start 2021-03-26 07:20:55 P-1 stage 1 end gpu compute, start stage 2 2021-03-26 07:46:35; stage 1 gpu time 0:25:40 P-1 stage 2 end gpu compute 2021-03-26 08:21:30, stage 2 gpu time 0:34:55, combined gpu time 1:00:35 P-1 stage 2 final gcd done & result written 2021-03-26 08:22:35 total elapsed time 1:01:40 (24hr/1:0035 = 23.77 wavefront P-1 to B1=1M, B2=30M per day on this gpu as configured; needed lower -maxalloc than v6.11-380 for possibly full performance; 23.77/31.35=75.82% of v6.11-380 throughput, even after optimistically neglecting the V7.2 stage 1 cost) V7.2-53 apparently does not support B2 extension after initial stage 2 completion, despite https://mersenneforum.org/showpost.p...8&postcount=61. Test method: B1=1000000,B2=30000000;PRP=AID,1,2,103276501,-1,76,2 in worktodo Run until those bounds are complete, results file contains the P-1 result line, halt program. Observe worktodo entry remains B1=1000000,B2=30000000;PRP=AID,1,2,103276501,-1,76,2, gpuowl did not change it to ...,76,0. Modify to B1=1000000,B2=40000000;PRP=AID,1,2,103276501,-1,76,2, restart program. Program continues PRP only, without P-1 stage 2 from 30M to 40M or from 1M to 40M. Perhaps the extension only works during the time window before B2=30000000 is reached. Last fiddled with by kriesel on 2021-03-30 at 14:11

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
GPUOWL AMD Windows OpenCL issues	xx005fs	GpuOwl	0	2019-07-26 21:37
Testing an expression for primality	1260	Software	17	2015-08-28 01:35
Testing Mersenne cofactors for primality?	CRGreathouse	Computer Science & Computational Number Theory	18	2013-06-08 19:12
Primality-testing program with multiple types of moduli (PFGW-related)	Unregistered	Information & Answers	4	2006-10-04 22:38

2021-01-14, 21:32	#2687
Xyzzy "Mike" Aug 2002 8056₁₀ Posts	https://www.mersenneforum.org/showpo...&postcount=119