mersenneforum.org  

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

Reply
 
Thread Tools
Old 2021-12-07, 06:14   #2740
Viliam Furik
 
Viliam Furik's Avatar
 
"Viliam Furík"
Jul 2018
Martin, Slovakia

13528 Posts
Default

Quote:
Originally Posted by Xyzzy View Post
6900XT (XTXH)
Could you test it on v6.11? It tends to have a bit better throughput.

Last fiddled with by Viliam Furik on 2021-12-07 at 06:15
Viliam Furik is offline   Reply With Quote
Old 2021-12-07, 14:31   #2741
Xyzzy
 
Xyzzy's Avatar
 
Aug 2002

23×1,051 Posts
Default

Quote:
Originally Posted by Viliam Furik View Post
Could you test it on v6.11? It tends to have a bit better throughput.
Code:
2021-12-07 08:20:10 gpuowl v6.11-380-g79ea0cc
2021-12-07 08:20:10 Note: not found 'config.txt'
2021-12-07 08:20:10 config: -prp 77936867
2021-12-07 08:20:10 device 0, unique id ''
2021-12-07 08:20:10 gfx1030-0 77936867 FFT: 4M 1K:8:256 (18.58 bpw)
2021-12-07 08:20:10 gfx1030-0 Expected maximum carry32: 583B0000
2021-12-07 08:20:11 gfx1030-0 OpenCL args "-DEXP=77936867u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=8u -DPM1=0 -DAMDGPU=1 -DMM_CHAIN=1u -DMM2_CHAIN=2u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0xa.c42d0d7cec038p-5 -DIWEIGHT_STEP_MINUS_1=-0x8.0e50c8817ddf8p-5  -cl-unsafe-math-optimizations -cl-std=CL2.0 -cl-finite-math-only "
2021-12-07 08:20:12 gfx1030-0 OpenCL compilation in 1.69 s
2021-12-07 08:20:13 gfx1030-0 77936867 OK        0 loaded: blockSize 400, 0000000000000003
2021-12-07 08:20:13 gfx1030-0 validating proof residues for power 8
2021-12-07 08:20:13 gfx1030-0 Proof using power 8
2021-12-07 08:20:13 gfx1030-0 77936867 OK      800   0.00%;  594 us/it; ETA 0d 12:52; 1579c241dc63eca6 (check 0.27s)
2021-12-07 08:22:13 gfx1030-0 77936867 OK   200000   0.26%;  599 us/it; ETA 0d 12:56; f0b04b45b0855bd2 (check 0.28s)
2021-12-07 08:24:13 gfx1030-0 77936867 OK   400000   0.51%;  601 us/it; ETA 0d 12:57; c03f94396a5aa29e (check 0.28s)
2021-12-07 08:26:14 gfx1030-0 77936867 OK   600000   0.77%;  601 us/it; ETA 0d 12:55; b9decd65ca71b629 (check 0.28s)
2021-12-07 08:28:14 gfx1030-0 77936867 OK   800000   1.03%;  601 us/it; ETA 0d 12:53; 21ebf3636148f663 (check 0.28s)
2021-12-07 08:30:15 gfx1030-0 77936867 OK  1000000   1.28%;  601 us/it; ETA 0d 12:51; 9bf9d9e6bff4286e (check 0.28s)
Xyzzy is offline   Reply With Quote
Old 2021-12-07, 17:07   #2742
tServo
 
tServo's Avatar
 
"Marv"
May 2009
near the Tannhäuser Gate

10110010002 Posts
Default

Quote:
Originally Posted by kriesel View Post
Yes, there was a lot of discussion and benchmark posting a while back (beginning ~https://www.mersenneforum.org/showpo...postcount=1528, going up to V6.11-2xx days I think).
A detailed multiple-version-specific treatise/tutorial? No. There were a great many commits, in V6.11 through v7.2, and along the way, some of the tune -use options best choices got incorporated into the code as defaults for AMD GPUs or for NVIDIA GPUs or became obsolete. One easy way to tune is to look at already collected benchmarks, and select a gpuowl version that does well for what you want to do.
There's also a little performance variables guidance.

What are you planning to run, on what hardware, and what parameters concern or interest you?

Thanks.
I am running Gpuowl v6.11 380 at the FPT wavefront and some DC with proofs.
My Gpus are Titan V and Radeon Vii on windoze 10.
Yes, I know linux would be faster but linux and I just don't get along.
CARRY32 looks interesting.
DrDerpenberg's post above shows an example of what tuning can do.
tServo is offline   Reply With Quote
Old 2021-12-08, 02:28   #2743
Viliam Furik
 
Viliam Furik's Avatar
 
"Viliam Furík"
Jul 2018
Martin, Slovakia

2·373 Posts
Default

Quote:
Originally Posted by Xyzzy View Post
Code:
2021-12-07 08:30:15 gfx1030-0 77936867 OK  1000000   1.28%;  601 us/it; ETA 0d 12:51; 9bf9d9e6bff4286e (check 0.28s)
Thanks!

The results have worse throughput, which doesn't match my past experience. But I guess that's even better.
Viliam Furik is offline   Reply With Quote
Old 2021-12-08, 10:09   #2744
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
"name field"
Jun 2011
Thailand

71×139 Posts
Default

@Mike: Any benchmarks sent to James?
LaurV is offline   Reply With Quote
Old 2021-12-22, 06:44   #2745
xx005fs
 
"Eric"
Jan 2018
USA

5×43 Posts
Default Newest gpuowl version (v7.2-86-gddf3314) performance regression

I am seeing a significant performance reduction using the newest revision of gpuowl compared to 6.11 on my Titan V, while also seeing a slight drop in performance with my Vega 56.

All tests were done what I found to be the fastest -use flags through my short testing and it does seem to be as fast as I can get them to be. I tested them all on wavefront exponents using 6M FFT.

Titan V using 6.11:
Code:
2021-12-20 21:30:09 Note: not found 'config.txt'
2021-12-20 21:30:09 config: -device 0 -carry short -use CARRY32,ORIG_SLOWTRIG,IN_WG=128,IN_SIZEX=16,IN_SPACING=4,OUT_WG=128,OUT_SIZEX=16,OUT_SPACING=4 -nospin -maxAlloc 10000 -B1 750000 -rB2 20 
2021-12-20 21:30:09 device 0, unique id ''
2021-12-20 21:30:09 NVIDIA TITAN V-0 115504451 FFT: 6M 1K:12:256 (18.36 bpw)
2021-12-20 21:30:09 NVIDIA TITAN V-0 Expected maximum carry32: 5ECF0000
2021-12-20 21:30:09 NVIDIA TITAN V-0 OpenCL args "-DEXP=115504451u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=12u -DPM1=0 -DMM_CHAIN=1u -DMM2_CHAIN=2u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0x8.f39dc6dc86f2p-4 -DIWEIGHT_STEP_MINUS_1=-0xb.7af4a364cb47p-5 -DCARRY32=1 -DIN_SIZEX=16 -DIN_SPACING=4 -DIN_WG=128 -DORIG_SLOWTRIG=1 -DOUT_SIZEX=16 -DOUT_SPACING=4 -DOUT_WG=128  -cl-unsafe-math-optimizations -cl-std=CL2.0 -cl-finite-math-only "
2021-12-20 21:30:09 NVIDIA TITAN V-0 

2021-12-20 21:30:09 NVIDIA TITAN V-0 OpenCL compilation in 0.06 s
2021-12-20 21:30:10 NVIDIA TITAN V-0 115504451 OK 20006400 loaded: blockSize 400, e23f0b6cd8d3813c
2021-12-20 21:30:10 NVIDIA TITAN V-0 validating proof residues for power 8
2021-12-20 21:30:13 NVIDIA TITAN V-0 Proof using power 8
2021-12-20 21:30:14 NVIDIA TITAN V-0 115504451 OK 20007200  17.32%; 1017 us/it; ETA 1d 02:58; 78071386ef269ffa (check 0.40s)
2021-12-20 21:32:30 NVIDIA TITAN V-0 115504451 OK 20200000  17.49%;  699 us/it; ETA 0d 18:30; 99b150da9675bdb7 (check 0.36s)
2021-12-20 21:34:48 NVIDIA TITAN V-0 115504451 OK 20400000  17.66%;  691 us/it; ETA 0d 18:16; d9b1eddb41c96e32 (check 0.37s)
2021-12-20 21:37:07 NVIDIA TITAN V-0 115504451 OK 20600000  17.83%;  692 us/it; ETA 0d 18:15; 79f93065f22e5dae (check 0.38s)
2021-12-20 21:39:26 NVIDIA TITAN V-0 115504451 OK 20800000  18.01%;  694 us/it; ETA 0d 18:16; 529b3a62d9a7908b (check 0.60s)
Titan V using newest version:
Code:
20211221 10:37:08 GpuOwl VERSION v7.2-86-gddf3314
20211221 10:37:08 Note: not found 'config.txt'
20211221 10:37:08 config: -device 0 -nospin -block 200 -maxAlloc 10000 -use CARRY32,NEWEST_FFT5,NEWEST_FFT8,OUT_WG=128,OUT_SIZEX=16,OUT_SPACING=4,IN_WG=128,IN_SIZEX=16,IN_SPACING=4,STATS 
20211221 10:37:08 device 0, unique id ''
20211221 10:37:08 NVIDIA TITAN V-0 115504451 FFT: 6M 1K:12:256 (18.36 bpw)
20211221 10:37:08 NVIDIA TITAN V-0 115504451 OpenCL args "-DEXP=115504451u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=12u -DMM_CHAIN=1u -DMM2_CHAIN=2u -DMAX_ACCURACY=1 -DWEIGHT_STEP=0.55947663955924698 -DIWEIGHT_STEP=-0.35875923073613414 -DIWEIGHTS={0,-0.1776205516677711,-0.32369204296077886,-0.44381823538738857,-0.085215094490870003,-0.24769969406475154,-0.38132368942480332,-0.49121331701295107,} -DFWEIGHTS={0,0.21598369466550077,0.47861634569236183,0.79797336702779942,0.093153148874317013,0.32925640480341822,0.61635411427064102,0.96546024775859707,} -DCARRY32=1 -DIN_SIZEX=16 -DIN_SPACING=4 -DIN_WG=128 -DNEWEST_FFT5=1 -DNEWEST_FFT8=1 -DOUT_SIZEX=16 -DOUT_SPACING=4 -DOUT_WG=128 -DSTATS=1  -cl-std=CL2.0 -cl-finite-math-only "
20211221 10:37:10 NVIDIA TITAN V-0 115504451 

20211221 10:37:10 NVIDIA TITAN V-0 115504451 OpenCL compilation in 1.49 s
20211221 10:37:10 NVIDIA TITAN V-0 115504451 maxAlloc: 9.8 GB
20211221 10:37:10 NVIDIA TITAN V-0 115504451 P1(0) 0 bits
20211221 10:37:11 NVIDIA TITAN V-0 115504451 OK    165800 on-load: blockSize 200, 4296258b82bdfc8c
20211221 10:37:11 NVIDIA TITAN V-0 115504451 validating proof residues for power 8
20211221 10:37:11 NVIDIA TITAN V-0 115504451 Proof using power 8
20211221 10:37:12 NVIDIA TITAN V-0 115504451 OK    166200   0.14% 92e2bc7084d61ca0 1123 us/it + check 0.30s + save 0.17s; ETA 1d 11:58
20211221 10:37:16 NVIDIA TITAN V-0 115504451    170000 5c11f6dd2b06d4d8 1131
20211221 10:37:27 NVIDIA TITAN V-0 115504451    180000 f81b5aef0616a108 1135
20211221 10:37:39 NVIDIA TITAN V-0 115504451    190000 e9c06dbbcc9f9573 1136
20211221 10:37:50 NVIDIA TITAN V-0 115504451 Roundoff: N=34169, mean 0.227767, SD 0.013053, CV 0.057309, max 0.322290, z 20.9 (pErr 0.015669%)
20211221 10:37:50 NVIDIA TITAN V-0 115504451 Carry: N=34168, max 563601ff, avg 3f7eb281; CarryM: N=1, max bbffb5ba, avg bbffb5ba
20211221 10:37:50 NVIDIA TITAN V-0 115504451 OK    200000   0.17% e52a80d2d6bba924 1142 us/it + check 0.31s + save 0.18s; ETA 1d 12:34
20211221 10:37:54 NVIDIA TITAN V-0 115504451 Stopping, please wait..
20211221 10:37:54 NVIDIA TITAN V-0 115504451 Roundoff: N=3416, mean 0.227717, SD 0.013281, CV 0.058323, max 0.310540, z 20.5 (pErr 0.024684%)
20211221 10:37:54 NVIDIA TITAN V-0 115504451 Carry: N=3415, max 4f342326, avg 3f6a0f6c; CarryM: N=1, max bbfc8853, avg bbfc8853
20211221 10:37:55 NVIDIA TITAN V-0 115504451 OK    203200   0.18% 47d1047eb3fe9511 1152 us/it + check 0.31s + save 0.26s; ETA 1d 12:54
20211221 10:37:55 NVIDIA TITAN V-0 Exiting because "stop requested"
20211221 10:37:55 NVIDIA TITAN V-0 Bye
Vega 56 using 6.11:
Code:
2021-12-20 22:41:20 Note: not found 'config.txt'
2021-12-20 22:41:20 config: -device 1 -carry short -nospin -block 400 -B1 750000 -rB2 20 
2021-12-20 22:41:20 device 1, unique id ''
2021-12-20 22:41:21 gfx900-1 115504531 FFT: 6M 1K:12:256 (18.36 bpw)
2021-12-20 22:41:21 gfx900-1 Expected maximum carry32: 5ECF0000
2021-12-20 22:41:21 gfx900-1 OpenCL args "-DEXP=115504531u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=12u -DPM1=0 -DAMDGPU=1 -DMM_CHAIN=1u -DMM2_CHAIN=2u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0x8.f38f5d3eee8d8p-4 -DIWEIGHT_STEP_MINUS_1=-0xb.7ae8c91a5a05p-5  -cl-unsafe-math-optimizations -cl-std=CL2.0 -cl-finite-math-only "
2021-12-20 22:41:21 gfx900-1 ASM compilation failed, retrying compilation using NO_ASM
2021-12-20 22:41:27 gfx900-1 OpenCL compilation in 5.57 s
2021-12-20 22:41:28 gfx900-1 115504531 OK 22400000 loaded: blockSize 400, 3096ca59db298cb1
2021-12-20 22:41:28 gfx900-1 validating proof residues for power 8
2021-12-20 22:41:32 gfx900-1 Proof using power 8
2021-12-20 22:41:35 gfx900-1 115504531 OK 22400800  19.39%; 2320 us/it; ETA 2d 12:00; 68544da0fa47c0da (check 1.04s)
2021-12-20 22:49:17 gfx900-1 115504531 OK 22600000  19.57%; 2315 us/it; ETA 2d 11:45; a8329bd6e6811c15 (check 1.05s)
2021-12-20 22:57:01 gfx900-1 115504531 OK 22800000  19.74%; 2314 us/it; ETA 2d 11:35; 683b935ead2e5e12 (check 1.05s)
2021-12-20 23:04:45 gfx900-1 115504531 OK 23000000  19.91%; 2314 us/it; ETA 2d 11:28; cd0d93c477364696 (check 1.04s)
Vega 56 using newest version:
Code:
20211221 10:43:37 GpuOwl VERSION v7.2-86-gddf3314
20211221 10:43:37 Note: not found 'config.txt'
20211221 10:43:37 config: -device 1 -nospin -block 200 -maxAlloc 10000 -use CARRY32,NEWEST_FFT5,NEWEST_FFT8,OUT_WG=128,OUT_SIZEX=16,OUT_SPACING=4,IN_WG=128,IN_SIZEX=16,IN_SPACING=4,STATS 
20211221 10:43:38 device 1, unique id ''
20211221 10:43:38 gfx900-1 115504451 FFT: 6M 1K:12:256 (18.36 bpw)
20211221 10:43:38 gfx900-1 115504451 OpenCL args "-DEXP=115504451u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=12u -DAMDGPU=1 -DMM_CHAIN=1u -DMM2_CHAIN=2u -DMAX_ACCURACY=1 -DWEIGHT_STEP=0.55947663955924698 -DIWEIGHT_STEP=-0.35875923073613414 -DIWEIGHTS={0,-0.1776205516677711,-0.32369204296077886,-0.44381823538738857,-0.085215094490870003,-0.24769969406475154,-0.38132368942480332,-0.49121331701295107,} -DFWEIGHTS={0,0.21598369466550077,0.47861634569236183,0.79797336702779942,0.093153148874317013,0.32925640480341822,0.61635411427064102,0.96546024775859707,} -DCARRY32=1 -DIN_SIZEX=16 -DIN_SPACING=4 -DIN_WG=128 -DNEWEST_FFT5=1 -DNEWEST_FFT8=1 -DOUT_SIZEX=16 -DOUT_SPACING=4 -DOUT_WG=128 -DSTATS=1  -cl-std=CL2.0 -cl-finite-math-only "
20211221 10:43:38 gfx900-1 115504451 ASM compilation failed, retrying compilation using NO_ASM
20211221 10:43:42 gfx900-1 115504451 OpenCL compilation in 4.63 s
20211221 10:43:43 gfx900-1 115504451 maxAlloc: 9.8 GB
20211221 10:43:43 gfx900-1 115504451 P1(0) 0 bits
20211221 10:43:44 gfx900-1 115504451 OK    266600 on-load: blockSize 200, bf34b4f7911a86b8
20211221 10:43:44 gfx900-1 115504451 validating proof residues for power 8
20211221 10:43:44 gfx900-1 115504451 Proof using power 8
20211221 10:43:46 gfx900-1 115504451 OK    267000   0.23% f8e4ff2b2fb857b5 2424 us/it + check 0.56s + save 0.19s; ETA 3d 05:36
20211221 10:43:53 gfx900-1 115504451    270000 38ebb1487e722f8b 2417
20211221 10:44:17 gfx900-1 115504451    280000 46fa05ed8132d4e3 2417
20211221 10:44:41 gfx900-1 115504451    290000 d9203a046d0dd6ba 2417
20211221 10:44:46 gfx900-1 115504451 Stopping, please wait..
20211221 10:44:46 gfx900-1 115504451 Roundoff: N=25526, mean 0.238955, SD 0.013561, CV 0.056753, max 0.355965, z 19.2 (pErr 0.122980%)
20211221 10:44:46 gfx900-1 115504451 Carry: N=25525, max 56790235, avg 3f73c614; CarryM: N=1, max c1bedaf1, avg c1bedaf1
20211221 10:44:47 gfx900-1 115504451 OK    292200   0.25% d5b81b2756ecbab2 2428 us/it + check 0.56s + save 0.18s; ETA 3d 05:42
20211221 10:44:47 gfx900-1 Exiting because "stop requested"
20211221 10:44:47 gfx900-1 Bye
xx005fs is offline   Reply With Quote
Old 2021-12-22, 06:55   #2746
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

41×149 Posts
Default

Quote:
Originally Posted by xx005fs View Post
I am seeing a significant performance reduction using the newest revision of gpuowl compared to 6.11 on my Titan V, while also seeing a slight drop in performance with my Vega 56.
Yes, see also https://mersenneforum.org/showpost.p...&postcount=197 re RadeonVII.
kriesel is offline   Reply With Quote
Old 2021-12-23, 04:18   #2747
xx005fs
 
"Eric"
Jan 2018
USA

D716 Posts
Default v6.11 Error

Is there any reason why an exponent would error out after a bit of running on Nvidia GPUs?

Code:
2021-12-22 16:37:40 gpuowl v6.11-364-g36f4e2a
2021-12-22 16:37:40 Note: not found 'config.txt'
2021-12-22 16:37:40 config: -device 0 -carry short -use CARRY32,ORIG_SLOWTRIG,IN_WG=128,IN_SIZEX=16,IN_SPACING=4,OUT_WG=128,OUT_SIZEX=16,OUT_SPACING=4 -nospin -maxAlloc 10000 -B1 750000 -rB2 20
2021-12-22 16:37:40 device 0, unique id ''
2021-12-22 16:37:40 NVIDIA TITAN V-0 115504457 FFT: 6M 1K:12:256 (18.36 bpw)
2021-12-22 16:37:40 NVIDIA TITAN V-0 Expected maximum carry32: 5ECF0000
2021-12-22 16:37:40 NVIDIA TITAN V-0 OpenCL args "-DEXP=115504457u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=12u -DPM1=0 -DMM_CHAIN=1u -DMM2_CHAIN=2u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0x8.f39cb2239e64p-4 -DIWEIGHT_STEP_MINUS_1=-0xb.7af3bfd2a5fa8p-5 -DCARRY32=1 -DIN_SIZEX=16 -DIN_SPACING=4 -DIN_WG=128 -DORIG_SLOWTRIG=1 -DOUT_SIZEX=16 -DOUT_SPACING=4 -DOUT_WG=128  -cl-unsafe-math-optimizations -cl-std=CL2.0 -cl-finite-math-only "
2021-12-22 16:37:40 NVIDIA TITAN V-0

2021-12-22 16:37:40 NVIDIA TITAN V-0 OpenCL compilation in 0.04 s
2021-12-22 16:37:41 NVIDIA TITAN V-0 115504457 OK 57405200 loaded: blockSize 400, d43727673128cc9e
2021-12-22 16:37:41 NVIDIA TITAN V-0 validating proof residues for power 8
2021-12-22 16:37:51 NVIDIA TITAN V-0 Proof using power 8
2021-12-22 16:37:52 NVIDIA TITAN V-0 115504457 OK 57406000  49.70%;  690 us/it; ETA 0d 11:08; 243e0adeedd63100 (check 0.34s)
2021-12-22 16:40:06 NVIDIA TITAN V-0 115504457 OK 57600000  49.87%;  689 us/it; ETA 0d 11:05; 595f9abe118390a5 (check 0.53s)
2021-12-22 16:42:26 NVIDIA TITAN V-0 115504457 OK 57800000  50.04%;  697 us/it; ETA 0d 11:10; 75c04091a029af74 (check 0.35s)
2021-12-22 16:44:51 NVIDIA TITAN V-0 115504457 OK 58000000  50.21%;  722 us/it; ETA 0d 11:32; 08e568ca391a7693 (check 0.35s)
2021-12-22 16:47:16 NVIDIA TITAN V-0 115504457 OK 58200000  50.39%;  728 us/it; ETA 0d 11:35; 86e0d3a2aef00688 (check 0.35s)
2021-12-22 16:49:43 NVIDIA TITAN V-0 115504457 OK 58400000  50.56%;  732 us/it; ETA 0d 11:36; d75315c6fcda4af5 (check 0.35s)
2021-12-22 16:52:10 NVIDIA TITAN V-0 115504457 OK 58600000  50.73%;  733 us/it; ETA 0d 11:36; 859b48b3e739a4cc (check 0.35s)
2021-12-22 16:54:38 NVIDIA TITAN V-0 115504457 OK 58800000  50.91%;  735 us/it; ETA 0d 11:35; ddbfbf24923805e5 (check 0.36s)
2021-12-22 16:57:05 NVIDIA TITAN V-0 115504457 OK 59000000  51.08%;  735 us/it; ETA 0d 11:32; 49f900d751d335c3 (check 0.35s)
2021-12-22 16:59:33 NVIDIA TITAN V-0 115504457 OK 59200000  51.25%;  737 us/it; ETA 0d 11:31; 638df61a1ae4aa8a (check 0.37s)
2021-12-22 17:02:00 NVIDIA TITAN V-0 115504457 OK 59400000  51.43%;  737 us/it; ETA 0d 11:29; 4f61a8a357443cb0 (check 0.35s)
2021-12-22 17:04:28 NVIDIA TITAN V-0 115504457 OK 59600000  51.60%;  738 us/it; ETA 0d 11:27; 5db27dfceb347b2d (check 0.37s)
2021-12-22 17:06:56 NVIDIA TITAN V-0 115504457 OK 59800000  51.77%;  739 us/it; ETA 0d 11:26; 25646a5903292d7c (check 0.35s)
2021-12-22 17:09:25 NVIDIA TITAN V-0 115504457 OK 60000000  51.95%;  742 us/it; ETA 0d 11:26; dad9657681d3751d (check 0.37s)
2021-12-22 17:11:57 NVIDIA TITAN V-0 115504457 OK 60200000  52.12%;  756 us/it; ETA 0d 11:37; fd177638070b0856 (check 0.37s)
2021-12-22 17:14:29 NVIDIA TITAN V-0 115504457 OK 60400000  52.29%;  760 us/it; ETA 0d 11:38; b6183c966a9c80c0 (check 0.38s)
2021-12-22 17:17:02 NVIDIA TITAN V-0 115504457 OK 60600000  52.47%;  764 us/it; ETA 0d 11:39; ba7471ff93d9c03d (check 0.39s)
2021-12-22 17:19:35 NVIDIA TITAN V-0 115504457 OK 60800000  52.64%;  765 us/it; ETA 0d 11:37; be56ea487ae74250 (check 0.38s)
2021-12-22 17:22:09 NVIDIA TITAN V-0 115504457 OK 61000000  52.81%;  766 us/it; ETA 0d 11:36; 4c329dee8db4b643 (check 0.37s)
2021-12-22 17:24:42 NVIDIA TITAN V-0 115504457 OK 61200000  52.98%;  766 us/it; ETA 0d 11:33; 738cd85c70e5bd49 (check 0.38s)
2021-12-22 17:27:16 NVIDIA TITAN V-0 115504457 OK 61400000  53.16%;  767 us/it; ETA 0d 11:31; 7e3c94d5bef33bde (check 0.58s)
2021-12-22 17:29:50 NVIDIA TITAN V-0 115504457 OK 61600000  53.33%;  764 us/it; ETA 0d 11:27; 692b651b7ad774b5 (check 0.38s)
2021-12-22 17:32:23 NVIDIA TITAN V-0 115504457 OK 61800000  53.50%;  764 us/it; ETA 0d 11:24; 29121cae63c8e851 (check 0.38s)
2021-12-22 17:34:56 NVIDIA TITAN V-0 115504457 OK 62000000  53.68%;  765 us/it; ETA 0d 11:22; ba96699741fd264e (check 0.38s)
2021-12-22 17:37:29 NVIDIA TITAN V-0 115504457 OK 62200000  53.85%;  764 us/it; ETA 0d 11:19; be6e6c3e1a405ce7 (check 0.38s)
2021-12-22 17:40:03 NVIDIA TITAN V-0 115504457 OK 62400000  54.02%;  765 us/it; ETA 0d 11:17; 760c6e2e16483061 (check 0.38s)
2021-12-22 17:42:36 NVIDIA TITAN V-0 115504457 OK 62600000  54.20%;  765 us/it; ETA 0d 11:15; 3175ce9938be2ac5 (check 0.38s)
2021-12-22 17:45:10 NVIDIA TITAN V-0 115504457 OK 62800000  54.37%;  768 us/it; ETA 0d 11:15; 929d73ee83f9e9e6 (check 0.38s)
2021-12-22 17:47:45 NVIDIA TITAN V-0 115504457 EE 63000000  54.54%;  770 us/it; ETA 0d 11:14; efcf198910c9b2f7 (check 0.36s)
2021-12-22 17:47:45 NVIDIA TITAN V-0 115504457 OK 62800000 loaded: blockSize 400, 929d73ee83f9e9e6
2021-12-22 17:49:02 NVIDIA TITAN V-0 115504457 EE 62900000  54.46%;  770 us/it; ETA 0d 11:15; 23671499502976e2 (check 0.37s) 1 errors
2021-12-22 17:49:03 NVIDIA TITAN V-0 115504457 OK 62800000 loaded: blockSize 400, 929d73ee83f9e9e6
2021-12-22 17:49:42 NVIDIA TITAN V-0 115504457 OK 62850000  54.41%;  769 us/it; ETA 0d 11:15; 70b38978f7efcc52 (check 0.60s) 2 errors
2021-12-22 17:50:21 NVIDIA TITAN V-0 115504457 EE 62900000  54.46%;  768 us/it; ETA 0d 11:14; 23671499502976e2 (check 0.36s) 2 errors
2021-12-22 17:50:21 NVIDIA TITAN V-0 115504457 OK 62850000 loaded: blockSize 400, 70b38978f7efcc52
2021-12-22 17:51:00 NVIDIA TITAN V-0 115504457 EE 62900000  54.46%;  769 us/it; ETA 0d 11:14; 23671499502976e2 (check 0.36s) 3 errors
2021-12-22 17:51:00 NVIDIA TITAN V-0 115504457 OK 62850000 loaded: blockSize 400, 70b38978f7efcc52
2021-12-22 17:51:39 NVIDIA TITAN V-0 115504457 EE 62900000  54.46%;  768 us/it; ETA 0d 11:14; 23671499502976e2 (check 0.37s) 4 errors
2021-12-22 17:51:39 NVIDIA TITAN V-0 3 sequential errors, will stop.
2021-12-22 17:51:39 NVIDIA TITAN V-0 Exiting because "too many errors"
2021-12-22 17:51:39 NVIDIA TITAN V-0 Bye
Even after completely rerunning the same exponent the error occurs on the same iterations. I am sure that the GPU is running stably. What could be happening here? I'm running CUDA 10.2 and the newest driver. I haven't seen any issue happening to my Vega 56 yet.

Last fiddled with by xx005fs on 2021-12-23 at 04:33
xx005fs is offline   Reply With Quote
Old 2021-12-23, 06:19   #2748
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

777410 Posts
Default

Try forcing a larger FFT size?
Prime95 is online now   Reply With Quote
Old 2021-12-23, 17:13   #2749
xx005fs
 
"Eric"
Jan 2018
USA

5×43 Posts
Default

Quote:
Originally Posted by Prime95 View Post
Try forcing a larger FFT size?
That did fix the issue, but I don't know why it's only happening to the Nvidia GPU.
xx005fs is offline   Reply With Quote
Old 2021-12-24, 01:20   #2750
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

2·132·23 Posts
Default

Quote:
Originally Posted by xx005fs View Post
That did fix the issue, but I don't know why it's only happening to the Nvidia GPU.
I suspect if you run the same exponent on an AMD GPU you would have the same problem. But maybe not as gpuowl chooses different default optimizations for the different architectures.

IIRC, Mihai was pretty aggressive in choosing FFT size. Something like willing to tolerate a 0.5% chance or a bad run. I think I lobbied for at most 0.1%. I don't remember what finally happened. No matter what target we choose, this scenario was going to be a possibility -- even a likelihood.
Prime95 is online now   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 1683 2022-01-22 21:51
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

All times are UTC. The time now is 05:59.


Sun Jan 23 05:59:27 UTC 2022 up 184 days, 28 mins, 0 users, load averages: 0.92, 0.99, 1.14

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, 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.

≠ ± ∓ ÷ × · − √ ‰ ⊗ ⊕ ⊖ ⊘ ⊙ ≤ ≥ ≦ ≧ ≨ ≩ ≺ ≻ ≼ ≽ ⊏ ⊐ ⊑ ⊒ ² ³ °
∠ ∟ ° ≅ ~ ‖ ⟂ ⫛
≡ ≜ ≈ ∝ ∞ ≪ ≫ ⌊⌋ ⌈⌉ ∘ ∏ ∐ ∑ ∧ ∨ ∩ ∪ ⨀ ⊕ ⊗ 𝖕 𝖖 𝖗 ⊲ ⊳
∅ ∖ ∁ ↦ ↣ ∩ ∪ ⊆ ⊂ ⊄ ⊊ ⊇ ⊃ ⊅ ⊋ ⊖ ∈ ∉ ∋ ∌ ℕ ℤ ℚ ℝ ℂ ℵ ℶ ℷ ℸ 𝓟
¬ ∨ ∧ ⊕ → ← ⇒ ⇐ ⇔ ∀ ∃ ∄ ∴ ∵ ⊤ ⊥ ⊢ ⊨ ⫤ ⊣ … ⋯ ⋮ ⋰ ⋱
∫ ∬ ∭ ∮ ∯ ∰ ∇ ∆ δ ∂ ℱ ℒ ℓ
𝛢𝛼 𝛣𝛽 𝛤𝛾 𝛥𝛿 𝛦𝜀𝜖 𝛧𝜁 𝛨𝜂 𝛩𝜃𝜗 𝛪𝜄 𝛫𝜅 𝛬𝜆 𝛭𝜇 𝛮𝜈 𝛯𝜉 𝛰𝜊 𝛱𝜋 𝛲𝜌 𝛴𝜎 𝛵𝜏 𝛶𝜐 𝛷𝜙𝜑 𝛸𝜒 𝛹𝜓 𝛺𝜔