mersenneforum.org  

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

Reply
 
Thread Tools
Old 2021-05-20, 17:00   #1156
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

13×367 Posts
Default After 5,000 Colab GPU attempts.

This started about the new year.
Comparing the first 1,000 you can see there have been a lot more K80s recently.

It doesn't show here but in the last 1,000 attempts 640 were unique; including 380 unique for T4s out of 594.
That tells me either I'm still a ways short of seeing all GPUs colab has ... or they have been replacing them regularly: probably a little of each.

Code:

TYPE:  TOT  UNIQ
ALL:  5000  3658
T4:   4182  2889
P100:  297   281
P4:    145   129
K80:   376   359

ALL:  1000   904
T4:    927   832
P100:   39    39
P4:     27    26
K80:     7     7
petrw1 is online now   Reply With Quote
Old 2021-05-20, 18:41   #1157
moebius
 
moebius's Avatar
 
Jul 2009
Germany

2·313 Posts
Default

Have you seen these before? I used them until just 5 minutes ago.
GPU 0: Tesla K80 (UUID: GPU-1e451911-17ec-e582-a61f-fbbaf862d9b5)

Last fiddled with by moebius on 2021-05-20 at 18:42
moebius is offline   Reply With Quote
Old 2021-05-20, 19:47   #1158
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

13×367 Posts
Default

Quote:
Originally Posted by moebius View Post
Have you seen these before? I used them until just 5 minutes ago.
GPU 0: Tesla K80 (UUID: GPU-1e451911-17ec-e582-a61f-fbbaf862d9b5)
Lately I get almost half K80 and almost half T4....with sporadic P4 and P100
petrw1 is online now   Reply With Quote
Old 2021-07-21, 18:11   #1159
moebius
 
moebius's Avatar
 
Jul 2009
Germany

11628 Posts
Default

Since one month I only be able to get T4/P4 or sometimes K80 instances (5-6 hours) at free google colab. Are the good times at colab over? Are P100 runtimes still available for other free users?

Last fiddled with by moebius on 2021-07-21 at 18:21
moebius is offline   Reply With Quote
Old 2021-07-21, 18:52   #1160
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

16AD16 Posts
Default

Count your blessings. Almost all T4 here if getting a gpu; rarely a K80. Usually one gpu session/account/day; gpu sessions often 3.5-4 hr, sometimes ~1.5 hours. CPU-only 6.5-7.1 hours typical.

Last fiddled with by kriesel on 2021-07-21 at 18:56
kriesel is offline   Reply With Quote
Old 2021-07-21, 19:14   #1161
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

13·367 Posts
Default

Quote:
Originally Posted by moebius View Post
Since one month I only be able to get T4/P4 or sometimes K80 instances (5-6 hours) at free google colab. Are the good times at colab over? Are P100 runtimes still available for other free users?
No P100 for me for several weeks.
And lately very few K80.
About 98% T4.
petrw1 is online now   Reply With Quote
Old 2021-07-22, 01:36   #1162
slandrum
 
Jan 2021
California

2·3·37 Posts
Default

Quote:
Originally Posted by kriesel View Post
Count your blessings. Almost all T4 here if getting a gpu; rarely a K80. Usually one gpu session/account/day; gpu sessions often 3.5-4 hr, sometimes ~1.5 hours. CPU-only 6.5-7.1 hours typical.
I find that the CPU gives you 12 hours/session IF you respond to an "are you human" request that appears some random time from 6 hours to 6:40 into the session. If you don't respond to that, your session is killed some time later (rarely lasting longer than 7 hours). If you do respond, the session will generally be killed after the 12 hour mark (for me it's generally a few seconds to about 5 minutes after the 12 hour mark). A new CPU session can be restarted immediately after the last one ends.

If the GPU session is going to last longer than 3 hours, shortly after 3 hours I get an "are you human" request. If I respond to that the session may last longer than 4 hours, and sometimes (very rarely) I've had a much longer GPU session. If I don't respond to the "are you human" prompt, then the session gets killed 3:45 to 3:50 (if it doesn't stop abruptly before that).

Sometimes I get just over an hour on the GPU, but I average around 3 hours.

For a while now I've only been getting T4 and the occasional K80 for GPU.

Sometimes an account can't get a GPU for 7-14 days, then one day it gets the GPU again with a very short session (maybe 65 minutes), and then the sessions get longer again after that it until it's back in with the normal mix of times I get on GPU sessions.
slandrum is offline   Reply With Quote
Old 2021-08-09, 02:52   #1163
Dylan14
 
Dylan14's Avatar
 
"Dylan"
Mar 2017

11128 Posts
Default

Been a long while since I did stuff with the tf1G stuff, but I have a new version of the tf1G.py script. This fixes a bug where mfaktx would enter a selftest loop if the worktodo.txt file is 1 byte in size (which can sometimes happen on my Ubuntu 20.04 build). On Windows this should have no effect.
Attached Files
File Type: txt tf1G.py.txt (15.0 KB, 27 views)
Dylan14 is offline   Reply With Quote
Old 2021-09-08, 01:25   #1164
moebius
 
moebius's Avatar
 
Jul 2009
Germany

2·313 Posts
Default

I now only do P-1 with colab, because the Tesla K80 GPUs appears too sparse at me.
When the free A100 instances are introduced at colab, someone will be happy to let me know.
Code:
2021-09-08 00:50:06 gpuowl v6.11-380-g79ea0cc
2021-09-08 00:50:06 Note: not found 'config.txt'
2021-09-08 00:50:06 config: -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 -block 100 -maxAlloc 850 -B1 750000 -rB2 20 
2021-09-08 00:50:06 device 0, unique id ''
2021-09-08 00:50:06 Tesla K80-0 106326947 FFT: 5.50M 1K:11:256 (18.44 bpw)
2021-09-08 00:50:06 Tesla K80-0 Expected maximum carry32: 5F500000
2021-09-08 00:50:07 Tesla K80-0 OpenCL args "-DEXP=106326947u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=11u -DPM1=1 -DCARRYM64=1 -DMM_CHAIN=1u -DMM2_CHAIN=2u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0x1.e937bfe53e1fcp-2 -DIWEIGHT_STEP_MINUS_1=-0x1.4b0e3250e5b22p-2 -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 "
2021-09-08 00:50:09 Tesla K80-0 

2021-09-08 00:50:09 Tesla K80-0 OpenCL compilation in 2.77 s
2021-09-08 00:50:10 Tesla K80-0 106326947 P1 B1=750000, B2=15000000; 1081625 bits; starting at 0
2021-09-08 00:50:47 Tesla K80-0 106326947 P1    10000   0.92%; 3787 us/it; ETA 0d 01:08; 7cb5d0b9b8af9639
2021-09-08 00:51:25 Tesla K80-0 106326947 P1    20000   1.85%; 3799 us/it; ETA 0d 01:07; 9affafff93c7d610
2021-09-08 00:52:03 Tesla K80-0 106326947 P1    30000   2.77%; 3798 us/it; ETA 0d 01:07; 88bbfd09f75881bd
2021-09-08 00:52:41 Tesla K80-0 106326947 P1    40000   3.70%; 3798 us/it; ETA 0d 01:06; c01dfe3472d8ba82
2021-09-08 00:53:19 Tesla K80-0 106326947 P1    50000   4.62%; 3797 us/it; ETA 0d 01:05; 45ac7a38db0b3eff
2021-09-08 00:53:57 Tesla K80-0 106326947 P1    60000   5.55%; 3798 us/it; ETA 0d 01:05; af4b53b6365b5136
2021-09-08 00:54:35 Tesla K80-0 106326947 P1    70000   6.47%; 3798 us/it; ETA 0d 01:04; 65d412eb1267f099
2021-09-08 00:55:10 Tesla K80-0 saved
2021-09-08 00:55:13 Tesla K80-0 106326947 P1    80000   7.40%; 3811 us/it; ETA 0d 01:04; 270dfe846d0df926
2021-09-08 00:55:51 Tesla K80-0 106326947 P1    90000   8.32%; 3798 us/it; ETA 0d 01:03; fe3abd04ee4dfca8
2021-09-08 00:56:29 Tesla K80-0 106326947 P1   100000   9.25%; 3797 us/it; ETA 0d 01:02; f5c61b55135ad1ba
2021-09-08 00:57:07 Tesla K80-0 106326947 P1   110000  10.17%; 3797 us/it; ETA 0d 01:01; be6c6a7ef3e49586
2021-09-08 00:57:45 Tesla K80-0 106326947 P1   120000  11.09%; 3797 us/it; ETA 0d 01:01; 208edfd77b5f593f
2021-09-08 00:58:23 Tesla K80-0 106326947 P1   130000  12.02%; 3796 us/it; ETA 0d 01:00; aee327860c605fd1
2021-09-08 00:59:01 Tesla K80-0 106326947 P1   140000  12.94%; 3796 us/it; ETA 0d 01:00; 981437fea2587d6c
2021-09-08 00:59:39 Tesla K80-0 106326947 P1   150000  13.87%; 3797 us/it; ETA 0d 00:59; 53f1b041f83663ee
2021-09-08 01:00:11 Tesla K80-0 saved
moebius is offline   Reply With Quote
Old 2021-09-08, 03:36   #1165
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

13·367 Posts
Default

Quote:
Originally Posted by moebius View Post
I now only do P-1 with colab, because the Tesla K80 GPUs appears too sparse at me.
When the free A100 instances are introduced at colab, someone will be happy to let me know.
For the last week or so I get K80's every time.
petrw1 is online now   Reply With Quote
Old 2021-09-16, 17:20   #1166
axn
 
axn's Avatar
 
Jun 2003

3·17·101 Posts
Default

Anyone having trouble opening colab notebooks? Not able to open notebooks, not even the colab intro notebook.

EDIT:- Nevermind. Looks like ISP issue. Switched to my mobile hotspot and it is going thru.

Last fiddled with by axn on 2021-09-16 at 17:27
axn is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Alternatives to Google Colab kriesel Cloud Computing 11 2020-01-14 18:45
Notebook enzocreti enzocreti 0 2019-02-15 08:20
Computer Diet causes Machine Check Exception -- need heuristics help Christenson Hardware 32 2011-12-25 08:17
Computer diet - Need help garo Hardware 41 2011-10-06 04:06
Workunit diet ? dsouza123 NFSNET Discussion 5 2004-02-27 00:42

All times are UTC. The time now is 18:54.


Mon Oct 25 18:54:56 UTC 2021 up 94 days, 13:23, 0 users, load averages: 2.36, 1.94, 1.70

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.