GPU FLOPS: Theory vs Reality

Message boards : Number crunching : GPU FLOPS: Theory vs Reality
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 8 · Next

AuthorMessage
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 10345
Credit: 138,979,954
RAC: 84,159
Australia
Message 1905508 - Posted: 8 Dec 2017, 7:19:02 UTC

I wonder what sort of performance this sucker would give?
GTX 1080Ti CUDA Cores 3584
Titan Xp CUDA Cores 3840
Titan V CUDA Cores 5120

For $3,000US you'd want it to kiss you good night & tuck you in as well as crunch like nothing else...
Grant
Darwin NT
ID: 1905508 · Report as offensive     Reply Quote
Profile Wiggo "Socialist"
Avatar

Send message
Joined: 24 Jan 00
Posts: 14937
Credit: 193,671,415
RAC: 54,210
Australia
Message 1905513 - Posted: 8 Dec 2017, 7:32:15 UTC

It must be time for another update. ;-)

Cheers.
ID: 1905513 · Report as offensive     Reply Quote
Profile Shaggie76
Avatar

Send message
Joined: 9 Oct 09
Posts: 267
Credit: 157,049,361
RAC: 190,225
Canada
Message 1905934 - Posted: 9 Dec 2017, 14:46:44 UTC

After all the server problems last week I didn't want to put strain on them running my script to crawl through the database. Maybe if they stay up this week I'll try again.
ID: 1905934 · Report as offensive     Reply Quote
Profile Shaggie76
Avatar

Send message
Joined: 9 Oct 09
Posts: 267
Credit: 157,049,361
RAC: 190,225
Canada
Message 1909613 - Posted: 30 Dec 2017, 17:15:15 UTC

New scan shows the 1070 Ti's doing a tiny bit more work than regular 1070's but probably use more power than it's worth.

ID: 1909613 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 5571
Credit: 386,928,160
RAC: 1,037,237
United States
Message 1909624 - Posted: 30 Dec 2017, 18:03:25 UTC

I would say that the 1070 Ti isn't any more power hungry than the 1070. Running both at full BOINC load and these are the nvidia-smi outputs.

GTX 1070Ti

keith@Darksider:~$ nvidia-smi
Sat Dec 30 09:53:01 2017       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.98                 Driver Version: 384.98                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 107...  On   | 00000000:03:00.0 Off |                  N/A |
|100%   48C    P2   119W / 180W |   2578MiB /  8114MiB |     95%      Default |
+-------------------------------+----------------------+----------------------+
|   1  GeForce GTX 107...  On   | 00000000:06:00.0  On |                  N/A |
|100%   62C    P2   110W / 180W |   2897MiB /  8110MiB |     99%      Default |
+-------------------------------+----------------------+----------------------+
|   2  GeForce GTX 107...  On   | 00000000:07:00.0 Off |                  N/A |
|100%   66C    P2   124W / 180W |   2580MiB /  8114MiB |     98%      Default |
+-------------------------------+----------------------+----------------------+


And this the GTX 1070

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi
Sat Dec 30 09:54:36 2017
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.94                 Driver Version: 384.94                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 1070   WDDM  | 00000000:01:00.0  On |                  N/A |
|100%   70C    P2   115W / 166W |   2449MiB /  8192MiB |     92%      Default |
+-------------------------------+----------------------+----------------------+
|   1  GeForce GTX 1070   WDDM  | 00000000:06:00.0 Off |                  N/A |
| 84%   64C    P2   108W / 166W |   2366MiB /  8192MiB |     94%      Default |
+-------------------------------+----------------------+----------------------+
|   2  GeForce GTX 970    WDDM  | 00000000:07:00.0 Off |                  N/A |
|100%   82C    P2   138W / 187W |   1281MiB /  4096MiB |     88%      Default |
+-------------------------------+----------------------+----------------------+


As you can see on the Windows machine, the old GTX 970 is more power hungry than the GTX 1070.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1909624 · Report as offensive     Reply Quote
Profile Shaggie76
Avatar

Send message
Joined: 9 Oct 09
Posts: 267
Credit: 157,049,361
RAC: 190,225
Canada
Message 1909711 - Posted: 31 Dec 2017, 0:22:42 UTC

Average of Median 60% Credit/Hour from my last run:

NVIDIA GeForce GTX 1070: 908.8006478
NVIDIA GeForce GTX 1070 Ti: 941.2174884

So approximately 3.6% more credit from a wide sampling of computers and tasks.

From your data below, average power:

NVIDIA GeForce GTX 1070: 111.5 W
NVIDIA GeForce GTX 1070 Ti: 117.7 W

So approximately 5.6% more power in this specific case.

I noticed your 1070's were rated for 166W not the stock 150W -- so maybe clocked a bit higher and drawing a bit more power too?
ID: 1909711 · Report as offensive     Reply Quote
Profile Brent Norman Special Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2395
Credit: 273,913,231
RAC: 686,334
Canada
Message 1909717 - Posted: 31 Dec 2017, 1:07:51 UTC - in response to Message 1909711.  

I wonder how badly you think mismatched cards in a system affect your numbers.
For instance my 4x1070 computer is actually a 1070/980/1060/1060.
ID: 1909717 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 5571
Credit: 386,928,160
RAC: 1,037,237
United States
Message 1909739 - Posted: 31 Dec 2017, 2:38:34 UTC - in response to Message 1909711.  

Average of Median 60% Credit/Hour from my last run:

NVIDIA GeForce GTX 1070: 908.8006478
NVIDIA GeForce GTX 1070 Ti: 941.2174884

So approximately 3.6% more credit from a wide sampling of computers and tasks.

From your data below, average power:

NVIDIA GeForce GTX 1070: 111.5 W
NVIDIA GeForce GTX 1070 Ti: 117.7 W

So approximately 5.6% more power in this specific case.

I noticed your 1070's were rated for 166W not the stock 150W -- so maybe clocked a bit higher and drawing a bit more power too?

No, not overclocked in the traditional sense. But use an offset in P2 state to get them back to where they should be running in P0 state which would be at their normal gaming mode P0 state spec.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 1909739 · Report as offensive     Reply Quote
Profile Shaggie76
Avatar

Send message
Joined: 9 Oct 09
Posts: 267
Credit: 157,049,361
RAC: 190,225
Canada
Message 1909892 - Posted: 31 Dec 2017, 21:09:10 UTC - in response to Message 1909717.  

I wonder how badly you think mismatched cards in a system affect your numbers.
For instance my 4x1070 computer is actually a 1070/980/1060/1060.

The script omits all multi-GPU systems for exactly this reason.
ID: 1909892 · Report as offensive     Reply Quote
Stephen "Heretic" Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 3706
Credit: 84,825,213
RAC: 130,298
Australia
Message 1909995 - Posted: 1 Jan 2018, 9:31:21 UTC - in response to Message 1905513.  

It must be time for another update. ;-)

Cheers.


. . Does that mean you are going to replace the 1060/3 cards? :)

Stephen

??
ID: 1909995 · Report as offensive     Reply Quote
Profile Shaggie76
Avatar

Send message
Joined: 9 Oct 09
Posts: 267
Credit: 157,049,361
RAC: 190,225
Canada
Message 1932562 - Posted: 28 Apr 2018, 22:51:58 UTC

New thread because my old ones keep getting locked. Feel free to unlock/merge if prefer. New this scan: Radeon Pro Vega 56's coming in barely faster than a GTX 1060 (I'm guessing these are in iMacs?)

ID: 1932562 · Report as offensive     Reply Quote
Profile betreger Project Donor
Avatar

Send message
Joined: 29 Jun 99
Posts: 8284
Credit: 19,809,404
RAC: 13,242
United States
Message 1932563 - Posted: 28 Apr 2018, 22:58:25 UTC

Thanx for your efforts.
ID: 1932563 · Report as offensive     Reply Quote
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 10345
Credit: 138,979,954
RAC: 84,159
Australia
Message 1932568 - Posted: 28 Apr 2018, 23:58:52 UTC - in response to Message 1932563.  
Last modified: 29 Apr 2018, 0:22:00 UTC

Thanx for your efforts.

Seconded.
Motion carried.


Interesting to see the GTX 750Ti, after all this time, still up near the top of the pack for Credit/W-Hr.
If Petri's TitanV's performance & power usage is anything to go by, consumer Volta cards might finally push the GTX 750Ti down the list, almost 4 years later.
Grant
Darwin NT
ID: 1932568 · Report as offensive     Reply Quote
juan BFP Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 16 Mar 07
Posts: 7133
Credit: 412,053,206
RAC: 281,101
Panama
Message 1932585 - Posted: 29 Apr 2018, 2:20:52 UTC

Is possible to run the script to compare the GPU`s running on Linux?
ID: 1932585 · Report as offensive     Reply Quote
Profile petri33
Volunteer tester

Send message
Joined: 6 Jun 02
Posts: 1614
Credit: 376,680,018
RAC: 357,918
Finland
Message 1936333 - Posted: 19 May 2018, 22:41:56 UTC - in response to Message 1905508.  

I wonder what sort of performance this sucker would give?
GTX 1080Ti CUDA Cores 3584
Titan Xp CUDA Cores 3840
Titan V CUDA Cores 5120

For $3,000US you'd want it to kiss you good night & tuck you in as well as crunch like nothing else...


I have Compared the TITAN V to a GTX 1080. The Titan is 92 % (shorties) - 140 % (guppi & vlar) faster.
To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones
ID: 1936333 · Report as offensive     Reply Quote
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6015
Credit: 84,803,404
RAC: 39,556
Russia
Message 1936633 - Posted: 22 May 2018, 6:56:16 UTC - in response to Message 1932562.  

New thread because my old ones keep getting locked. Feel free to unlock/merge if prefer. New this scan: Radeon Pro Vega 56's coming in barely faster than a GTX 1060 (I'm guessing these are in iMacs?)



Thanks for your work!

Could you consider to put some info about number of hosts of particular type averaged for single line of graph, please.
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1936633 · Report as offensive     Reply Quote
Profile Shaggie76
Avatar

Send message
Joined: 9 Oct 09
Posts: 267
Credit: 157,049,361
RAC: 190,225
Canada
Message 1937251 - Posted: 26 May 2018, 13:04:33 UTC - in response to Message 1936633.  

The thresholds I'm using are at least 10 unique host IDs and at least 25 work units to qualify for each; my results can be combined if I do another scan the next week but I periodically reset to avoid older results with perhaps outdated drivers.

The last scan I posted was a single snapshot and the host counts are here:



Sorry for the image size -- I didn't see how to post tables so I did a hack job of a high-dpi copy from Excel.
ID: 1937251 · Report as offensive     Reply Quote
Profile iwazaru
Volunteer tester
Avatar

Send message
Joined: 31 Oct 99
Posts: 84
Credit: 424,498
RAC: 0
Greece
Message 1937260 - Posted: 26 May 2018, 14:02:33 UTC - in response to Message 1937251.  

Shaggie, since you're taking requests...
Could you please give Raistmer a rough estimate of how many jigowatts we'd be saving at Seti if he were to improve app efficiency by 10%?
He says he is in need of "motivation" so...
:P

(I promise, I am teasing in the friendliest ways possible. Please LOL!)
ID: 1937260 · Report as offensive     Reply Quote
Profile Raistmer
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 16 Jun 01
Posts: 6015
Credit: 84,803,404
RAC: 39,556
Russia
Message 1937318 - Posted: 26 May 2018, 21:39:59 UTC - in response to Message 1937260.  
Last modified: 26 May 2018, 21:44:10 UTC

@Shaggie76
Thanks for data

@ iwazaru
I never said that need additional motivation besides project's goal.
All I said in another thread is that CreditScrew discourages optimization. It does. But who cares? ;)
And, just to be precise, your proposal for additioanl motivation is wrong one.
If app performance will be increased GPU power consumption will increase too. We will do more work per hour... but will consume more energy also.
So, greenpeace will be disappointed :P
SETI apps news
We're not gonna fight them. We're gonna transcend them.
ID: 1937318 · Report as offensive     Reply Quote
Profile iwazaru
Volunteer tester
Avatar

Send message
Joined: 31 Oct 99
Posts: 84
Credit: 424,498
RAC: 0
Greece
Message 1937326 - Posted: 26 May 2018, 23:14:02 UTC - in response to Message 1937318.  

I hope you know I was completely kidding :)
But damn, I can't believe I forgot power usually goes up with optimizations :P

Touché!
ID: 1937326 · Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 8 · Next

Message boards : Number crunching : GPU FLOPS: Theory vs Reality


 
©2018 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.