Advanced search

Message boards : News : WU: CASP

Author Message
Stefan
Volunteer moderator
Project developer
Project scientist
Send message
Joined: 5 Mar 13
Posts: 258
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 44670 - Posted: 13 Oct 2016 | 14:54:07 UTC

Hey all. The next weeks I will be sending out some simulations starting with the CASP name. It will consist of a total of 3600 simulations running at any time (meaning that once some of them complete, new ones are sent to reach again 3600) probably for a month or so.
It will probably be my most thorough and organized method test so far so it should be fun :D

The simulations have their length written in their name (1ns, 5ns, 20ns and 50ns). Some will end up in the short queue and some in the long queue. Longest simulation time I measured on a 780 was 8 hours for one of the proteins at 50ns and shortest at 0.1 hour.

Right now I sent out the first protein called NTL9 to test if everything works as expected (so no 3600 simulations yet) but the next week I will probably send out another 4 proteins. So there should be lots to crunch in the coming days :)

Profile Logan Carr
Send message
Joined: 12 Aug 15
Posts: 193
Credit: 25,979,525
RAC: 26
Level
Val
Scientific publications
wat
Message 44679 - Posted: 13 Oct 2016 | 21:00:28 UTC - in response to Message 44670.

Thank you very much, Stefan!

I'll be more than happy to assist with the short ones! I'm looking forward to trying them out!


____________
Cruncher/Learner in progress.

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 1813
Credit: 9,929,924,294
RAC: 5,877,361
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44682 - Posted: 13 Oct 2016 | 23:23:51 UTC
Last modified: 13 Oct 2016 | 23:26:19 UTC

From my experience these workunits tolerate less overclocking.
http://www.gpugrid.net/result.php?resultid=15343984
http://www.gpugrid.net/result.php?resultid=15343983
http://www.gpugrid.net/result.php?resultid=15343948

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1073
Credit: 4,463,718,154
RAC: 432,624
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44683 - Posted: 13 Oct 2016 | 23:29:31 UTC - in response to Message 44670.
Last modified: 13 Oct 2016 | 23:31:22 UTC

The simulations have their length written in their name (1ns, 5ns, 20ns and 50ns). Some will end up in the short queue and some in the long queue. Longest simulation time I measured on a 780 was 8 hours for one of the proteins at 50ns and shortest at 0.1 hour.

Right now I sent out the first protein called NTL9 to test if everything works as expected (so no 3600 simulations yet) but the next week I will probably send out another 4 proteins. So there should be lots to crunch in the coming days :)

Just had 3 run through my GTX 670 (Win7-64):

Two 1ns took about 7 minutes each.
The 50ns took just over 6 hours.

All 3 WUs ran fine.

Bedrich Hajek
Send message
Joined: 28 Mar 09
Posts: 331
Credit: 3,757,194,209
RAC: 228,066
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44689 - Posted: 14 Oct 2016 | 1:43:57 UTC - in response to Message 44683.

I got 2 of these tasks, both completed successfully:

15344547 30790 14 Oct 2016 | 0:26:27 UTC 14 Oct 2016 | 1:34:35 UTC Completed and validated 214.88 207.75 600.00 Short runs (2-3 hours on fastest card) v8.48 (cuda65)

http://www.gpugrid.net/workunit.php?wuid=11766563

15341790 263612 13 Oct 2016 | 14:55:46 UTC 13 Oct 2016 | 21:49:50 UTC Completed and validated 11,318.60 11,237.88 63,750.00 Long runs (8-12 hours on fastest card) v8.48 (cuda65)

http://www.gpugrid.net/workunit.php?wuid=11765049

Even the long runs seemed short. It seems to me that all the tasks have been getting shorter and shorter, recently.


Profile Logan Carr
Send message
Joined: 12 Aug 15
Posts: 193
Credit: 25,979,525
RAC: 26
Level
Val
Scientific publications
wat
Message 44690 - Posted: 14 Oct 2016 | 2:06:53 UTC - in response to Message 44689.
Last modified: 14 Oct 2016 | 2:07:52 UTC

https://www.gpugrid.net/result.php?resultid=15344597

https://www.gpugrid.net/result.php?resultid=15344565

https://www.gpugrid.net/result.php?resultid=15344564

https://www.gpugrid.net/result.php?resultid=15343964

^ These are all of the short runs I did so far. Hopefully the results can help a few people or the scientists. I have ran these just fine on my GTX 960 with windows XP and it worked amazing!
____________
Cruncher/Learner in progress.

Stefan
Volunteer moderator
Project developer
Project scientist
Send message
Joined: 5 Mar 13
Posts: 258
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 44695 - Posted: 14 Oct 2016 | 8:42:53 UTC - in response to Message 44689.

@Bedrich The experiment I'm doing dictates the length of the simulations. If I make them longer I won't be able to prove my point. So unfortunately I can't make them longer.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1073
Credit: 4,463,718,154
RAC: 432,624
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44697 - Posted: 14 Oct 2016 | 12:59:58 UTC - in response to Message 44695.
Last modified: 14 Oct 2016 | 13:22:09 UTC

@Bedrich The experiment I'm doing dictates the length of the simulations. If I make them longer I won't be able to prove my point. So unfortunately I can't make them longer.

Hi Stefan, I don't think Bedrich was complaining. Personally very happy that WUs have returned to a more manageable size. Everything's running smoother and I thank you guys for that. BTW, the 1ns WUs are taking about 9-10 minutes on a 750Ti and the 5ns WUs average about 47 minutes on a 750Ti (Win7-64). Looks like the 20ns are running about 3 hours and the 50ns about 8.5 hours on the same GPUs.

Edit: made a quick list since someone asked on the other thread. SDOERR_CASP times:
GTX 670: 1ns about 7 minutes, 5ns about 35 minutes, 20ns about 2.5 hours, 50 ns about 6 hours.
GTX 750Ti: 1ns about 9.5 minutes, 5ns about 47 minutes, 20ns about 3 hours, 50ns about 8.5 hours.

Profile caffeineyellow5
Avatar
Send message
Joined: 30 Jul 14
Posts: 225
Credit: 2,658,976,345
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwat
Message 44704 - Posted: 14 Oct 2016 | 22:03:36 UTC
Last modified: 14 Oct 2016 | 22:29:09 UTC

10.5 mins
1ns
https://www.gpugrid.net/result.php?resultid=15348016

7 mins
1ns
https://www.gpugrid.net/result.php?resultid=15348045

7 mins
1ns
https://www.gpugrid.net/result.php?resultid=15348088

24.4 mins
5ns
https://www.gpugrid.net/result.php?resultid=15345748

01:35:00
20ns
https://www.gpugrid.net/result.php?resultid=15346512

25 mins
5ns
https://www.gpugrid.net/result.php?resultid=15346067

24.8 mins
5ns
https://www.gpugrid.net/result.php?resultid=15345539

01:40:00
20ns
https://www.gpugrid.net/result.php?resultid=15344861

07:33:00
50ns
https://www.gpugrid.net/result.php?resultid=15342889

This one here ran and errored out about 6 hours ago. It was a 50ns one. It says the runtime was about 30.5 minutes, but from the time it was sent to the time it reported back was almost 2 hours. It must have gotten lost in transmission to and/or from me to the server for the other hour and a half. (And obviously, the second user [10esseeTony] failed without running because the card is a 1070 with an expected 100% error rate since Oct 3.)
https://www.gpugrid.net/workunit.php?wuid=11768069

This 20ns WU is currently running at 6.5% after 13:47 minutes.
https://www.gpugrid.net/result.php?resultid=15348080

This 50ns WU is currently running at 32.7% after 02:02:50.
https://www.gpugrid.net/result.php?resultid=15347994

This 50ns WU is currently running at 56% after 02:17:50.
https://www.gpugrid.net/result.php?resultid=15347985

These are on GTX 980TI Classified cards running 2 WUs per card adding 18% usage to each card for their runs along side long runs on the same card.
____________
1 Corinthians 9:16 "For though I preach the gospel, I have nothing to glory of: for necessity is laid upon me; yea, woe is unto me, if I preach not the gospel!"
Ephesians 6:18-20, please ;-)
http://tbc-pa.org

Bedrich Hajek
Send message
Joined: 28 Mar 09
Posts: 331
Credit: 3,757,194,209
RAC: 228,066
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44706 - Posted: 15 Oct 2016 | 1:46:28 UTC - in response to Message 44697.

@Bedrich The experiment I'm doing dictates the length of the simulations. If I make them longer I won't be able to prove my point. So unfortunately I can't make them longer.

Hi Stefan, I don't think Bedrich was complaining. Personally very happy that WUs have returned to a more manageable size. Everything's running smoother and I thank you guys for that. BTW, the 1ns WUs are taking about 9-10 minutes on a 750Ti and the 5ns WUs average about 47 minutes on a 750Ti (Win7-64). Looks like the 20ns are running about 3 hours and the 50ns about 8.5 hours on the same GPUs.

Edit: made a quick list since someone asked on the other thread. SDOERR_CASP times:
GTX 670: 1ns about 7 minutes, 5ns about 35 minutes, 20ns about 2.5 hours, 50 ns about 6 hours.
GTX 750Ti: 1ns about 9.5 minutes, 5ns about 47 minutes, 20ns about 3 hours, 50ns about 8.5 hours.


For the record, I was merely stating my observation and opinion, not complaining nor intending to be derogatory.

And here is some more: I was able to finish the 50ns task in under 3 hours, so that would technically make it a short run, not a long run:

15345412 30790 14 Oct 2016 | 6:05:02 UTC 14 Oct 2016 | 13:46:26 UTC Completed and validated 9,970.77 9,892.05 63,750.00 Long runs (8-12 hours on fastest card) v8.48 (cuda65)

http://www.gpugrid.net/workunit.php?wuid=11767231

Also, the GPU usage on windows 10 computer was 72% and power usage was 59%, which are low. On the windows xp machine, the GPU usage was 87% and power usage was 65%, which are also a little low.



Betting Slip
Send message
Joined: 5 Jan 09
Posts: 574
Credit: 1,904,392,225
RAC: 1,826,399
Level
His
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44707 - Posted: 15 Oct 2016 | 3:13:06 UTC - in response to Message 44706.



For the record, I was merely stating my observation and opinion, not complaining nor intending to be derogatory.


Oh, come on be derogatory, this forum could do with it.

Bedrich Hajek
Send message
Joined: 28 Mar 09
Posts: 331
Credit: 3,757,194,209
RAC: 228,066
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44710 - Posted: 15 Oct 2016 | 14:01:19 UTC - in response to Message 44707.



For the record, I was merely stating my observation and opinion, not complaining nor intending to be derogatory.


Oh, come on be derogatory, this forum could do with it.



You can take the lead on derogatory, I'll just be sarcastic.




Profile caffeineyellow5
Avatar
Send message
Joined: 30 Jul 14
Posts: 225
Credit: 2,658,976,345
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwat
Message 44719 - Posted: 15 Oct 2016 | 22:50:49 UTC - in response to Message 44710.

01:30:00
5ns
https://www.gpugrid.net/result.php?resultid=15349944

On a Quadro K2100M laptop video card.

Bedrich Hajek
Send message
Joined: 28 Mar 09
Posts: 331
Credit: 3,757,194,209
RAC: 228,066
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44773 - Posted: 18 Oct 2016 | 23:32:12 UTC

I had 2 tasks (labeled as a long runs) each completed in a little over 1 hour:

15364004 30790 18 Oct 2016 | 20:16:24 UTC 18 Oct 2016 | 22:24:38 UTC Completed and validated 3,955.64 3,920.84 25,500.00 Long runs (8-12 hours on fastest card) v8.48 (cuda65)

http://www.gpugrid.net/workunit.php?wuid=11781198

15360119 30790 18 Oct 2016 | 9:23:49 UTC 18 Oct 2016 | 11:54:12 UTC Completed and validated 3,956.91 3,925.63 25,500.00 Long runs (8-12 hours on fastest card) v8.48 (cuda65)

http://www.gpugrid.net/workunit.php?wuid=11778232


I also had short runs finish in a little over 3 minutes.

Maybe, it's time to redefine short and longs runs.

Short runs would take under 3 hours to complete on the fastest cards.
Long runs would take 3 to 12 hours to complete on the fastest cards.
We could add another category for runs over 12 hours on the fastest cards.

That is what has been basically happening, recently, for the most part.

And please put the tasks in the correctly category!



frederikhk
Send message
Joined: 14 Apr 14
Posts: 8
Credit: 57,034,536
RAC: 0
Level
Thr
Scientific publications
watwatwat
Message 44779 - Posted: 19 Oct 2016 | 10:37:54 UTC - in response to Message 44773.

I had 2 tasks (labeled as a long runs) each completed in a little over 1 hour:

15364004 30790 18 Oct 2016 | 20:16:24 UTC 18 Oct 2016 | 22:24:38 UTC Completed and validated 3,955.64 3,920.84 25,500.00 Long runs (8-12 hours on fastest card) v8.48 (cuda65)

http://www.gpugrid.net/workunit.php?wuid=11781198

15360119 30790 18 Oct 2016 | 9:23:49 UTC 18 Oct 2016 | 11:54:12 UTC Completed and validated 3,956.91 3,925.63 25,500.00 Long runs (8-12 hours on fastest card) v8.48 (cuda65)

http://www.gpugrid.net/workunit.php?wuid=11778232


I also had short runs finish in a little over 3 minutes.

Maybe, it's time to redefine short and longs runs.

Short runs would take under 3 hours to complete on the fastest cards.
Long runs would take 3 to 12 hours to complete on the fastest cards.
We could add another category for runs over 12 hours on the fastest cards.

That is what has been basically happening, recently, for the most part.

And please put the tasks in the correctly category!





Stefan already answered you why "@Bedrich The experiment I'm doing dictates the length of the simulations. If I make them longer I won't be able to prove my point. So unfortunately I can't make them longer."

eXaPower
Send message
Joined: 25 Sep 13
Posts: 260
Credit: 870,516,892
RAC: 2,861,846
Level
Glu
Scientific publications
watwatwatwatwatwat
Message 44781 - Posted: 19 Oct 2016 | 12:18:54 UTC

CASP WU on (2) GTX 970 at 1.5GHz are 2.5~2.7x slower with PCIe 2.0 x1 compared to PCIe 3.0 x4.
WU's require PCIe 3.0 x8 for proper OC scaling.

1ns: 900/sec vs. 350/sec
5ns: 4770/sec vs. 1761/sec

PCIe 2.0 x1: 46% GPU / 7% MCU / 80% BUS usage / 75W GPU power
PCIe 3.0 x4: 57% GPU / 12% MCU / 40% BUS / 107W


Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1073
Credit: 4,463,718,154
RAC: 432,624
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44783 - Posted: 19 Oct 2016 | 16:37:21 UTC - in response to Message 44781.

CASP WU on (2) GTX 970 at 1.5GHz are 2.5~2.7x slower with PCIe 2.0 x1 compared to PCIe 3.0 x4.
WU's require PCIe 3.0 x8 for proper OC scaling.

Or PCIe 2.0 x16...

Bedrich Hajek
Send message
Joined: 28 Mar 09
Posts: 331
Credit: 3,757,194,209
RAC: 228,066
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44788 - Posted: 19 Oct 2016 | 22:39:58 UTC - in response to Message 44779.

I had 2 tasks (labeled as a long runs) each completed in a little over 1 hour:

15364004 30790 18 Oct 2016 | 20:16:24 UTC 18 Oct 2016 | 22:24:38 UTC Completed and validated 3,955.64 3,920.84 25,500.00 Long runs (8-12 hours on fastest card) v8.48 (cuda65)

http://www.gpugrid.net/workunit.php?wuid=11781198

15360119 30790 18 Oct 2016 | 9:23:49 UTC 18 Oct 2016 | 11:54:12 UTC Completed and validated 3,956.91 3,925.63 25,500.00 Long runs (8-12 hours on fastest card) v8.48 (cuda65)

http://www.gpugrid.net/workunit.php?wuid=11778232


I also had short runs finish in a little over 3 minutes.

Maybe, it's time to redefine short and longs runs.

Short runs would take under 3 hours to complete on the fastest cards.
Long runs would take 3 to 12 hours to complete on the fastest cards.
We could add another category for runs over 12 hours on the fastest cards.

That is what has been basically happening, recently, for the most part.

And please put the tasks in the correctly category!





Stefan already answered you why "@Bedrich The experiment I'm doing dictates the length of the simulations. If I make them longer I won't be able to prove my point. So unfortunately I can't make them longer."


My post was not about the length of the tasks. It was about classifying the tasks correctly, and updating the definition of each category.

Please read the post more carefully next time.



Profile Logan Carr
Send message
Joined: 12 Aug 15
Posts: 193
Credit: 25,979,525
RAC: 26
Level
Val
Scientific publications
wat
Message 44816 - Posted: 22 Oct 2016 | 20:28:58 UTC - in response to Message 44788.

I notice with the CASP units I have around 80% GPU usage using windows XP and a gtx 960. I see that other's are having lower gpu usage as well. Will the casp units always use this low of a gpu usage?

I'm not complaining by the way I'm just curious. Thanks.
____________
Cruncher/Learner in progress.

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 1813
Credit: 9,929,924,294
RAC: 5,877,361
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44817 - Posted: 22 Oct 2016 | 21:14:44 UTC - in response to Message 44816.
Last modified: 22 Oct 2016 | 21:49:21 UTC

I notice with the CASP units I have around 80% GPU usage using windows XP and a gtx 960. I see that other's are having lower gpu usage as well. Will the casp units always use this low of a gpu usage?

Most probably they will. I think it's because the models these units are simulating have "only" 11340 atoms, while others have 2-5 times of this. The smaller the model is, the more frequent the CPU has to do the DP part of the simulation, resulting in lower GPU usage. (However there were high atom count batches with low GPU usage in the past, so a larger model could also need relatively high CPU-GPU interaction.)

Profile Logan Carr
Send message
Joined: 12 Aug 15
Posts: 193
Credit: 25,979,525
RAC: 26
Level
Val
Scientific publications
wat
Message 44818 - Posted: 22 Oct 2016 | 23:26:20 UTC - in response to Message 44817.

If my CPU and GPU have to interact a lot more often I might be in trouble since the only machine I have access to right now has a pentium D CPU. I'll just hope for the best until I can get my hands on a newer machine.


____________
Cruncher/Learner in progress.

Stefan
Volunteer moderator
Project developer
Project scientist
Send message
Joined: 5 Mar 13
Posts: 258
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 44829 - Posted: 24 Oct 2016 | 12:21:08 UTC

Hm okay I can move some stuff from long to short queue. I was stretching the definition a bit but Gianni said to go with it. But if it's an issue I will send only 4+ hours to long and under that to short.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1073
Credit: 4,463,718,154
RAC: 432,624
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44832 - Posted: 25 Oct 2016 | 13:58:39 UTC - in response to Message 44829.
Last modified: 25 Oct 2016 | 13:59:03 UTC

Hm okay I can move some stuff from long to short queue. I was stretching the definition a bit but Gianni said to go with it. But if it's an issue I will send only 4+ hours to long and under that to short.

Stefan, I don't see it as an issue at all. Leave them in the long queue.

Stefan
Volunteer moderator
Project developer
Project scientist
Send message
Joined: 5 Mar 13
Posts: 258
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 44834 - Posted: 25 Oct 2016 | 15:45:24 UTC
Last modified: 25 Oct 2016 | 15:47:56 UTC

Yeah I switched to the 4 hour thing but Gianni also told me to revert it back. I guess for this project we can "suffer" some shorter sims in the long queue.

I am sending now some thousands more simulations including a new protein called alpha3D (a3D).

Awesome job by the way on the crunching. I guess since I never did many simulations before I hadn't realized just how awesome the throughput of GPUGRID is :P Totally enjoying it now, hehe.

Profile Beyond
Avatar
Send message
Joined: 23 Nov 08
Posts: 1073
Credit: 4,463,718,154
RAC: 432,624
Level
Arg
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44844 - Posted: 25 Oct 2016 | 20:50:50 UTC - in response to Message 44834.

Yeah I switched to the 4 hour thing but Gianni also told me to revert it back. I guess for this project we can "suffer" some shorter sims in the long queue.

I am sending now some thousands more simulations including a new protein called alpha3D (a3D).

Awesome job by the way on the crunching. I guess since I never did many simulations before I hadn't realized just how awesome the throughput of GPUGRID is :P Totally enjoying it now, hehe.

Thumbs up on the long queue decision and a double thumbs up on being a DOCTOR!
BTW, I've a krick in my knee...

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 1813
Credit: 9,929,924,294
RAC: 5,877,361
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 44845 - Posted: 25 Oct 2016 | 21:01:13 UTC - in response to Message 44834.

Yeah I switched to the 4 hour thing but Gianni also told me to revert it back. I guess for this project we can "suffer" some shorter sims in the long queue.

It's ok to have the 50ns and the 20ns long simulations in the long queue, until there are any in the long queue; as the 5ns and the 1ns long simulations take only 14 minutes and 3 minutes (respectively) to process, the data transfer of a long task takes more time than the calculation of a short task, which makes my hosts to download tasks from my backup project (because there's the limit of two GPUGrid workunits).

I am sending now some thousands more simulations including a new protein called alpha3D (a3D).

This should be no problem. :)

Awesome job by the way on the crunching. I guess since I never did many simulations before I hadn't realized just how awesome the throughput of GPUGRID is :P Totally enjoying it now, hehe.

Besides enjoying it, it is your job and responsibility to use and nourish this huge computing power as wisely as possible. As most of us (your volunteer crunchers) are not into biochemistry this job partly consists of making us motivated to support a research whose results we can't comprehend as much as you do (to be polite :) ).

Stefan
Volunteer moderator
Project developer
Project scientist
Send message
Joined: 5 Mar 13
Posts: 258
Credit: 0
RAC: 0
Level

Scientific publications
wat
Message 44855 - Posted: 26 Oct 2016 | 12:58:42 UTC - in response to Message 44844.
Last modified: 26 Oct 2016 | 12:59:56 UTC

BTW, I've a krick in my knee...




Zoltan, once we have some results I will make a post about it to explain it. I know we are not really top at the communication aspect. Right now it's still a bit in the making and a bit under the covers to avoid competition etc. But if the results are nice it will probably make a quite important publication.

eXaPower
Send message
Joined: 25 Sep 13
Posts: 260
Credit: 870,516,892
RAC: 2,861,846
Level
Glu
Scientific publications
watwatwatwatwatwat
Message 45007 - Posted: 29 Oct 2016 | 15:22:47 UTC - in response to Message 44781.
Last modified: 29 Oct 2016 | 15:31:59 UTC

CASP WU on (2) GTX 970 at 1.5GHz are 2.5~2.7x slower with PCIe 2.0 x1 compared to PCIe 3.0 x4.
WU's require PCIe 3.0 x8 for proper OC scaling.

1ns: 900/sec vs. 350/sec
5ns: 4770/sec vs. 1761/sec

PCIe 2.0 x1: 46% GPU / 7% MCU / 80% BUS usage / 75W GPU power
PCIe 3.0 x4: 57% GPU / 12% MCU / 40% BUS / 107W

CASP runtimes (atom and step amount) vary so this just a general reference.

GTX 1070 (PCIe 3.0 x8) CASP runtimes:

-- 1ns (ntl9) 600 credits = 240/sec @ 2.1GHz / 59% GPU / 15% MCU / 37% BUS / 78W power
-- 1ns (a3d)= 1,350 credits = 330/sec @ 2.1GHz / 70% GPU / 24% MCU / 39% BUS / 96W power

-- 5ns (ntl9) 3,150 credits = 1,200/sec @ same usage and power numbers as 1ns
-- 5ns (a3d) 6,900 credits = 1,600/sec @ same usage and power numbers as 1ns

GTX 1060 (3GB) PCIe 3.0 x4 CASP runtimes:

-- 1ns (ntl9) 600 credits = 300/sec @ 2.1GHz / 63% GPU / 17% MCU / 51% BUS / 74W power
-- 1ns (a3d) 1,350 credits = 450/sec @ 2.1GHz / 74% GPU / 24% MCU / 59% BUS / 88W power

-- 5ns (ntl9) 3,150 credits = 1,500/sec @ same GPU usage and power as 1ns
-- 5ns (a3d) = 6,900 credits = 2.275/sec @ same GPU usage and power as 1ns

IMO: a (1152CUDA GTX 1060) is on par with (2048CUDA GTX 980) and ~20% faster than a (1664CUDA GTX 970).
The (1920CUDA GTX 1070) is as (if not) ~5% faster than a (2816CUDA GTX 980ti).

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 1813
Credit: 9,929,924,294
RAC: 5,877,361
Level
Tyr
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 45037 - Posted: 30 Oct 2016 | 10:18:26 UTC - in response to Message 45007.

IMO: a (1152CUDA GTX 1060) is on par with (2048CUDA GTX 980) and ~20% faster than a (1664CUDA GTX 970).
The (1920CUDA GTX 1070) is as (if not) ~5% faster than a (2816CUDA GTX 980ti).

This is true if you compare these cards under WDDM os.
I guesstimate that my GTX980Ti@1390MHz/3500MHz will be ~6% faster under Windows XP x64 than my GTX1080@2000MHz/4750MHz under Windows 10x64 (while processing CASP11_crystal_contacts_**ns_a3D workunits). I also guesstimate that the TITAN X (Pascal) cards won't scale well under WDDM os especially with low atom-count workunits, the TITAN X (Pascal) will be just slightly faster than a GTX 1080. In general, under a WDDM os the TITAN X (Pascal) GPUs won't be as faster as they should be taking the ratio of the CUDA cores of the GTX 1080 and the TITAN X (Pascal) GPUs in consideration.

PappaLitto
Send message
Joined: 21 Mar 16
Posts: 239
Credit: 955,117,381
RAC: 3,478,910
Level
Glu
Scientific publications
watwat
Message 45300 - Posted: 17 Nov 2016 | 15:15:47 UTC - in response to Message 44855.

BTW, I've a krick in my knee...






Dr. Who?

Profile caffeineyellow5
Avatar
Send message
Joined: 30 Jul 14
Posts: 225
Credit: 2,658,976,345
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwat
Message 45347 - Posted: 21 Nov 2016 | 9:41:17 UTC - in response to Message 45300.

hah!

Post to thread

Message boards : News : WU: CASP