1050ti Optimize

Message boards : Number crunching : 1050ti Optimize
Message board moderation

To post messages, you must log in.

AuthorMessage
elec999 Project Donor

Send message
Joined: 24 Nov 02
Posts: 375
Credit: 416,969,548
RAC: 141
Canada
Message 1879032 - Posted: 19 Jul 2017, 12:53:51 UTC

Good Day,
Can anyone tell me the best way to optimized the work for 1050ti?
Thank you
ID: 1879032 · Report as offensive
Alfred and Pauline
Volunteer tester

Send message
Joined: 13 Oct 99
Posts: 37
Credit: 27,139,561
RAC: 0
Canada
Message 1879062 - Posted: 19 Jul 2017, 16:55:04 UTC

I would try the following (it works reasonably well for me)

-sbs 1024 -period_iterations_num 10 -spike_fft_tresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 32 -oclfft_tune_wg 256 -oclftt_tune_ls 512 -oclfft_tune_bn 64 -oclftt_tune_cw 64

Hopefully the above is without mistook.
ID: 1879062 · Report as offensive
elec999 Project Donor

Send message
Joined: 24 Nov 02
Posts: 375
Credit: 416,969,548
RAC: 141
Canada
Message 1879080 - Posted: 19 Jul 2017, 18:49:14 UTC

Thanks I put it in, hopefully it will help. ANyway to check if its working or done it correctly?
ID: 1879080 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1879082 - Posted: 19 Jul 2017, 19:03:04 UTC - in response to Message 1879080.  

Look at the stderr once the results are returned after it is installed.

Should see the changes listed in there.
ID: 1879082 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1879199 - Posted: 20 Jul 2017, 8:10:30 UTC - in response to Message 1879062.  

I would try the following (it works reasonably well for me)

-sbs 1024 -period_iterations_num 10 -spike_fft_tresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 32 -oclfft_tune_wg 256 -oclftt_tune_ls 512 -oclfft_tune_bn 64 -oclftt_tune_cw 64

Hopefully the above is without mistake.


Two other things you cam play with. This can go down to -period_iterations_num 1 if you don't get "screen lag" you can't live with. And you might try -tt1500 to see if that speeds you up.

Depending on if your system is "dual use" or "dedicated" you can play with -hp and -high_perform Both of these usually will cause major keyboard/mouse/screen lag. So are pretty much "dedicated".

If you want to run "dedicated" but still have a usable system you can setup your "when the computer is in use" preferences to suspend the GPU when the computer is in use. This will (usually) give you good usability while using the more extreme parameters. The catch is BOINC doesn't always respond promptly to your "wake up" so suspend the gpu already signal of typing/moving mouse.

HTH,
Tom
A proud member of the OFA (Old Farts Association).
ID: 1879199 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1879248 - Posted: 20 Jul 2017, 14:30:48 UTC - in response to Message 1879199.  



Two other things you cam play with. This can go down to -period_iterations_num 1 if you don't get "screen lag" you can't live with. And you might try -tt1500 to see if that speeds you up.

Depending on if your system is "dual use" or "dedicated" you can play with -hp and -high_perform


2 errors I see.

First -tt1500 should be -tt 1500. There needs to be a space between the -tt and the 1500

Next , it's -high_perf not -high_perform

The commandlines get very picky about missing spaces or wrong words.

Good luck
ID: 1879248 · Report as offensive
KLiK
Volunteer tester

Send message
Joined: 31 Mar 14
Posts: 1304
Credit: 22,994,597
RAC: 60
Croatia
Message 1891297 - Posted: 22 Sep 2017, 11:19:00 UTC

Not to open up new topic, it's about 1050Ti.

So, right now I'm running it on drivers v370. But from time to time I also play a Mass Effect Andromeda on that computer, which asks every time for v378 drivers or later.
Recently (stupid Windows 10) have updated by themselves to WHQL v385, which produced GSOD (sorry wingmans).

Anyone has a stable v378+ on 1050Ti?

p.s. Will post my setting when I arrive home. ;)


non-profit org. Play4Life in Zagreb, Croatia, EU
ID: 1891297 · Report as offensive
Profile tullio
Volunteer tester

Send message
Joined: 9 Apr 04
Posts: 8797
Credit: 2,930,782
RAC: 1
Italy
Message 1891310 - Posted: 22 Sep 2017, 13:27:22 UTC
Last modified: 22 Sep 2017, 13:29:18 UTC

On my Windows 10 PC, which strangely enough has been good for more than one day, I have a GTX 1050Ti. I get my drivers from Geforce and the latest I have is 385.41, while I have 384.69 on my main Linux Box, with GTX 750Ti, driver provided by SuSE.
Tullio
ID: 1891310 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1891360 - Posted: 22 Sep 2017, 16:22:38 UTC - in response to Message 1891297.  

Don't allow Windows to update your drivers. If you must update, go to the nvidia website and download the drivers directly from them.

When Windows installs your drivers, they drop OpenCl from THEIR drivers, making your GPU useless for Seti.

Nvidia doesn't drop the OpenCl so it always better just to get them directly
http://www.nvidia.com/Download/index.aspx?lang=en-us
ID: 1891360 · Report as offensive
KLiK
Volunteer tester

Send message
Joined: 31 Mar 14
Posts: 1304
Credit: 22,994,597
RAC: 60
Croatia
Message 1891575 - Posted: 23 Sep 2017, 20:38:52 UTC - in response to Message 1891360.  

Don't allow Windows to update your drivers. If you must update, go to the nvidia website and download the drivers directly from them.

When Windows installs your drivers, they drop OpenCl from THEIR drivers, making your GPU useless for Seti.

Nvidia doesn't drop the OpenCl so it always better just to get them directly
http://www.nvidia.com/Download/index.aspx?lang=en-us

I found out that "the hard way". Stupid guys at Microsoft. Wrote them some messages with a lot of (not so) nice adjectives! :D

Anyway, here's my setting for SETi@home on standard bunch of apps:
<app_config>
    <app>
      <name>setiathome_v8</name>
      <gpu_versions>
          <gpu_usage>0.25</gpu_usage>
      </gpu_versions>
    </app>
	<app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>cuda60</plan_class>
	   <max_concurrent>3</max_concurrent>
       <ngpus>0.25</ngpus>
       <cmdline></cmdline>
    </app_version>
    <app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>cuda50</plan_class>
	   <max_concurrent>3</max_concurrent>
       <ngpus>0.25</ngpus>
       <cmdline></cmdline>
    </app_version>
    <app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>cuda42</plan_class>
	   <max_concurrent>3</max_concurrent>
       <ngpus>0.25</ngpus>
    </app_version>
    <app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>cuda32</plan_class>
	   <max_concurrent>3</max_concurrent>
       <ngpus>0.25</ngpus>
    </app_version>
	   <app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>opencl_nvidia_SoG</plan_class>
	   <max_concurrent>4</max_concurrent>
	   <cmdline>-high_precision_timer -use_sleep -sbs 1024 -period_iterations_num 12 -tt 360 -instances_per_device 4</cmdline>
       <ngpus>0.25</ngpus>
    </app_version>
	    <app_version>
       <app_name>setiathome_v8</app_name>
       <plan_class>opencl_nvidia_sah</plan_class>
	   <max_concurrent>4</max_concurrent>
	   <cmdline>-high_precision_timer -use_sleep -sbs 1024 -period_iterations_num 12 -tt 360 -instances_per_device 4</cmdline>
       <ngpus>0.25</ngpus>
    </app_version>
</app_config>

Any good suggestion is welcomed.

& these settings will give 12-20k RAC on 1050Ti (single cooler). ;)


non-profit org. Play4Life in Zagreb, Croatia, EU
ID: 1891575 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1891594 - Posted: 24 Sep 2017, 1:07:08 UTC - in response to Message 1891575.  

Ok, a couple of things,

First I would use the Beta Lunatics 6 to install the SoG. That way you can control which apps are running on your computer. I would recommend the Lunatics Win64 0.45 Beta 6

http://mikesworld.eu/download.html

Second, you don't need all those cudas in your app_info.xml. SoG is going to be the best of the standard apps for your card. If you use the installer it will remove those so you don't need to worry about going in and removing them yourself. The installer will create a new app_info.xml

Third, trying to run 4 work units per cards is bottlenecking your card. You would be better off running only 2 per card. It would increase your efficency and probably get you better results.

Fourth, you do not need to add -instances_per_device to you commandline. That was something that was required when we were still using -cpu_lock. So I'd just remove that all together.

Next, I don't see where you have any AstroPulses instructions at all in your app_info.xml or even the application. I would also encourage you to have the installer put that in there as well. Even though they are rare, they give more credits than any other work unit.

Lastly, you can just make a app_config.xml after you use the lunatics installer to control your cards. It's a lot less trouble than having to go through each line of the app_info and make changes. I've added your commandline minus the -instances_per_device. Since you have -use_sleep I've reduced the percentage of CPU in the file down to 0.25. I don't have an AstroPulses commandline for the 1050Ti but if you ask Mike,

https://setiathome.berkeley.edu/show_user.php?userid=9826

I'm sure he can give you one for that card. You will need to create the app_config.xml and place it in the Seti@home folder next to the app_info.xml

<app_config>
  <app_version>
    <app_name>setiathome_v8</app_name>
    <plan_class>opencl_nvidia_SoG</plan_class>
    <avg_ncpus>.25</avg_ncpus>
    <ngpus>0.50</ngpus>
    <cmdline>-high_precision_timer -use_sleep -sbs 1024 -period_iterations_num 12 -tt 360</cmdline>
  </app_version>
   <app_version>
    <app_name>astropulse_v7</app_name>
    <plan_class>opencl_nvidia_100</plan_class>
    <avg_ncpus>.25</avg_ncpus>
    <ngpus>0.50</ngpus>
    <cmdline> </cmdline>
  </app_version>
</app_config>



Good luck,

Zalster
ID: 1891594 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13720
Credit: 208,696,464
RAC: 304
Australia
Message 1891602 - Posted: 24 Sep 2017, 2:07:34 UTC - in response to Message 1891594.  

Third, trying to run 4 work units per cards is bottlenecking your card. You would be better off running only 2 per card. It would increase your efficency and probably get you better results.

SoG on a GTX 1050Ti?
Any more than 1 would be detrimental to output.
Grant
Darwin NT
ID: 1891602 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1891605 - Posted: 24 Sep 2017, 2:23:51 UTC - in response to Message 1891602.  

He's running -use_sleep. The 1050Ti should be able to handle 2 at a time
ID: 1891605 · Report as offensive
Profile Bill G Special Project $75 donor
Avatar

Send message
Joined: 1 Jun 01
Posts: 1282
Credit: 187,688,550
RAC: 182
United States
Message 1891635 - Posted: 24 Sep 2017, 14:51:49 UTC - in response to Message 1891605.  
Last modified: 24 Sep 2017, 14:54:14 UTC

He's running -use_sleep. The 1050Ti should be able to handle 2 at a time

I run 2 at a time on my system with 2 1050Ti's.
(Normal RAC is 32,000. It is a bit lower right now because of down time with W10 upgrades coming every other day it seems)

SETI@home classic workunits 4,019
SETI@home classic CPU time 34,348 hours
ID: 1891635 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1891704 - Posted: 25 Sep 2017, 0:57:26 UTC - in response to Message 1891297.  

Not to open up new topic, it's about 1050Ti.

So, right now I'm running it on drivers v370. But from time to time I also play a Mass Effect Andromeda on that computer, which asks every time for v378 drivers or later.
Recently (stupid Windows 10) have updated by themselves to WHQL v385, which produced GSOD (sorry wingmans).

Anyone has a stable v378+ on 1050Ti?

p.s. Will post my setting when I arrive home. ;)


. . If Windows 10 update replaced your video drivers it might be best to go straight to nvidia and get the good ones :) M$ scramble the drivers and remove essential things like OpenCL support. The *&^*%$&*('s.

Stephen

. .
ID: 1891704 · Report as offensive
Profile betreger Project Donor
Avatar

Send message
Joined: 29 Jun 99
Posts: 11360
Credit: 29,581,041
RAC: 66
United States
Message 1891708 - Posted: 25 Sep 2017, 1:53:12 UTC - in response to Message 1891704.  

M$ scramble the drivers and remove essential things like OpenCL support. The *&^*%$&*('s.

A question I ask is what is their motivation?
ID: 1891708 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1891714 - Posted: 25 Sep 2017, 2:38:40 UTC - in response to Message 1891708.  

M$ scramble the drivers and remove essential things like OpenCL support. The *&^*%$&*('s.

A question I ask is what is their motivation?

That is a question I have pondered also. Why are the messing with the developer sourced drivers in the first place. You should just package what Nvidia releases without any changes.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1891714 · Report as offensive

Message boards : Number crunching : 1050ti Optimize


 
©2024 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.