Optimizing the apps

Message boards : Number crunching : Optimizing the apps
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 . . . 12 · Next

AuthorMessage
Brent

Send message
Joined: 25 Jun 14
Posts: 40
Credit: 423,953,904
RAC: 160,377
Message 168 - Posted: 25 Apr 2018, 15:59:39 UTC - in response to Message 125.  
Last modified: 25 Apr 2018, 15:59:58 UTC

[quote]Thank you for this info. I am successfully running with the default values (no config file) and would like to optimize my settings. What would help greatly is if I knew what the default settings are to use as a starting point, since I have not been able to find any settings for my NVIDIA GT 730 GPU. While I fully realize this is not a powerful GPU, it is nevertheless better than nothing. Any help in this area would be greatly appreciated, since am not a developer or programmer.[quote]

BUMP
Brent
ID: 168 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
EG

Send message
Joined: 9 Jun 13
Posts: 19
Credit: 45,764,477,742
RAC: 0
Message 169 - Posted: 25 Apr 2018, 16:04:27 UTC - in response to Message 168.  

[quote]Thank you for this info. I am successfully running with the default values (no config file) and would like to optimize my settings. What would help greatly is if I knew what the default settings are to use as a starting point, since I have not been able to find any settings for my NVIDIA GT 730 GPU. While I fully realize this is not a powerful GPU, it is nevertheless better than nothing. Any help in this area would be greatly appreciated, since am not a developer or programmer.[quote]

BUMP


Try these:
verbose=1
kernels_per_reduction=48
threads=8
lut_size=15
sleep=0
cache_sieve=1
reduce_cpu=0
sieve_size=30

Adjust from there
ID: 169 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
JOHN

Send message
Joined: 8 Feb 10
Posts: 6
Credit: 36,165,642,743
RAC: 20,409,857
Message 194 - Posted: 29 Apr 2018, 21:41:00 UTC

just installed a 1070 ti in my other puter.gonna start with the settings i use for my 1080ti and adjust from there
ID: 194 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 11 Aug 09
Posts: 963
Credit: 24,557,133,931
RAC: 31,138
Message 195 - Posted: 30 Apr 2018, 2:26:25 UTC - in response to Message 194.  

just installed a 1070 ti in my other puter.gonna start with the settings i use for my 1080ti and adjust from there


These are the 1070 settings:

verbose=1
kernels_per_reduction=48
threads=8
lut_size=17
sleep=1
reduce_cpu=0
sieve_size=28
ID: 195 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Arnulf

Send message
Joined: 30 Oct 17
Posts: 7
Credit: 6,211,530,570
RAC: 0
Message 207 - Posted: 1 May 2018, 13:52:54 UTC

This is my current config on my Vega 64:

verbose=1
kernels_per_reduction=48
threads=8
lut_size=17
sieve_size=30
cache_sieve=1
sleep=0
reduce_cpu=0

Note that I have managed to shave off some seconds by setting cache_sieve to 1, and sleep to 0.
Combined with undervolting the Vega - this allows it to run cooler and at higher speeds.
Results now: around 05:35, down from 06:06.
ID: 207 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Brent

Send message
Joined: 25 Jun 14
Posts: 40
Credit: 423,953,904
RAC: 160,377
Message 252 - Posted: 7 May 2018, 2:51:00 UTC - in response to Message 169.  

Thanks for the help. I have tried this config file and while I can see no real sign of increased performance, nothing has crashed yet either.
Brent
ID: 252 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 11 Aug 09
Posts: 963
Credit: 24,557,133,931
RAC: 31,138
Message 256 - Posted: 7 May 2018, 10:16:14 UTC - in response to Message 252.  

Thanks for the help. I have tried this config file and while I can see no real sign of increased performance, nothing has crashed yet either.


It should be finishing the workunits in roughly half the time of not using the settings, it will start with the first workunit after you save the file.
ID: 256 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
mmonnin

Send message
Joined: 29 May 16
Posts: 15
Credit: 1,213,278,633
RAC: 666
Message 280 - Posted: 9 May 2018, 13:11:51 UTC - in response to Message 252.  

Thanks for the help. I have tried this config file and while I can see no real sign of increased performance, nothing has crashed yet either.


DId you tell boinc manager to re-read the config files to pick up the changes?
ID: 280 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Brent

Send message
Joined: 25 Jun 14
Posts: 40
Credit: 423,953,904
RAC: 160,377
Message 284 - Posted: 10 May 2018, 1:15:48 UTC - in response to Message 280.  

Thanks for the help. I have tried this config file and while I can see no real sign of increased performance, nothing has crashed yet either.


DId you tell boinc manager to re-read the config files to pick up the changes?


Yes I did. I also just tried bumping up the lut_size to 17 and no apparent change in performance, but no crashes yet either (knock on wood!)
Brent
ID: 284 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Slicker
Project administrator

Send message
Joined: 11 Jun 09
Posts: 79
Credit: 943,644,517
RAC: 0
Message 286 - Posted: 10 May 2018, 5:41:28 UTC - in response to Message 284.  

Thanks for the help. I have tried this config file and while I can see no real sign of increased performance, nothing has crashed yet either.


DId you tell boinc manager to re-read the config files to pick up the changes?


Yes I did. I also just tried bumping up the lut_size to 17 and no apparent change in performance, but no crashes yet either (knock on wood!)


FYI, bumping up the lut size on GPUs can slow down the processing. The goal is to find the lut size and cache that fit within the GPUs cached RAM. Anything larger is slower and anything smaller is slower. For example, on my laptop
s nVidia 970M, while it can do 1024 threads, it works best at 256 with a lut of 12 and a sieve of 28 or 29. It doesn't have the oomph to support lut 14 or sieve 31 withouth swapping memory and/or over heating such that it throttles itself back to a slower speed.
ID: 286 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Brent

Send message
Joined: 25 Jun 14
Posts: 40
Credit: 423,953,904
RAC: 160,377
Message 312 - Posted: 12 May 2018, 6:10:30 UTC - in response to Message 286.  

Well, I just deleted the config file and re-read the files to pick up the standard configuration, since I could see no improvement in performance.
Brent
ID: 312 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile IDEA

Send message
Joined: 30 May 17
Posts: 119
Credit: 37,173,545,890
RAC: 4
Message 316 - Posted: 12 May 2018, 12:13:58 UTC

Any suggestions for optimising performance from NVIDIA Quadro P600 2GB GPUs?

Can't experiment myself yet as there are no work units available :(
ID: 316 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 11 Aug 09
Posts: 963
Credit: 24,557,133,931
RAC: 31,138
Message 318 - Posted: 12 May 2018, 15:38:38 UTC - in response to Message 316.  

Any suggestions for optimising performance from NVIDIA Quadro P600 2GB GPUs?

Can't experiment myself yet as there are no work units available :(


Scroll up/down depending on your settings to post #71 in this thread, I posted all the codes I have in there, maybe you can find
something that is close to yours an experiment from there.
ID: 318 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile IDEA

Send message
Joined: 30 May 17
Posts: 119
Credit: 37,173,545,890
RAC: 4
Message 361 - Posted: 13 May 2018, 19:54:36 UTC - in response to Message 318.  

Bit difficult experimenting with slow GPUs as they take an hour or more to complete a single unit -- so I was hoping somebody might have experience with the Nviidia Quadro range.

Or a method of working with slow GPUs?

Working with one variable at a time in a set order perhaps?

What config is used by default?

Can you dump the settings out to give a starting point?

Questions, questions :)
ID: 361 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 11 Aug 09
Posts: 963
Credit: 24,557,133,931
RAC: 31,138
Message 367 - Posted: 14 May 2018, 10:13:24 UTC - in response to Message 361.  

Bit difficult experimenting with slow GPUs as they take an hour or more to complete a single unit -- so I was hoping somebody might have experience with the Nviidia Quadro range.

Or a method of working with slow GPUs?

Working with one variable at a time in a set order perhaps?

What config is used by default?

Can you dump the settings out to give a starting point?

Questions, questions :)


As I said post #71 in this thread gives all the data I know, personally I would pick one close to yours and go from there.
ID: 367 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Slicker
Project administrator

Send message
Joined: 11 Jun 09
Posts: 79
Credit: 943,644,517
RAC: 0
Message 368 - Posted: 14 May 2018, 14:25:40 UTC
Last modified: 14 May 2018, 14:26:43 UTC

One way to check the speed on various settings without having to run the entire WU is to:

1. Copy the app, to a temp folder.
2. Copy the collatz config file to the temp folder but rename it to collatz.config
3. Copy a collatz WU file to the temp folder and rename it to in.txt
4. Run the WU for 15 minutes.
5. copy stderr.txt to stderr_test_N.txt changing N to a new number each time
6. delete the boinc_lockfile
7. delete the out.txt (probably won't exist unless the WU finished)
8. delete the checkpoint.txt file
9. delete the stderr.txt file
10. edit the config and try new settings
11. go back to step 4
12. compare the new stderr to the previous one and see which reports numbers in less time e.g. 1234567890 - 123 steps @ 1:03 vs 1234567890 - 123 steps @ 0:57

For GPU apps, you will also need to have an init_data.xml file in the temp folder to tell it which GPU type and number to use. You can copy one from https://github.com/BOINC/boinc/tree/master/samples/openclapp/INIT_DATA%20test%20files

Note that when changing the sieve size, it creates a new sieve file which will be re-used on subsequent runs so the time will be reduced by 1-2 seconds on subsequent tests with the same sieve size.
ID: 368 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile IDEA

Send message
Joined: 30 May 17
Posts: 119
Credit: 37,173,545,890
RAC: 4
Message 369 - Posted: 14 May 2018, 17:44:29 UTC - in response to Message 368.  

Cheers Slicker. I'll report back once I've had a chance to run some tests.

Shame the machine is busy RIPping and printing at the moment. Pesky clients are getting in the way of number crunching :(
ID: 369 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
vseven

Send message
Joined: 24 Apr 18
Posts: 6
Credit: 1,483,437,063
RAC: 0
Message 375 - Posted: 15 May 2018, 12:21:17 UTC - in response to Message 367.  
Last modified: 15 May 2018, 12:21:34 UTC

As I said post #71 in this thread gives all the data I know, personally I would pick one close to yours and go from there.


Is there a reason they say AMD 1070, AMD 1080, AMD 1080 ti, etc? Shouldn't those all be GTX or nVidia?
ID: 375 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 11 Aug 09
Posts: 963
Credit: 24,557,133,931
RAC: 31,138
Message 380 - Posted: 16 May 2018, 20:44:32 UTC - in response to Message 375.  

As I said post #71 in this thread gives all the data I know, personally I would pick one close to yours and go from there.


Is there a reason they say AMD 1070, AMD 1080, AMD 1080 ti, etc? Shouldn't those all be GTX or nVidia?


Those are all different models of Nvidia cards, and yes that seems to b a typo.
ID: 380 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Anthony Ayiomamitis

Send message
Joined: 21 Jan 15
Posts: 14
Credit: 10,000,363,396
RAC: 0
Message 388 - Posted: 17 May 2018, 17:05:43 UTC - in response to Message 380.  
Last modified: 17 May 2018, 17:07:08 UTC

Mike,

Have you looked at the performance when running two uw's at the same time on one GPU? With my current config, my GPU is running at 99-100% efficiency but I wonder if there is better throughput with two units running simultaneously.

Not that I am complaining with the 345 seconds per unit right now (GTX 1080).
ID: 388 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4 · 5 . . . 12 · Next

Message boards : Number crunching : Optimizing the apps


©2022 Jon Sonntag; All rights reserved