Advanced search

Message boards : Number crunching : GTX 1080 hovers around 80-82% utilization

Author Message
wolfman1360
Send message
Joined: 19 Feb 17
Posts: 5
Credit: 36,563,552
RAC: 0
Level
Val
Scientific publications
wat
Message 52588 - Posted: 6 Sep 2019 | 22:54:19 UTC

Hello, everyone.
I've got an i7-7700 crunching gpugrid along with world community grid.
I have the CPU percentage set to give gpugrid ample cores (85% usage, just in case gpugrid needs a little more than one thread) for the 1080.
Nevertheless I'm getting the same utilization if I bump it down a little from 87. Even when I suspend cpu tasks it doesn't seem to change anything.
In reading, there appears to be an environment variable, but does this still apply and how would I go about applying it if so?
I am running Windows 10 pro x64. The machine is running on an SSD with 64 GB of ram.
The upside of this is I am able to run at higher clock speeds, but I'm not sure if I can ever make that equal to 20% of my gpu's utilization.
Any help appreciated.

rod4x4
Send message
Joined: 4 Aug 14
Posts: 266
Credit: 2,219,935,054
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 52590 - Posted: 6 Sep 2019 | 23:33:19 UTC - in response to Message 52588.
Last modified: 6 Sep 2019 | 23:42:33 UTC

This utilization is about right for your setup.
You can add swan_sync=1 to your Windows environment, this will increase utilization a bit, but wont take it to 100%
This guide describes accessing Environment variables in windows:
https://www.architectryan.com/2018/08/31/how-to-change-environment-variables-on-windows-10/
You need to ADD the variable "swan_sync" (without quotes)
Set the value to 1
swan_sync will dedicate 1 CPU to each GPU task you run.
GPU utilization on Linux with swan_sync will reach 100%.
Otherwise you can run two task at once. Searching the forum you will find plenty of guides on setting that up.
Stopping CPU tasks will improve the GPU task runtime, but wont increase GPU utilization.
Personally, I run both CPU tasks and single GPU task on same PC with swan_sync enabled on both Windows and Linux hosts.

wolfman1360
Send message
Joined: 19 Feb 17
Posts: 5
Credit: 36,563,552
RAC: 0
Level
Val
Scientific publications
wat
Message 52591 - Posted: 7 Sep 2019 | 0:27:08 UTC - in response to Message 52590.

This utilization is about right for your setup.
You can add swan_sync=1 to your Windows environment, this will increase utilization a bit, but wont take it to 100%
This guide describes accessing Environment variables in windows:
https://www.architectryan.com/2018/08/31/how-to-change-environment-variables-on-windows-10/
You need to ADD the variable "swan_sync" (without quotes)
Set the value to 1
swan_sync will dedicate 1 CPU to each GPU task you run.
GPU utilization on Linux with swan_sync will reach 100%.
Otherwise you can run two task at once. Searching the forum you will find plenty of guides on setting that up.
Stopping CPU tasks will improve the GPU task runtime, but wont increase GPU utilization.
Personally, I run both CPU tasks and single GPU task on same PC with swan_sync enabled on both Windows and Linux hosts.


Thank you very much. This info is appreciated.

rod4x4
Send message
Joined: 4 Aug 14
Posts: 266
Credit: 2,219,935,054
RAC: 0
Level
Phe
Scientific publications
watwatwatwatwatwatwatwatwatwat
Message 52595 - Posted: 7 Sep 2019 | 7:32:23 UTC - in response to Message 52591.

Thank you very much. This info is appreciated.


Your welcome.

Profile Retvari Zoltan
Avatar
Send message
Joined: 20 Jan 09
Posts: 2346
Credit: 16,293,065,968
RAC: 7,113,235
Level
Trp
Scientific publications
watwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwatwat
Message 52597 - Posted: 7 Sep 2019 | 12:59:12 UTC - in response to Message 52588.

In reading, there appears to be an environment variable, but does this still apply and how would I go about applying it if so?

It does still apply.
Here's the guide. (you may have to click two times to get directly to my post)

Post to thread

Message boards : Number crunching : GTX 1080 hovers around 80-82% utilization

//