All things Nvidia GPU

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6453
Credit: 9579749842
RAC: 7443635

We clearly have tools that

We clearly have tools that allow overclocking GPU's in Linux. And controlling fans speeds. But what about under voting Nvidia GPU'S in Linux?

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

Skip Da Shu
Skip Da Shu
Joined: 18 Jan 05
Posts: 152
Credit: 1043529300
RAC: 712851

Voltage controls taken out of

Voltage controls taken out of nvidia-smi (I believe by Nvidia) some time ago.

 

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3956
Credit: 46950852642
RAC: 64634797

You can’t directly control

You can’t directly control voltage in Linux. 
 

you can indirectly alter the voltage with the power limit function. Setting the power limit lower will force the GPU to run lower voltage, but you can’t really control exactly how much or at what point the card ends up pulling back voltage or core clocks. 
 

I’ve never felt that power limiting wasn’t good enough though. 

_________________________________________________________________________

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6453
Credit: 9579749842
RAC: 7443635

Tom M wrote: A solid case

Tom M wrote:

A solid case can be made that a Titan V is more "efficient" (uses less electricity for a given level of production) than a EVGA rtx 3080 ti.  But what happens to the rtx 3080 ti's production if you were to power limit it to say 200 watts?

My Titan V's are running ~150 watts or less.  If I can get the equivalent production at 200 watts on my rtx 3080 ti running the same number of threads....

Experiment has started.

 

Preliminary results seem to indicate I can get ~2.3M RAC or maybe higher.  So from a cost of electricity vs. results POV this is looking very encouraging.   The graphics clock is not nearly as stable as the Titan V clocks seem to be but otherwise it appears to be crunching right along.  And I don't have to buy a Titan V to get the results.

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

pututu
pututu
Joined: 6 Apr 17
Posts: 63
Credit: 653417392
RAC: 5

Tom M wrote:Tom M wrote:A

Tom M wrote:

Tom M wrote:

A solid case can be made that a Titan V is more "efficient" (uses less electricity for a given level of production) than a EVGA rtx 3080 ti.  But what happens to the rtx 3080 ti's production if you were to power limit it to say 200 watts?

My Titan V's are running ~150 watts or less.  If I can get the equivalent production at 200 watts on my rtx 3080 ti running the same number of threads....

Experiment has started.

 

Preliminary results seem to indicate I can get ~2.3M RAC or maybe higher.  So from a cost of electricity vs. results POV this is looking very encouraging.   The graphics clock is not nearly as stable as the Titan V clocks seem to be but otherwise it appears to be crunching right along.  And I don't have to buy a Titan V to get the results.

Tom M

Last I ran O3AS, it was producing close to 3.4M RAC at 174W average board power with single VII. I think it can do 3M RAC at 150W but I haven't tried it out yet. 

 

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6453
Credit: 9579749842
RAC: 7443635

Pututu, Your making a case

Pututu,

Your making a case for jumping to a Radeon VII. My fallback was another Titan V.

Sigh.

At least the outside temperatures have fallen.

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3956
Credit: 46950852642
RAC: 64634797

Tom M wrote:Pututu,Your

Tom M wrote:

Pututu,

Your making a case for jumping to a Radeon VII. My fallback was another Titan V.

Sigh.

At least the outside temperatures have fallen.

Keep in mind, pututu’s results were with a 7950x CPU that wasn’t running any other CPU work. You won’t get the same result with your slower EPYC CPUs and running CPU work on them too

_________________________________________________________________________

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6453
Credit: 9579749842
RAC: 7443635

Ian& SteveC,Thank you for

Ian& SteveC,

Thank you for the reminder. Just for giggles I am starting a 4x test at 200 watts power limit.

System

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

mikey
mikey
Joined: 22 Jan 05
Posts: 12689
Credit: 1839094599
RAC: 3726

Tom M wrote: Ian&

Tom M wrote:

Ian& SteveC,

Thank you for the reminder. Just for giggles I am starting a 4x test at 200 watts power limit.

System

Certainly not taking 4 times or even twice as long for each task!!

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6453
Credit: 9579749842
RAC: 7443635

mikey wrote: Tom M

mikey wrote:

Tom M wrote:

Ian& SteveC,

Thank you for the reminder. Just for giggles I am starting a 4x test at 200 watts power limit.

System

Certainly not taking 4 times or even twice as long for each task!!

I was running 5 tasks on the GPU. Now I am running 4 tasks. The graphics clock may now be averaging above the graphics clock speed reported on the Titan V's. A widely varying graphics clock is the most visible difference between a 200 watt power limit and a 300 watt power limit.

I believe I am getting a 2M+ RAC average. So I am a net winner for this experiment.

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.