Running multiple tasks concurrently per GPU - How to revert to just one task.

maeax
maeax
Joined: 6 Aug 12
Posts: 21
Credit: 1413446572
RAC: 2425994

Have only WCG for

Have only WCG for CPU's.

Overloading is not seen so far. Yes, the hardware must be carefully balanced.

Two O3AS (one for each GPU) have the same performance as those four with .5 GPU.

Never thought,  it would be the same performance.

More GPU-use is not running and not planned.

 

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 3060
Credit: 4961384353
RAC: 1395568

maeax wrote: Gary, have no

maeax wrote:

Gary,

have no Hyperthreading.

You do have Hyperthreading available to you.  I don't think you need to do anything about it, just use them.

 

George

Proud member of the Old Farts Association

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3945
Credit: 46589342642
RAC: 64192692

GWGeorge007 wrote: maeax

GWGeorge007 wrote:

maeax wrote:

Gary,

have no Hyperthreading.

You do have Hyperthreading available to you.  I don't think you need to do anything about it, just use them.



he has intentionally disabled SMT. which is a choice he's made for whatever reason. either for speed in whatever use-case, or for security threat mitigation.

_________________________________________________________________________

maeax
maeax
Joined: 6 Aug 12
Posts: 21
Credit: 1413446572
RAC: 2425994

Sorry for the

Sorry for the misunderstanding.

HT is for me disabled. Of course, this CPU can do it.

The more Performance of HT is about 30%.

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 3060
Credit: 4961384353
RAC: 1395568

Ian&Steve C. wrote: he has

Ian&Steve C. wrote:

he has intentionally disabled SMT. which is a choice he's made for whatever reason. either for speed in whatever use-case, or for security threat mitigation.

Okay, I was unaware of SMT being able to disable hyperthreading.  But...  how did you know that?  I wasn't able to see his BIOS, assuming that is where it is disabled.

George

Proud member of the Old Farts Association

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3945
Credit: 46589342642
RAC: 64192692

GWGeorge007 wrote:Ian&Steve

GWGeorge007 wrote:

Ian&Steve C. wrote:

he has intentionally disabled SMT. which is a choice he's made for whatever reason. either for speed in whatever use-case, or for security threat mitigation.

Okay, I was unaware of SMT being able to disable hyperthreading.  But...  how did you know that?  I wasn't able to see his BIOS, assuming that is where it is disabled.



SMT = Hyperthreading. it's the same thing. AMD calls it "SMT", Intel calls it "Hyperthreading" and is their trademark. It's just marketing.

I could tell because BOINC is reporting his host as having "64 processors". if SMT was enabled it would say "128 processors". I made the assumption that he hadn't messed with any BOINC settings to affect that reported value, and it seemed most likely that he had turned off SMT.

_________________________________________________________________________

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 3060
Credit: 4961384353
RAC: 1395568

Thanks Ian!  Makes sense.

Thanks Ian!  Makes sense.

George

Proud member of the Old Farts Association

KLiK
KLiK
Joined: 1 Apr 14
Posts: 60
Credit: 426006605
RAC: 1374916

Has anybody started to use

Has anybody started to use multiple All-Sky Gravitational Wave search on O3 v1.07 () windows_x86_64 in order to use GPU 100%?

As these "analysis & saves" around 50% & 100% take quite some time to do it with only CPU time! As it take ~20min for crunching on GPU & ~15min for CPU analysis.

What am I talking about? This, 12min gap of only using CPU:

& how does your app_config.xml looks like that multiple app crunching? ????

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4963
Credit: 18697479698
RAC: 6213632

You need to try to stagger

You need to try to stagger the task startups on the gpu so that they all don't converge at the 49% and 99% cpu offload points when running gpu task multiples.

Just suspend a task that starts anew on the gpu until the previous one had made progress halfway to one of its cpu compute intervals, then un-suspend the task.

 

KLiK
KLiK
Joined: 1 Apr 14
Posts: 60
Credit: 426006605
RAC: 1374916

Keith Myers wrote: You need

Keith Myers wrote:

You need to try to stagger the task startups on the gpu so that they all don't converge at the 49% and 99% cpu offload points when running gpu task multiples.

Just suspend a task that starts anew on the gpu until the previous one had made progress halfway to one of its cpu compute intervals, then un-suspend the task.

Thank you.

Immediate result can be seen:

Comment viewing options

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