Brp7/MeerKat 1x vs 2x crunching speeds

Tom M
Tom M
Joined: 2 Feb 06
Posts: 5,624
Credit: 7,713,011,738
RAC: 2,404,454
Topic 230124

Sometimes you pickup crunching speeds (eg total production) when you run 2 GPU tasks at once.

Brp7 seems to process faster at 1x on Linux with Nvidia rtx 3080 ti's.

Are other brand/model/operating systems getting the same results?

Eg. Windows. Radeon GPU's. Slower Nvidia GPU's.

Thank you.

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)

Boca Raton Community HS
Boca Raton Comm...
Joined: 4 Nov 15
Posts: 216
Credit: 8,431,541,988
RAC: 431,444

The only thing I have looked

The only thing I have looked at is that 3x on the 4090 is a little slower than 2x (by only 2 seconds though and on Linux). I have not had time to try 1x vs 2x at this point but plan to soon. 

Tom M
Tom M
Joined: 2 Feb 06
Posts: 5,624
Credit: 7,713,011,738
RAC: 2,404,454

Tom M wrote: Sometimes you

Tom M wrote:

Sometimes you pickup crunching speeds (eg total production) when you run 2 GPU tasks at once.

Brp7 seems to process faster at 1x on Linux with Nvidia rtx 3080 ti's.

Got a PM that said while the speeds between 1x and 2x were a "wash"

Now running BRP7 x2 has reduced power consumption ~13% compared to running FGRP x2 with all clocks the same.

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)

Magiceye04
Magiceye04
Joined: 18 Feb 06
Posts: 29
Credit: 783,817,765
RAC: 295,228

My Geforce 3070 runs with 2x

My Geforce 3070 runs with 2x WUs exactly twice the time compared to only one WU. (Ubuntu)

So for me there is no advantage. Also the power consumption does not differentiate (170 vs. 172W). I run the GPU now at 110W fixed and it takes only 6-7% longer then the unlimited 170W setting.

I think this is the normal behaviour on geforce GPUs. AMD need often multiple WUs running parallel, nvidia not (maybe the 4090 is the exception here).

Tom M
Tom M
Joined: 2 Feb 06
Posts: 5,624
Credit: 7,713,011,738
RAC: 2,404,454

Magiceye04 wrote: My Geforce

Magiceye04 wrote:

My Geforce 3070 runs with 2x WUs exactly twice the time compared to only one WU. (Ubuntu)

So for me there is no advantage. Also the power consumption does not differentiate (170 vs. 172W). I run the GPU now at 110W fixed and it takes only 6-7% longer then the unlimited 170W setting.

I think this is the normal behaviour on geforce GPUs. AMD need often multiple WUs running parallel, nvidia not (maybe the 4090 is the exception here).

It sounds like I need to revisit my 2x testing.  I would swear the run time was slightly more than doubled.  :)

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3,700
Credit: 34,353,996,994
RAC: 38,110,286

it likely varies with card

it likely varies with card used. like all things "it depends" so you wont get a result that holds true for all platforms/hardware, and that's subject to change if the apps change or the data files change.

AMD historically has benefited from multiples, while Nvidia has not (or not much). that probably still holds true here. something very powerful like the 4090 might see benefit if a single task cannot use the whole GPU effectively due to sheer size/power.

_________________________________________________________________________

Tom M
Tom M
Joined: 2 Feb 06
Posts: 5,624
Credit: 7,713,011,738
RAC: 2,404,454

Has anyone been able to come

Has anyone been able to come up with a combination of e@h tasks that produces a higher RAC than Brp7/MeerKat?

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)

GWGeorge007
GWGeorge007
Joined: 8 Jan 18
Posts: 2,782
Credit: 4,588,428,885
RAC: 3,042,337

Tom M wrote: Has anyone been

Tom M wrote:

Has anyone been able to come up with a combination of e@h tasks that produces a higher RAC than Brp7/MeerKat?

I just don't get why you are so obsessive with higher RAC values when you won't leave your computers alone long enough to achieve the high RAC you aspire to.  You keep changing things all the time trying to achieve a higher RAC and then don't let them run long enough to do so.  In the process of changing things you are constantly inhibiting your computers from achieving your goals.

Will you ever learn that leaving a computer alone long enough, like a month or more, can actually have a benefit in RAC?

I tend to leave mine alone and apply a goal of achieving a high total credit score, but then will also look at the RAC values.

You must remember that RAC is a function of:


Computing the current value of Recent Average Credit

BOINC updates 'recent average credit' (RAC) only when new credit is granted. Interfaces that export RAC also export that time at which it was last updated. To obtain the current value of RAC, you must 'decay' it based on the time that has elapsed since it was updated:

function decay_average($avg, $avg_time, $now = 0) {
   $M_LN2 = 0.693147180559945309417;
   $credit_half_life = 86400 * 7;
   if ($now == 0) {
       $now = time();
   }
   $diff = $now - $avg_time;
   $weight = exp(-$diff * $M_LN2/$credit_half_life);
   $avg *= $weight;
   return $avg;
}

If you don't apply this decay, inactive entities will have incorrectly high RAC.


 

George

Proud member of the Old Farts Association

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 4,720
Credit: 17,646,040,754
RAC: 6,355,996

GWGeorge007 wrote: I just

GWGeorge007 wrote:

I just don't get why you are so obsessive with higher RAC values when you won't leave your computers alone long enough to achieve the high RAC you aspire to. 

+100

 

Paul
Paul
Joined: 3 May 07
Posts: 121
Credit: 1,654,345,650
RAC: 16,469

I don't think these are

I don't think these are unreasonable statements.  After FGRPB1G ended, my RAC tanked and that means I have to make a change to my configuration.  But, being interested in RAC doesn't mean waiting a month after every change.  You have to calculate your personalized credit/s rate by looking at your validated tasks and doing a little math.  It doesn't much matter what is the algorithm for calculating RAC, you're not trying to trick it or beat the system.  Whatever it is, you have to wait for it to settle.  Fine.  So, in the short term, you can calculate your own version of RAC based on data in the validated tasks list.  It still takes a few days get get a reasonable sample of WUs validated, but you cannot test every combination of configurations anyway, so I don't actually see another way to do it or any point waiting for RAC to stabilize.  If you want more credit, just look at your credit.  You can figure it out. RAC changes just tells you whether or not you are doing the right thing...a month later.  But, that's not the only measurement you have; you can see the credit you're getting and how long it took to complete that work.  That's literally what counts.  It's not rocket sci...well, I guess technically it is astrophysics, but you get my meaning; your BONIC credit is not difficult to track.  And, that fine tuning you do after RAC settles is accomplished the exact same way, so, why not do that now and get a feel for the numbers, make some records.

Tom M
Tom M
Joined: 2 Feb 06
Posts: 5,624
Credit: 7,713,011,738
RAC: 2,404,454

GWGeorge007 wrote: I just

GWGeorge007 wrote:

I just don't get why you are so obsessive with higher RAC values

Or lower RAC values since I have been stare-ring at the bottom RAC of the top 50 list too.

Since George bought a dual CPU MB for the purpose of possibly becoming very competitive in the universe at home project.  And an rtx 3090 hybrid to possibly compete with Freewill's rtx 3090 on e@h tasks....

I claim "kettle, pot, black"  ;)

Tom M

 

 

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)

Comment viewing options

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