• We would like to remind our members that this is a privately owned, run and supported forum. You are here at the invitation and discretion of the owners. As such, rules and standards of conduct will be applied that help keep this forum functioning as the owners desire. These include, but are not limited to, removing content and even access to the forum.

    Please give yourself a refresher on the forum rules you agreed to follow when you signed up.

Block CPU Usage Values

Patzag

Fractal Fanatic
Here is a quick compilation of CPU usage by Block.
Nothing super scientific. I suspect one of the CPUs handles the amp as the usage is barely above that of a volume control!

Input
1.5

Output
0.6

Shunt (1)
0.2

Shunt (10)
1.9

Amp
0.6

Amp 2
0.6

Cab - 1 Active
5.2

Cab - 2 Active
9.7

Cab - 3 Active
14.0

Cab - 4 Active
18.2

Chorus
3.4

Compressor
1.8

Crossover
0.7

Delay
3.6

Drive
6.8

Enhancer
1.0

Filter
0.4

Flanger
2.1

Formant
1.0

Gate
1.4

send/return
0.4

GEQ
0.5

Looper
3.8

Megatap Delay
5.4

Midi
0.2

Mixer
0.3

MultiDelay
3.5

Multiband Comp
5.1

Multiplexer
0.5

PEQ
0.6

Phaser
1.5

Pitch
3.0

Pitch Quad Harmony/Delay
8.4

Plex Delay
2.7

RTA
0.2

Resonator
3.8

Reverb low
7.7

Reverb High
10.5

Ring Mod
1.2

Rotary
4.3

Synth
3.3

10-tap delay
2.0

Tonematch
7.5

Trem/Pan
0.8

Vocoder
6.3

Vol/Pan
0.4

Wah
1.2
 

trancegodz

Power User
Thanks! I did that for everything in the Axe FXII a long time ago. I did it for every type of every effect because they often used different amounts of CPU. Especially the different types of reverb, different types of drive blocks, etc. Spring reverb used less CPU that other types of reverbs for instance.
 

∞Fractals

Power User
How can the amp block just be 0.6? In the Axe II a whole Tigershark DSP was dedicated to amp modeling iirc.???

Are these blocks processing any audio? My Understanding is that CPU usage goes up depending on input signal ... ???

Thanks for providing this.
 

Patzag

Fractal Fanatic
The cab block can hold up to 4 IRs. The processing increases as you add IRs so I showed here how much the cab uses with 1, 2, 3 or 4 cabs activated.
The amp block is probably not included in the CPU usage. I think what we are seeing here is most likely a single CPU - with the other handling the amps. But that's just conjecture.
 

Patzag

Fractal Fanatic
Thanks! I did that for everything in the Axe FXII a long time ago. I did it for every type of every effect because they often used different amounts of CPU. Especially the different types of reverb, different types of drive blocks, etc. Spring reverb used less CPU that other types of reverbs for instance.
I included the min/max values. For example hi and low quality reverbs, Whammy and quad harmony pitch, etc. I don't have time to do every permutation. Sorry! :D
 

Dr. Dipwad

Inspired
So, just to confirm...

It looks to me like...
...if you started with Input 1 and Output 1;
...and between them in parallel, you put Amp 1 and Amp 2 on the grid, simultaneously;
...and, after the amps, you also had Cab 1 and Cab 2 on the grid, simultaneously;
...and, before the amps, you also had Drive 1 and Drive 2 on the grid, simultaneously;
...and you added a (single) Reverb at the end;

...there would still be...what? More than 50% of the CPU still free for adding more blocks?

Do I have that right?

(I'm trying to gauge just how much CPU the Axe FX III has, even when 2 instances of some of its heaviest blocks are engaged.)
 

unix-guy

Legend!
I assume that similar to the Axe Fx II the CPU consumption may vary a lot depending on which Drive model is in use, and similar for a few others. But thanks for posting!
 

Patzag

Fractal Fanatic
So, just to confirm...

It looks to me like...
...if you started with Input 1 and Output 1;
...and between them in parallel, you put Amp 1 and Amp 2 on the grid, simultaneously;
...and, after the amps, you also had Cab 1 and Cab 2 on the grid, simultaneously;
...and, before the amps, you also had Drive 1 and Drive 2 on the grid, simultaneously;
...and you added a (single) Reverb at the end;

...there would still be...what? More than 50% of the CPU still free for adding more blocks?

Do I have that right?

(I'm trying to gauge just how much CPU the Axe FX III has, even when 2 instances of some of its heaviest blocks are engaged.)
I started with a shunt. Anything put in the grid adds about 5% CPU.
Then yes, I put an input, and made an excel grid that measured the difference in CPU when adding a block. I kept adding until I ran out of CPU, then I cleaned up and started again. A bit tedious, especially because some blocks CPU vary depending on settings, but the result is enough to get an approximation of what blocks use up gobs of memory. Like the Cab. That's the big bad boy if you use 4 IRs.

On the other side, I have not made a preset yet -to be actually used - which uses all the CPU. This was more for edification and sharing than to solve a problem. There's lots of power in this machine!
 

Patzag

Fractal Fanatic
As a note, Axe Edit does not poll CPU increase when adding IRs to CABS or changing models of delays. I had to look on the hardware for that.
 

haffner1

Inspired
How can the amp block just be 0.6? In the Axe II a whole Tigershark DSP was dedicated to amp modeling iirc.???

Are these blocks processing any audio? My Understanding is that CPU usage goes up depending on input signal ... ???

Thanks for providing this.
My guess is that the percentage of DSP shown is only that of one processor. Like the axe II, 1 DSP is dedicated to amp modeling only. That .6% is just what it takes to bridge back and forth between the DSPs.
 

yek

Moderator
Moderator
To add to Patzag's list:

I tested all Drive blocks in the III, for the update of my Drive Guide. Here you go. Numbers have been rounded.

BB Pre — 8
Bender Fuzz — 5
Bit Crusher — 7
Blackglass 7k — 5
Blues OD — 5
ACB — 7
RCB — 7
Eternal Love — 8
Face Fuzz — 7
FAS Boost — 8
LED-Drive — 8
Fat Rat — 7
FET Boost — 5
FET Preamp — 5
Full OD — 7
Hard Fuzz — 4
M-Zone Dist — 7
Master Fuzz — 7
Micro Boost — 5
Mid Boost — 7
Octave Dist — 5
Pi Fuzz — 7
Plus Dist — 7
Rat Dist — 7
Ruckus — 7
SDD — 5
Shimmer — 7
Shred Dist — 7
Super OD — 8
T808 MOD — 8
T808 — 8
Tape Dist — 7
Timothy — 7
Treb Boost— 7
Tube 3-knob — 6
Tube 4-knob — 6
Zen Master — 8
 

Dr. Dipwad

Inspired
Hey guys,

I'm not quite sure whether the answer to my earlier question was "yes" or "no."

Here's a slight variation of the question, (but it comes down to the same underlying idea):

THE QUESTION:
If I had an Axe III, and I build a Preset containing:
Input 1
Drive 1 - with a heavier-than-usual model selected
Amp 1 - with a heavier-than-usual model selected
Cab 1 - with all 4 active
Drive 1 - with a different heavier-than-usual model selected
Amp 2 - with a different heavier-than-usual model selected
Cab 2 - with all 4 active
Reverb 1 - with a pretty big reverb selected
Output 1

...about how much CPU would I have left?

MY GUESS AT THE ANSWER:
From what you've said thus far, it seems that the items on the grid would "weigh" as follows:
(having anything on the grid at all) - 5%
Input 1 - 1.5%

Drive 1 - Zen Master, 8%
Amp 1 - 0.6%...? Really?! Is that right?
Cab 1 - with all 4 active, 18.2%

Drive 1 - Eternal Love, 8%
Amp 2 - 0.6%...? Really?! Is that right?
Cab 2 - with all 4 active, 18.2%

Reverb 1 - 10.5%

Output 1 - 0.6%

Adding all that up, it looks like it comes to 71.2% used.

BUT, somehow I doubt that the real weight of those Amp models is 0.6% each.

So, what's the real truth of it? Am I failing to account for something?
 

Dr. Dipwad

Inspired
See post #9 above,
Rex,

Are you saying that...

(a.) in the Axe FX III, an entire chip is dedicated to the Amp modeling whether or not there is an Amp Block on the grid;

...and,

(b.) in the Axe FX III, the CPU burden of the chip that handles Amp modeling is not counted as part of the "available CPU" (because it's wholly used for Amp modeling even when there's no Amp Block on the grid);

...therefore,

(c.) when you do add an Amp Block to the grid, the only increase in CPU you see is 0.6% ...because that's the amount it takes to get the signal path out of the "everything else" chip, over to the "amp modeling only" chip, and back?

Is all of that correct?
 
Top Bottom