Optimizing my FX III via CPU Block usage

[Moondog Wily]

I must first of all thank @Yerbluesrob for getting my mind spinning on this topic! He had posted a thread ( Unused channels for CPU economy? ) asking about using lower CPU drawing blocks when they are in bypass mode. This got my head spinning on how I could optimize my kitchen sink preset, allowing more room for other things I want to do with my Axe FXIII MkII Turbo. Twenty four hours later, I have savings in my preset ranging from 4.4% to 9.9% depending on the scene! Wooooo Hooooo!

Others have mentioned that having a CPU usage by block type chart would be handy (see for one, this thread - Is there a chart showing approximate CPU usage by block type? ) and I agree wholeheartedly! I decided to make one for the Axe FXIII MkII Turbo!! I am about half way done, here is the link to view what I have so far: Axe FXIII CPU Usage per Block. If you would like to contribute to flushing this out and/or starting one for other units (FM3, FM9, non Turbo FXIII, etc.), PM me a email address and I will set you up to be able to edit this spreadsheet (I will always have my own backup and between that and history, I will be able to control any changes that are made so don't worry about messing anything up). It took me about 3-4 hours to get up to MultiTap Delay of the blocks in alphabetical order. Here are some details and notes (more in the various tabs of the spreadsheet).

1. This data was gathered for/from an Axe FXIII MkII Turbo running 19.07FW. Because these are CPU values, the values will not directly translate to other units. You can still discern which blocks use more CPU in general, but the number will vary from what is displayed here.
2. My CPU usage values with a blank grid are 3.6% at it's lowest and 4.9% at it's highest.
3. I used an empty preset, and placed a single block on the grid at a time. I then observed the CPU usage over the course of 10-15 seconds, noting the highs and lows of the CPU values over that time frame.
4. If a block has "types" available, I cycled through them in most cases. The two exceptions are, I did NOT cycle through all amps or all cabs (tried a dozen or so different amps and saw no significant changes in CPU%, cabs just too many for this first go round but only changes I saw were changing number of cabs, going from standard to ultra to full res and the IR Length setting for those cabs), but I intend on cycling through all other blocks different types and have done that in the ones I have tested thus far. I did NOT change any settings on any blocks, using only the factory default values for these tests. I DID test on the Wah whether adding LFO to enable auto wah, and it seemed to have zero effect on the CPU usage.
5. In the spreadsheet, each tab (along the bottom) represents a block type (with a few other non block type tabs as well). In each of those worksheets there are columns for the "type" chosen for that block (some don't have such choices as you know), Low CPU % value, high CPU % value, Low Adjusted CPU value, High Adjusted CPU value and Description. The Low Adjusted and High Adjusted use a formula to reference the "Empty Preset" (second tab in spreadsheet) CPU usage values, which get subtracted from the Low/High CPU values, resulting in an approximation of the actual CPU usage for that block and the selected type.

Keep in mind, these are estimates. I do believe however it can create a valuable tool in building efficient presets. As mentioned above, I had significant CPU savings in the first pass at this based on choosing block types that had lower CPU draw. This can be done for blocks that are active and for blocks that are bypassed. The original intent was to simply set the bypassed blocks to a channel with the lowest CPU draw for that block, but it will also help in choosing a active block type when you want to reduce your CPU usage.

Attached are my presets before (MW 1_2022May07) and after (MW 1_2022May13) if you would like to review where I saved the CPU%'s.

Feel free to let me know if you see other ways I can reduce my CPU usage as I am doing this in part to allow me to connect all 3 of my mics to the Axe in my one man band show! These presets have one mic connected via input 2 which is in rows 5 and 6, so feel free to ignore those and/or remove them if you would like to play with this preset.

Also, any thoughts on how to make this better or more complete are of course appreciated! If you think I am wasting my time, please save your time and move on!! Not interested in the naysayers tripe!!!

Hope others can find in this work the advantages I have garnered already in a very short period of time! May your optimizations bring you a thousand rays of golden sunlight (and a cooler CPU with room to add more) ;~))

 

[sprint]

Here's a version with a few changed bits to save 7%

• Cab resolution from Max to 512 (big savings - try 1024 if going to 512 is audible to u).
• Shaved a bit off reverbs density (8 to 7 for a tad less cpu)
• Changed gates from Intelligent to Classic - saves a smidge.
• Remove a shunt / save a pinch.

Other Possibilities:

• Turn any unused global EQs off.
• Turn any unused gates off. (input 2?).
• If u can, use the amp's input dynamics knob instead of comp1 in pre position (or amp's output compressor instead of a comp in post position (not a substitute if u want to comp your drive)).
• Use Pedal Compressors if possible.
• Reverb density or quality down more (I go to the lowest I find minimally audible listening with reverb mix at 100%).
• If possible, use Cab block hi/lo cut instead of a filter block at end of chain.

 

[Moondog Wily]

Here's a version with a few changed bits to save 7%

• Cab resolution from Max to 512 (big savings - try 1024 if going to 512 is audible to u).
• Shaved a bit off reverbs density (8 to 7 for a tad less cpu)
• Changed gates from Intelligent to Classic - saves a smidge.
• Remove a shunt / save a pinch.

Other Possibilities:

• Turn any unused global EQs off.
• Turn any unused gates off. (input 2?).
• If u can, use the amp's input dynamics knob instead of comp1 in pre position (or amp's output compressor instead of a comp in post position (not a substitute if u want to comp your drive)).
• Use Pedal Compressors if possible.
• Reverb density or quality down more (I go to the lowest I find minimally audible listening with reverb mix at 100%).
• If possible, use Cab block hi/lo cut instead of a filter block at end of chain.

Thanks @sprint , I will look at utilizing those recommendations! Your insights are very much appreciated!! And thanks for the modified Preset, truly above and beyond kind!!!

 

[fuzzyoctaves]

This is a much needed project! My (admittedly complex) kitchen sink preset is at 82% and I am always tweaking it to have it do more.

I would think that a separate thread entitled: "Axe FXIII CPU Usage per Block" would be best to focus the discussion. And when it is done, it should go up on the Wiki.

I'd be happy to help, but before I dig in, I am curious how the methodology stands up to the scrutiny of the super experts on here.
Also - there shouldn't be any difference between turbo and non as far as CPU % goes right?

 

[Moondog Wily]

This is a much needed project! My (admittedly complex) kitchen sink preset is at 82% and I am always tweaking it to have it do more.

I would think that a separate thread entitled: "Axe FXIII CPU Usage per Block" would be best to focus the discussion. And when it is done, it should go up on the Wiki.

I'd be happy to help, but before I dig in, I am curious how the methodology stands up to the scrutiny of the super experts on here.
Also - there shouldn't be any difference between turbo and non as far as CPU % goes right?

I totally agree on wanting input from the experts! I also believe that the greatest expert on these products most likely already has all this data, so a dump from the source would be awesome if it does exist!! As for it being a part of the Wiki, agree again, and maybe even part of the Blocks details PDF from FAS!!! Thanks for your enthusiasm and support for this work, I know we are not alone in wanting such a dataset ;~))

 

[Moondog Wily]

Here's a version with a few changed bits to save 7%

• Cab resolution from Max to 512 (big savings - try 1024 if going to 512 is audible to u).
• Shaved a bit off reverbs density (8 to 7 for a tad less cpu)
• Changed gates from Intelligent to Classic - saves a smidge.
• Remove a shunt / save a pinch.

Other Possibilities:

• Turn any unused global EQs off.
• Turn any unused gates off. (input 2?).
• If u can, use the amp's input dynamics knob instead of comp1 in pre position (or amp's output compressor instead of a comp in post position (not a substitute if u want to comp your drive)).
• Use Pedal Compressors if possible.
• Reverb density or quality down more (I go to the lowest I find minimally audible listening with reverb mix at 100%).
• If possible, use Cab block hi/lo cut instead of a filter block at end of chain.

Played around with your optimized version of my preset just now and it is great! Looks like 5-6% of that 7% CPU savings was the IR Length at 512 rather than max!! Sounds a bit "crunchier" to me at 512, which is good in some cases, not so in others!!! Did a couple of other adjustments including your suggestion of using Amp compressor rather than Comp block and I now have it down to about 50% for some of the scenes which is KILLER!!! Thanks so much for your input @sprint , very much appreciated! My spreadsheet is complete for first pass if you want to review it! I will dig deeper eventually, but for now, I think it has a fair amount of good data that can help users save many %'s in their CPU draw!

 

[Sauro_on_guitar]

I can't check your preset on my Axe III now but as a "trick" (not a secret) to have less cpu usage is known that shaunts use cpu as well... so less shaunts as possible!

 

[TSJMajesty]

Also - there shouldn't be any difference between turbo and non as far as CPU % goes right?

From the Wiki:
"A TURBO module is available as an option when buying a new Mark II. This provides a 25% higher clock speed allowing for more complex presets which would exceed the limits of the standard DSP module."

Edit: I may have misunderstood. Maybe you meant that the amount of CPU used per block would be the same between Turbo and non-Turbo models?

 

[Moondog Wily]

I can't check your preset on my Axe III now but as a "trick" (not a secret) to have less cpu usage is known that shaunts use cpu as well... so less shaunts as possible!

Thanks for the thoughts Sauro_on_guitar and yes, I am hip to the shunts CPU draw! @sprint did an optimization on my preset and shaved an additional 7% (5-6% of that being Cab IR Length at 512 rather than Max) on top of the 10% I had gotten by using lower CPU% type blocks when one was bypassed! Then I shaved about 3% more off using some other techniques including suggestions from @sprint , so I am now around 50-54% for the 8 scenes in my kitchen sink preset!! Totally psyched! Now I have the CPU to add my other two mics which will not draw a whole lot of processing power, but will allow me to eliminate the mix board from my kit assuming all goes as planned!!

 

[Moondog Wily]

I have completed my rookie pass of all Blocks on the Axe FXIII MkII FW 19.07. Here is the link: Axe FXIII CPU Usage per Block. There is of course still plenty to do on this, including getting into some of the variations within certain blocks that affect CPU usage, but I think this is a good start in seeing the overall picture of Block CPU usage in a structured format! I am now down about 20% on my CPU usage!!!

 

[yek]

I would think that a separate thread entitled: "Axe FXIII CPU Usage per Block" would be best to focus the discussion. And when it is done, it should go up on the Wiki.

The wiki has always contained links to such user-created charts. But the reality is that the charts get outdated quickly, aren't updated and loose their value, so they get removed from the wiki.

 

[yek]

Also - there shouldn't be any difference between turbo and non as far as CPU % goes right?

Percentages will be different.