All blocks muted, but CPU maxed

woodslanding

woodslanding

Member
Muting blocks sometimes doesn't seem to bring CPU down at all. Other scenes in this preset function fine.... what's going on? I tried muting blocks to figure out what was going on, and ended up here:fm3CPU llimit.png
 
woodslanding said:
Muting blocks sometimes doesn't seem to bring CPU down at all.
Click to expand...

The manual says
Blocks basically use the same amount of CPU whether they are engaged or bypassed […]
Click to expand...

and the Wiki says
Important: bypassing a block on the grid does NOT lower CPU usage. Switching a block between X/Y or changing channels may affect CPU usage, depending on the parameter values. Removing a block will decrease CPU usage.
Click to expand...

Aleksi said:
Muting doesn't reduce cpu use. The blocks have to reserve the cpu needed so that when you unmute the block, it won't overload the cpu.
Click to expand...
There’s no “reserving” of the CPU.

As I understand it, blocks process the sound always, even when they are bypassed so they are immediately able to output the sound they’re responsible for when enabled. If they didn’t the amp, pitch, flanger, chorus, delay and reverb blocks would have to listen until they had a chance to process the appropriate amount of sound before beginning to output it, which wouldn’t be very useful, especially for long delays and reverb. Different models in the blocks can use different amounts of CPU, depending on how the code works, so the processing load can vary somewhat when switching scenes but the blocks are still running and processing, they just might not be outputting the sound.
 
Greg Ferguson said:
There’s no “reserving” of the CPU.

As I understand it, blocks process the sound always, even when they are bypassed so they are immediately able to output the sound they’re responsible for when enabled. If they didn’t the amp, pitch, flanger, chorus, delay and reverb blocks would have to listen until they had a chance to process the appropriate amount of sound before beginning to output it, which wouldn’t be very useful, especially for long delays and reverb. Different models in the blocks can use different amounts of CPU, depending on how the code works, so the processing load can vary somewhat when switching scenes but the blocks are still running and processing, they just might not be outputting the sound.
Click to expand...
Well, that's splitting hairs a bit I think. As far as the user is concerned, they can think of it as reserving the CPU.
 
ConstantM said:
also empty shunts still consume dsp. try getting rid of them if you dont plan to put a block there
Click to expand...
I actually tried to verify this on the FM3, and didn't notice any difference when I removed shunts. I'll have to check again, but I wonder if this is still true for this generation?
 
Greg Ferguson said:
If they didn’t the amp, pitch, flanger, chorus, delay and reverb blocks would have to listen until they had a chance to process the appropriate amount of sound before beginning to output it, which wouldn’t be very useful, especially for long delays and reverb.
Click to expand...
Whether this occurs depends on the bypass mode, so it's more like: The effect algorithm keeps performing its math on something, and that "something" being all zero-amplitude samples vs. anything else has little to no impact on CPU usage.
 
That doesn't seem like a large preset. Start by removing the shunts. Then if that doesn't donit check the reverb quality. If that doesn't do it, remove the plex delay.
 
Dickie Fredericks said:
That doesn't seem like a large preset. Start by removing the shunts. Then if that doesn't donit check the reverb quality. If that doesn't do it, remove the plex delay.
Click to expand...

Reverb quality is the first place I'm always going if I need more power. In fact, I think I pretty much always keep it at the lowest quality settings. But I'm not a reverb junkie...
 
Yeah, check reverb quality and plex type/amount of delays. The shimmer verbs for example use crazy cpu in the plex block, and I actually think the econo versions sound better. I will definitely compromise almost anywhere except reverb quality, unless I have to.
 
Aleksi said:
I actually tried to verify this on the FM3, and didn't notice any difference when I removed shunts. I'll have to check again, but I wonder if this is still true for this generation?
Click to expand...
It’s still true. They use a very small amount, less than 1% individually; I think it came out to 0.20 or 0.25% each when I measured. I only use them when necessary to make a parallel path line up correctly, never for cosmetic reasons.

There’s a point where swapping long chains of shunts with the Feedback blocks makes sense because the Feedback blocks can traverse more grid slots without additional CPU consumption.
 
Will Chen said:
Reverb quality is the first place I'm always going if I need more power. In fact, I think I pretty much always keep it at the lowest quality settings. But I'm not a reverb junkie...
Click to expand...
^Same here^

This is where the FM9 has a major advantage because Reverb is in a dedicated core so it doesn’t affect the CPU %. I still keep it at Economy though because I use the same presets on all three units.

This is from the FM3 manual in the Preset section:

The maximum allowable load is approximately 80%. There are safeguards to prevent you from pushing the FM3 too far. Should CPU levels rise above the allowable limit, the FM3 will disable sound processing and flash the warning, “CPU LIMIT - Muted”. This allows you to remove blocks or change settings to solve the problem.

Another safeguard is that the FM3 will prevent you from inserting a block that would push the CPU over the limit. If this happens, you can make changes to reduce the current CPU load and try again. You might remove an effect that is less important. Adjusting certain parameters can also help.

Here are a few of the common block parameters whose settings have a notable effect on CPU usage:
◽ Amp: Output Comp: Set to "0" for the least CPU usage.
◽ Cab: Mute: Using two IRs demands more CPU than using one. Set the mic Preamp Type to NONE and set Room Level to 0%.
◽ Reverb: Quality and Echo Density. Lower = less CPU.
◽ Compressor: Set Type to one of the "PEDAL" options to use less CPU.
◽ Phaser: Stages. Lower = less CPU.
◽ Filter: Order and Q. Lower = less CPU.
◽ Multitap Delay: Type. Different types have different CPU requirements.
◽ Plex Delay: Number of Delays affects CPU usage.
◽ Synth: Turn extra voice type parameters OFF to conserve CPU
◽ Modifiers also affect CPU usage.

TIP: The Fractal Audio Wiki keeps a list of CPU saving ideas.
Click to expand...
 
Aleksi said:
I actually tried to verify this on the FM3, and didn't notice any difference when I removed shunts. I'll have to check again, but I wonder if this is still true for this generation?
Click to expand...
Shunts do use CPU so minimize unneeded ones, and you can also disable the Global EQ's if they're not needed to save a bit, plus you can try using the Multi-Delay in Diffusor Mode for a vintage Reverb effect instead of the CPU hungry Reverb Block. Oh yeah, Standard IR's vs Ultra-Rez saves a bit too.

This kitchen sink preset runs on average at 75%, though it uses a UA OX Stomp in the FX Loop for Cab/Mics/Room sim and Reverb (Compressor and EQ not used), so extending the FM3 with outboard is another tactic. Using the OX Stomp for the Cab Sim saves about 5% CPU, with the Room Sim and Reverb saving much more.

 
Last edited:
Greg Ferguson said:
It’s still true. They use a very small amount, less than 1% individually; I think it came out to 0.20 or 0.25% each when I measured. I only use them when necessary to make a parallel path line up correctly, never for cosmetic reasons.

There’s a point where swapping long chains of shunts with the Feedback blocks makes sense because the Feedback blocks can traverse more grid slots without additional CPU consumption.
Click to expand...

Aleksi said:
Thanks, tested again with a large enough amount of shunts and they do use a little bit of cpu
Click to expand...
See above. Already checked and indeed they do use a bit of cpu. Inserting a shunt into all 48 slots and connecting the shunts on each row together used about 6-7%.
 
You must log in or register to reply here.
Back
Top Bottom