Why the AFxII has one dsp chip dedicated to amp(s)?

When running two Amps it uses "normal" instead of "highest quality" amp modeling. These names might be incorrect but you get the idea. This is because running 1 high quality amp uses more than 50% of DSP power on this chip. The amp modeling is extremely deep and high quality, and this is the tradeoff for that.


still sounds amazing with two amps :lol
 
  • Like
Reactions: sam
Thank you guys for your thoughtful (and sometimes funny) answers, I don't have idea of how coding and dsps works, that's why I was asking. Obviously the axe remains the best modeler on the market despite not being flexible as other cheaper ones, and anyway I rarely hit the cpu limit so it's not really an issue for me.
Is it possible to even know how they are using those processors? We know FAS uses one dedicated only to AMP because Cliff told us so. Has Atomic and L6 publicly said how theirs works
No they haven't. From what I understand they use parallel processing and the modeler behave like it has a single cpu, everything you add to the chain uses a certain percentage of the processing power, so for example on the helix you can choose to have 4 amps or a bunch of FX blocks or both.
 
Do you really care why ? Do we really need to break down every aspect of competing products and debate what's best ?

The one you spend your hard earned money and on , plug into , and enjoy the most is the best.

Do you play your axe fx an wish the DSP was handled different ?
 
I was looking at how other dual dsp modelers work (amplifire and helix in particular) and I asked myself this question. I think it's kind of a waste to have one of the two cpus dedicated to just this task in the case you want to use the axe as an effects only (or amp only) unit, you would have about twice the power if it had a more flexible architecture. Is there a technical reason for this choice? And in case, wouldn't be possible to have a global setting to switch between say "FX + amp" and "FX only" mode?
Sorry if this has already been answered or if it's a stupid question

It was said that the Helix uses one of its cpus for the user interface and the other cpu does all the amp and effects modeling. So if it can run 4 amps and many instances of effects at once (according to line6) just exactly how detailed and realistic do you think its algorithms could be?
I'm glad Cliff chose the dedicate one processor to modeling instead of the user interface.
 
The Axe is truly flexible. Routing possibilities galore. And sounding always amazing. Maybe you need to play more and read less about unnecessary stuff. Who needs 4 amps at a time? I mean, I strive to get a couple dual-amp setups sounding good. 4 amps at a time sounds like useless overkill.


Sent from my iPhone using Tapatalk
 
The Axe is truly flexible. Routing possibilities galore. And sounding always amazing. Maybe you need to play more and read less about unnecessary stuff. Who needs 4 amps at a time? I mean, I strive to get a couple dual-amp setups sounding good. 4 amps at a time sounds like useless overkill.


Sent from my iPhone using Tapatalk

Kind of an inflammatory comment. He can decide how he wants to spend his time.

It appears the take away is that we all love the Axe FX, and trust Cliff (with good reason, IMO) that the processing / architecture is the way it is for very good reason. Other than that, the only honest answer is that none of us truly know.
 
I asked this question not 'cuz I have to determine which product is better (flaming this stupid fanboy war that still goes on) nor to teach anyone what's the best way to use a dsp, I already stated I don't have the required knowledge.
I asked just because I always like to understand and learn technical details about the stuff I use when possible.. and I thought this was the purpose of this forum, but apparently I was wrong given half of the replies. The topic here seems: the reason of why I asked the question and the way I spend my time. That's why I think I won't be writing anymore.
I think someone should spend less of his time in front of a pc or tablet screen and more in the real life, just to care a bit less of what other people do.
Anyway, thanks to those that kindly answered the question and offered their constructive opinion on the subject. Cheers!
 
It was said that the Helix uses one of its cpus for the user interface and the other cpu does all the amp and effects modeling. So if it can run 4 amps and many instances of effects at once (according to line6) just exactly how detailed and realistic do you think its algorithms could be?
I'm glad Cliff chose the dedicate one processor to modeling instead of the user interface.
I think you're wrong here, if I remeber correctly they said it uses one mcu for the UI, not one dsp.. anyway the point was not to compare the algos of the units, it's obvious that line6 ones are way less detailed given that it can run 4 amps with the same chipset of the AX8. And that's why I now own an axe fx and not a pod hd anymore.
The question was to understand why fractal has not implemented that apparently more flexible design and if there would be tradeoffs in doing so
 
I was looking at how other dual dsp modelers work (amplifire and helix in particular) and I asked myself this question. I think it's kind of a waste to have one of the two cpus dedicated to just this task in the case you want to use the axe as an effects only (or amp only) unit, you would have about twice the power if it had a more flexible architecture. Is there a technical reason for this choice? And in case, wouldn't be possible to have a global setting to switch between say "FX + amp" and "FX only" mode?
Sorry if this has already been answered or if it's a stupid question

One of the Axe-Fx's DSPs is more powerful than both DSPs combined in the Helix. Our tests show that the TigerSHARC DSP used in the Axe-Fx is over twice as fast as the DSP used in the Helix clock-for-clock . Now add that our DSP is clocked 50% higher the net power is about three times greater. So a single TigerSHARC is about 50% faster than both DSPs combined on the Helix. Yes, if we decided to use both DSPs for effects then you would be able to run more effects. 99% of our customers buy the unit for amp modeling so it doesn't make sense to invest the time and resources to making this possible.

Our algorithms are studio-quality and use more processing power than competing products. We've always been about quality over quantity. For example, our variable delay algorithm (chorus, flanger, etc.) uses poly-phase interpolation. EVERY competing product I have tested uses simple linear interpolation (or occasionally polynomial interpolation) which is far less computationally intensive but doesn't sound as good. This is demonstrable and measurable.

I highly doubt the Amplifire has two DSPs. My guess is that their claim of "blazing dual DSPs" is marketing for a DSP with two cores. My guess would be a Freescale Symphony (DSP5672x) DSP which is a dual-core DSP running at 200 or 250 MHz. While a dual-core DSP will be faster than the same DSP with a single core it will not have twice the power as the cores will share buses and contention for those buses will hamper peak throughput. The unit's capabilities are far below the Helix and Axe-Fx so I doubt their using a DSP with anywhere near the processing power. This is complete speculation though.
 
One of the Axe-Fx's DSPs is more powerful than both DSPs combined in the Helix. Our tests show that the TigerSHARC DSP used in the Axe-Fx is over twice as fast as the DSP used in the Helix clock-for-clock . Now add that our DSP is clocked 50% higher the net power is about three times greater. So a single TigerSHARC is about 50% faster than both DSPs combined on the Helix. Yes, if we decided to use both DSPs for effects then you would be able to run more effects. 99% of our customers buy the unit for amp modeling so it doesn't make sense to invest the time and resources to making this possible.

Our algorithms are studio-quality and use more processing power than competing products. We've always been about quality over quantity. For example, our variable delay algorithm (chorus, flanger, etc.) uses poly-phase interpolation. EVERY competing product I have tested uses simple linear interpolation (or occasionally polynomial interpolation) which is far less computationally intensive but doesn't sound as good. This is demonstrable and measurable.

I highly doubt the Helix has two DSPs. My guess is that their claim of "blazing dual DSPs" is marketing for a DSP with two cores. My guess would be a Freescale Symphony (DSP5672x) DSP which is a dual-core DSP running at 200 or 250 MHz. While a dual-core DSP will be faster than the same DSP with a single core it will not have twice the power as the cores will share buses and contention for those buses will hamper peak throughput. The unit's capabilities are far below the Helix and Axe-Fx so I doubt their using a DSP with anywhere near the processing power. This is complete speculation though.

Yes but it's shinier and has more pretty colors.
 
One of the Axe-Fx's DSPs is more powerful than both DSPs combined in the Helix. Our tests show that the TigerSHARC DSP used in the Axe-Fx is over twice as fast as the DSP used in the Helix clock-for-clock . Now add that our DSP is clocked 50% higher the net power is about three times greater. So a single TigerSHARC is about 50% faster than both DSPs combined on the Helix. Yes, if we decided to use both DSPs for effects then you would be able to run more effects. 99% of our customers buy the unit for amp modeling so it doesn't make sense to invest the time and resources to making this possible.

Our algorithms are studio-quality and use more processing power than competing products. We've always been about quality over quantity. For example, our variable delay algorithm (chorus, flanger, etc.) uses poly-phase interpolation. EVERY competing product I have tested uses simple linear interpolation (or occasionally polynomial interpolation) which is far less computationally intensive but doesn't sound as good. This is demonstrable and measurable.

I highly doubt the Helix has two DSPs. My guess is that their claim of "blazing dual DSPs" is marketing for a DSP with two cores. My guess would be a Freescale Symphony (DSP5672x) DSP which is a dual-core DSP running at 200 or 250 MHz. While a dual-core DSP will be faster than the same DSP with a single core it will not have twice the power as the cores will share buses and contention for those buses will hamper peak throughput. The unit's capabilities are far below the Helix and Axe-Fx so I doubt their using a DSP with anywhere near the processing power. This is complete speculation though.

Do you think that now Yamaha are involved that they might be using existing Yamaha tech used in the higher end Motif workstations?
 
One of the Axe-Fx's DSPs is more powerful than both DSPs combined in the Helix. Our tests show that the TigerSHARC DSP used in the Axe-Fx is over twice as fast as the DSP used in the Helix clock-for-clock . Now add that our DSP is clocked 50% higher the net power is about three times greater. So a single TigerSHARC is about 50% faster than both DSPs combined on the Helix. Yes, if we decided to use both DSPs for effects then you would be able to run more effects. 99% of our customers buy the unit for amp modeling so it doesn't make sense to invest the time and resources to making this possible.

Our algorithms are studio-quality and use more processing power than competing products. We've always been about quality over quantity. For example, our variable delay algorithm (chorus, flanger, etc.) uses poly-phase interpolation. EVERY competing product I have tested uses simple linear interpolation (or occasionally polynomial interpolation) which is far less computationally intensive but doesn't sound as good. This is demonstrable and measurable.

I highly doubt the Helix has two DSPs. My guess is that their claim of "blazing dual DSPs" is marketing for a DSP with two cores. My guess would be a Freescale Symphony (DSP5672x) DSP which is a dual-core DSP running at 200 or 250 MHz. While a dual-core DSP will be faster than the same DSP with a single core it will not have twice the power as the cores will share buses and contention for those buses will hamper peak throughput. The unit's capabilities are far below the Helix and Axe-Fx so I doubt their using a DSP with anywhere near the processing power. This is complete speculation though.

This is a great explanations, exactly what I was looking for. Thank you Sir! Chapeau!
Just a note, I recall one of the line6 guys stating on tgp that helix uses two sharc dsps at 450 MHz, that would be the same as your upcoming AX-8.. but I may be wrong
 
Last edited:
Fuck me!!!!!! This why is just play the damn thing lol my brain can only process that. The technicality of the box is too much! We need more playing and less Han Solo keyboard physicist


Sent from my iPad using Tapatalk
 
One of the Axe-Fx's DSPs is more powerful than both DSPs combined in the Helix. Our tests show that the TigerSHARC DSP used in the Axe-Fx is over twice as fast as the DSP used in the Helix clock-for-clock . Now add that our DSP is clocked 50% higher the net power is about three times greater. So a single TigerSHARC is about 50% faster than both DSPs combined on the Helix. Yes, if we decided to use both DSPs for effects then you would be able to run more effects. 99% of our customers buy the unit for amp modeling so it doesn't make sense to invest the time and resources to making this possible.

Our algorithms are studio-quality and use more processing power than competing products. We've always been about quality over quantity. For example, our variable delay algorithm (chorus, flanger, etc.) uses poly-phase interpolation. EVERY competing product I have tested uses simple linear interpolation (or occasionally polynomial interpolation) which is far less computationally intensive but doesn't sound as good. This is demonstrable and measurable.

I highly doubt the Helix has two DSPs. My guess is that their claim of "blazing dual DSPs" is marketing for a DSP with two cores. My guess would be a Freescale Symphony (DSP5672x) DSP which is a dual-core DSP running at 200 or 250 MHz. While a dual-core DSP will be faster than the same DSP with a single core it will not have twice the power as the cores will share buses and contention for those buses will hamper peak throughput. The unit's capabilities are far below the Helix and Axe-Fx so I doubt their using a DSP with anywhere near the processing power. This is complete speculation though.
That last paragraph looks a bit odd? ...doubt the Helix has two DSPs...the unit's capabilities are far below the Helix and...

The Helix can't be far below the Helix. The Helix's DSPs are well known.

Me thinks you were talking about the Amplifire?
 
Last edited:
Yes, if we decided to use both DSPs for effects then you would be able to run more effects. 99% of our customers buy the unit for amp modeling so it doesn't make sense to invest the time and resources to making this possible.

Yes almost all customers buy the unit for amp modeling because it's great. I would not like to have more effects blocks in my presets but more amp blocks instead. Could it be possible to have :
1 amp block in a preset --> high quality simulation
2 amp blocks in a preset --> medium quality simulation
4 amp blocks in a preset --> low quality simulation (new feature)

Considering what Cliff explained about DSPs in the Axe vs DSPs in the Helix, even with 4 amp blocks at low quality in the Axe, I am sure the sound would always be better than what we can have with the four simultaneous amp sims of the Helix.
 
That last paragraph looks a bit odd? ...doubt the Helix has two DSPs...the unit's capabilities are far below the Helix and...

The Helix can't be far below the Helix. The Helix's DSPs are well known.

Me thinks you were talking about the Amplifire?

Yes, meant amplifier.
 
One of the Axe-Fx's DSPs is more powerful than both DSPs combined in the Helix. Our tests show that the TigerSHARC DSP used in the Axe-Fx is over twice as fast as the DSP used in the Helix clock-for-clock . Now add that our DSP is clocked 50% higher the net power is about three times greater. So a single TigerSHARC is about 50% faster than both DSPs combined on the Helix.

SHOTS. FUCKING. FIRED.
 
Back
Top Bottom