FM3 Firmware Version 5.03 beta 6 (release candidate)

Status
Not open for further replies.
All this CPU talk, I don't know if some of you guys have had the chance of trying an FM3 with an FM9 side by side, I did last night and man, what a HUGE difference. And let me tell you this, I have a particular preset that I always use so my ear is very very accustom to it, loading this preset to the FM9, to me, it sounds more lively. I don't know what's going on there but it simply feel more natural.
 
nicolasrivera said:
All this CPU talk, I don't know if some of you guys have had the chance of trying an FM3 with an FM9 side by side, I did last night and man, what a HUGE difference. And let me tell you this, I have a particular preset that I always use so my ear is very very accustom to it, loading this preset to the FM9, to me, it sounds more lively. I don't know what's going on there but it simply feel more natural.
Click to expand...
Interesting, with the exact same settings they "should" sound and feel identical.
 
Vio said:
Thanks @hippietim, my point exactly.
Processors/cores peak briefly towards the max 100% or even stay there for a couple a seconds but they usually stay in the 5%-20%.
Badly designed software (tight loops, no multi-process/multi-threading when useful) can hog a processor not at 80% but at 100%.
The FM3 hardware, running the signal through the blocks can reach the 80% and still be stable (probably way less when no signal is present but the CPU usage is there to warn you)
From my point of view this is a logical representation of the machine resources and I can use that to know how far I can push it.
Click to expand...
I don't necessarily agree with your assessment of what is badly designed software.

For starters, tight loops are the best kind!

I do agree that using multiple threads can be a really good way to go. However, multi-threading is far more difficult to get right than what even experienced developers often understand. Synchronization is what most people worry about. That's important to get right but the patterns are well understood and there's plenty of sound approaches to follow depending on the problem at hand. The thing that often gets overlooked is the data access patterns - in particular, what happens to memory caches. The OS doesn't know what your data usage will be so your work could easily get bounced around cores and CPUs. Crossing core boundaries can be bad enough but if you have to cross NUMA boundaries you can thrash the cache lines horribly. Understanding this and managing processor and NUMA affinity is not trivial.

As far as hogging the processor, sometimes that's what you're striving for! A typical interactive GUI app should not be pegging the CPU too often. But if you're running simulations, large data models, rendering graphics or video, etc. you absolutely want to hammer all of the CPUs as hard as you can. When I was working on CAD tools for chip design, I promise you that we worked hard to hit 100% CPU usage on as many cores as we could. I'm working on optimizing some modeling for Quantum computer design by augmenting Python code with C++ code in hot spots - again, I'm doing what I can to keep sucking up the CPU and trying not to blow caches. Fortunately, threading is a non-issue since the problem partitions unbelievably good at a process level so we can spawn as many processes as there are available cores and they can run fully independent of each other.
 
shanesiegle said:
Interesting, with the exact same settings they "should" sound and feel identical.
Click to expand...
I exported the preset from the FM3 and imported it in the FM9, identical settings. The preset just feels better in the FM9, I went back and forth between the 2 units just to make sure my ears where no playing games, maybe is a well played game my head has, but I feel the difference.
 
nicolasrivera said:
I exported the preset from the FM3 and imported it in the FM9, identical settings. The preset just feels better in the FM9, I went back and forth between the 2 units just to make sure my ears where no playing games, maybe is a well played game my head has, but I feel the difference.
Click to expand...
Since this is audio we are talking about...could it be possible that having the extra CPU with the FM9 allows it to more accurately replicate that preset? Maybe it is at the limits of the FM3 and that is not a problem with the FM9?
 
hippietim said:
I don't necessarily agree with your assessment of what is badly designed software.

For starters, tight loops are the best kind!

I do agree that using multiple threads can be a really good way to go. However, multi-threading is far more difficult to get right than what even experienced developers often understand. Synchronization is what most people worry about. That's important to get right but the patterns are well understood and there's plenty of sound approaches to follow depending on the problem at hand. The thing that often gets overlooked is the data access patterns - in particular, what happens to memory caches. The OS doesn't know what your data usage will be so your work could easily get bounced around cores and CPUs. Crossing core boundaries can be bad enough but if you have to cross NUMA boundaries you can thrash the cache lines horribly. Understanding this and managing processor and NUMA affinity is not trivial.

As far as hogging the processor, sometimes that's what you're striving for! A typical interactive GUI app should not be pegging the CPU too often. But if you're running simulations, large data models, rendering graphics or video, etc. you absolutely want to hammer all of the CPUs as hard as you can. When I was working on CAD tools for chip design, I promise you that we worked hard to hit 100% CPU usage on as many cores as we could. I'm working on optimizing some modeling for Quantum computer design by augmenting Python code with C++ code in hot spots - again, I'm doing what I can to keep sucking up the CPU and trying not to blow caches. Fortunately, threading is a non-issue since the problem partitions unbelievably good at a process level so we can spawn as many processes as there are available cores and they can run fully independent of each other.
Click to expand...
Agree to disagree :D
 
nicolasrivera said:
I exported the preset from the FM3 and imported it in the FM9, identical settings. The preset just feels better in the FM9, I went back and forth between the 2 units just to make sure my ears where no playing games, maybe is a well played game my head has, but I feel the difference.
Click to expand...
Interesting. Not what I would expect.
 
nicolasrivera said:
I exported the preset from the FM3 and imported it in the FM9, identical settings. The preset just feels better in the FM9, I went back and forth between the 2 units just to make sure my ears where no playing games, maybe is a well played game my head has, but I feel the difference.
Click to expand...
I'm not arguing your POV, but I do think it would bear further investigation. There are a few blind A/B videos out comparing the FM3 and the III with no apparent difference. I think Fractal's goal is to have a standard result reproduced across all three platforms.
 
nicolasrivera said:
I exported the preset from the FM3 and imported it in the FM9, identical settings. The preset just feels better in the FM9, I went back and forth between the 2 units just to make sure my ears where no playing games, maybe is a well played game my head has, but I feel the difference.
Click to expand...
Things to check:

1. Same volume
2. Same output
3. Same monitor system
4. same parameters/eq/resolutions… i.E maybe the reverbs and cabs are set to higher resolutions in the FM9
5: same firmware

Still… you’d have to be able to switch back and forth instantaneously to really test given all other parameters being equal.

If you still hear/feel a difference perhaps there is something to it.

Sean Meredith-Jones
 
nicolasrivera said:
I exported the preset from the FM3 and imported it in the FM9, identical settings. The preset just feels better in the FM9, I went back and forth between the 2 units just to make sure my ears where no playing games, maybe is a well played game my head has, but I feel the difference.
Click to expand...
The only difference between the two running the same preset that I can think of is the variable input impedance on the FM9
 
nicolasrivera said:
All this CPU talk, I don't know if some of you guys have had the chance of trying an FM3 with an FM9 side by side, I did last night and man, what a HUGE difference. And let me tell you this, I have a particular preset that I always use so my ear is very very accustom to it, loading this preset to the FM9, to me, it sounds more lively. I don't know what's going on there but it simply feel more natural.
Click to expand...
The two aren't at parity in their firmware yet. Once they are they'll be identical, that's just how computers work.

Also, be careful with familiarity. Being used to something is going to color your idea of what sounds good. Step away from the FM9 for a week then compare the two and your opinion might change.

Think of it this way: The two modelers share the same DAC and ADC, and when the firmware uses the same algorithms they will generate the same output to their ADC.

framos said:
The devices are supposed to sound the same, however the firmware/algorithms may be at different iterations?
Click to expand...
Exactly.
 
Robbie Boyette said:
Since this is audio we are talking about...could it be possible that having the extra CPU with the FM9 allows it to more accurately replicate that preset? Maybe it is at the limits of the FM3 and that is not a problem with the FM9?
Click to expand...
No. Fractal insists on a consistent level of quality of the sound and will remove features if they're keeping that from happening, which is why the FM3 doesn't have some of the esoteric parameters in the amp and delay blocks.

From the Wiki:

  • FM3: aside from modifications designed to optimize CPU on the FM3 (see below), the FM3 runs the same algorithms and therefore sounds the same when identical settings are used. Exceptions are possible, when later improvements and new features of the III can't be ported to the FM3 because of CPU constraints.
Click to expand...
"The core modeling technologies are the same, but not all features are present on the lighter platform. For example, the FM3 doesn't have input dynamics or bias trem." [3]
Click to expand...
(comparing the FM3's amp modeling to the Axe-Fx III) "They are the same quality. Certain features were removed to allow the algorithms to run including the bias tremolo, input dynamics processing (...)." [5]
Click to expand...
(FM3) "We removed all the superfluous stuff (bias tremolo, dynamic presence/depth, etc.) in order to get the core amp modeling to run on the slower processor." [6]
Click to expand...
 
Greg Ferguson said:
The easy way to confirm that the preset is the same is to create it on the FM3 and then push it to the FM9, and make sure all their global settings are the same.
Click to expand...
You can also try to after importing to FM9, export it and import it back to the FM3 in a slot next to the original then flip back and forth between the two :D
 
fcs101 said:
I'm not arguing your POV, but I do think it would bear further investigation. There are a few blind A/B videos out comparing the FM3 and the III with no apparent difference. I think Fractal's goal is to have a standard result reproduced across all three platforms.
Click to expand...
Absolutely agree with you. I want my FM3 to sound as good as the Are FX III, let alone the FM9.

Another great thing about the FM9 is seeing how you can put 2 amps, 2 cabs, 2 drives and 2 reverbs with high settings and the CPU barely ranges to the 70%. I'm so accustom to always watch my resources, its a freeing experience for sure.
 
Last edited:
smj said:
Things to check:

1. Same volume
2. Same output
3. Same monitor system
4. same parameters/eq/resolutions… i.E maybe the reverbs and cabs are set to higher resolutions in the FM9
5: same firmware

Still… you’d have to be able to switch back and forth instantaneously to really test given all other parameters being equal.

If you still hear/feel a difference perhaps there is something to it.

Sean Meredith-Jones
Click to expand...
Good points, will double check tonight, thanks.
 
Status
Not open for further replies.
Back
Top Bottom