Overloud Sculptube

[AC222]

Yes, the low CPU with Overloud plugins is completely puzzling, even more because they don't sound like shit. Have someone finally done some work in assembler? I doubt it... There's not much talk about these Overloud GEM plugins around. I wonder why? I'd like to hear some opinions from people who really used them.

I was told that Overloud utilizes assembly code and if this is true, it would yield superior results both from a sonic perspective and from the vantage point of dsp efficiency (assembly code will be fixed point processing). You will probably find some a-hole that will come on here and dispute this fact but fixed point processing always sound better than floating point, all things equal that is as they say in economics.

Assembly code is harder to program but those that have used this process can testify that it yields a "smoother sound", more dimensional sound (ok people it's not a vague description, it can be likened to the difference of a plugin wrapping its sound around the source instead of sounding like it is plastered on top), and more open high frequency response. The other huge benefit is less cpu. The tradeoff is that is extremely complicated to program, hence why it is rarely used.

There is an antiquated DAW that I often still use called SawStudio that is all assembly code with fixed point processing. You can download for free and try it yourself but if you render at the same bit rate and sample rate as say protools, reaper, cubase the SS renders sound significantly better. It is not the differences in dither algorithms and pan law ok, it is the difference in rounding errors. Math is math they say but those same people never point out that fixed point doesn't have rounding errors like floating point.

The improvements from fixed point from a sonic standpoint can be likened to OLED tvs versus LCDs. OLED technology allows for infinite blacks while the best LEDs can allow for 99.999% black scales. Yet there is still a very visible and noticeable improvement from the OLED blacks in dark scenes over the very best LED tvs, which cannot reach infinite blacks. Thank for taking the time to be educated all.

Source: See Post on GS forum

poshook 6th December 2017 04:58 AM
Quote:
Originally Posted by kj.metissage (Post 12999438)
Don't get me wrong, but when I read things like "more than 1500 instances" when we talk about a plugin modeling tubes, I'm worried.

I'll have to try it of course. Looking forward to reading your opinions guys, as I'm away from my DAW for 1 more week. Damn that's long !
From what I remember, Overloud does not use classic hi level code like C++ with faster workflow but limited optimisation capabilities. They use assembler which means low level code where you have to write everything manually from the ground however application written in assembler runs much faster

[AC222]

Also, for full disclosure, I own and love Sculptube myself. I also find that their overdrive boxes to (Imho) sound the most accurate when it comes to emulating things like the Ibanez Tube Screamer and Boss Super Overdrive as not only does the tone match most closely versus the competition, it lacks the artifacts that come with other digital plugins (too fizzy or fizzy, added boxy frequencies in the 450 - 900 hz range) etc.

[lacandon]

It sounds very close to the hw and the nebula one the uad not really or at least from.what I heard in a shooutout the nebula.was nearly identical to the hw uad sounded completely different more agressive and bigger sounding but duller.

[EvilDragon]

(assembly code will be fixed point processing)

Not necessarily. Assembly code can also deal with floating points, ESPECIALLY if SSE and AVX instructions are used (which would very much be the case with today's CPUs, as these instructions do more in less time, with good reason). Also today's CPUs don't really gain much in speed when doing fixed vs floating point calculations. That was maybe the case over a decade ago. Get on with the times

There is an antiquated DAW that I often still use called SawStudio that is all assembly code with fixed point processing. You can download for free and try it yourself but if you render at the same bit rate and sample rate as say protools, reaper, cubase the SS renders sound significantly better.

Newsflash: Reaper also has fixed point processing modes, several flavors of it. Since quite some time now. Also PT HD used 48-bit fixed point too, IIRC.


BTW, fixed point is not the end all be all. It has some cons, as well.

From what I remember, Overloud does not use classic hi level code like C++ with faster workflow but limited optimisation capabilities. They use assembler which means low level code where you have to write everything manually from the ground however application written in assembler runs much faster

There are plenty of ways to optimize code when writing C++. Using inline ASM is one of them. Various compiler optimizations is another.

Also, nobody HAS to write EVERYTHING manually from the GROUND UP in assembler. Nobody does graphical interfaces in assembler because that just doesn't make any sense. The DSP engine, sure, there are a few examples, but the GUI is 100% of the time done in a much more advanced and modern way (WYSIWYG drag&drop editing of the GUI etc. etc.).

[imrae]

Assembly code is harder to program but those that have used this process can testify that it yields a "smoother sound", more dimensional sound (ok people it's not a vague description, it can be likened to the difference of a plugin wrapping its sound around the source instead of sounding like it is plastered on top), and more open high frequency response.

This is utter nonsense.

[AC222]

Assembly code is harder to program but those that have used this process can testify that it yields a "smoother sound", more dimensional sound (ok people it's not a vague description, it can be likened to the difference of a plugin wrapping its sound around the source instead of sounding like it is plastered on top), and more open high frequency response.

This is utter nonsense.

It sounds like someone doesn’t like to hear the truth . No nonsense. Only facts

[resynthesis]

So, if I use /FA on my VC++ compiler to output the assembly language and then assemble it, everything will sound smoother? LMAO, codswallop!

PS I've been writing in various assembly languages since the 80s.

[AC222]

So, if I use /FA on my VC++ compiler to output the assembly language and then assemble it, everything will sound smoother? LMAO, codswallop!

PS I've been writing in various assembly languages since the 80s.

Yes that is correct. Speaking of languages, codswallop? What gives, can’t think of something more creative?

[imrae]

My day job involves computational physics. If code compiled from C/Fortran gave different results from an assembly implementation of the same algorithm and datatypes, especially in a manner related to the frequency content of the data (!), we'd be demanding refunds from Intel for a defective compiler.

[EvilDragon]

So, if I use /FA on my VC++ compiler to output the assembly language and then assemble it, everything will sound smoother?

Yes that is correct.

EEEEERP, WRONG!

It sounds like someone doesn’t like to hear the truth . No nonsense. Only facts

You're arguing against actual programmers here. Your "facts" are not really facts. Perhaps you should stay out of topics you know jack squat about.

[jens]

You will probably find some a-hole that will come on here and dispute this fact but fixed point processing always sound better than floating point, all things equal that is as they say in economics.

So whoever dares to question the wild claims you make is an arsehole?


Way to go!

[AC222]

You will probably find some a-hole that will come on here and dispute this fact but fixed point processing always sound better than floating point, all things equal that is as they say in economics.

So whoever dares to question the wild claims you make is an arsehole?


Way to go!

Not wild claims. Facts. have a good day

[EvilDragon]

Those are not facts but rather subjective statements.

[VariKusBrainZ]

Hew would regute t even if the Dev(il) himself showed up and said he was talking crap

[whyterabbyt]

I was told that Overloud utilizes assembly code and if this is true, it would yield superior results both from a sonic perspective

Absolutely. But in fact, assembly code is pretty poor, really, what you ideally want to do is skip the assembler and directly enter the code into memory, bit at a time, word by word. Assembly code aliases too much.

(assembly code will be fixed point processing).

Well, except for the floating point assembly code, of course.

https://docs.oracle.com/cd/E19120-01/op ... index.html

But nobody uses that.

Assembly code is harder to program but those that have used this process can testify that it yields a "smoother sound", more dimensional sound

Dont be silly, that actually only works when the developer ensures that their keyboard cable is oxygen-free copper, and they have the right quartz crystals on their mouse. When we had proper computers, like Crays, we also used to be able to ensure we could align our operands to the nearest significant ley-line, but you cant do that anymore, which is why all mobile devices have such shit audio.