Future for pianoteq - less "simplification" ?

Hi,

I think your post is very appropriate - and this question was phrased differently in previous posts e.g.: "could we have a more complete off-line model that renders a better piano sound into disk - thus not being limited by CPU strength and real-time limitations?"
And if my memory serves me right - Phillipe answered one of these questions: "The problem is not just the computing power, but also what to do with it - i.e. what to model?"
So - do Modartt have a "Holy-Grail" model in their drawer waiting for the adequate hardware to arrive - thus cutting us back from having the perfect virtual piano? I seriously doubt that. That is - unless we're talking about a generic physical "kinetic-material-simulator" that takes into account wood, felt, metal and air molecules and computes movement and interaction between gazillions of particles over in nanosecond resolution... Nah ... I don't think such a model will be practical on any machine in the foreseeable future ... but I'd be glad to be wrong here.
So we need a simpler model.
Modartt keeps researching and improving their models, and I'm sure they have unreleased improvements in development, but I don't think they are cutting back on us...

Regards,
Eran

The simplification it's complex, and I'm sure the degree of simplification take CPU power in consideration.  But once the simpolification is made and the algorithm is ready, a faster CPU do not change the sound naturality.

I think the software architecture and the simplification, how it works, do not allow to creat a slider for the simplification degree.

Modartt probably could create a new model just for "turbo PCs", but that would be a small market.

I'm afraid it's not that simple.

The accoustic of a piano, in modelled terms, probably have a lot of variances that we could not understand well to choose in logic strategy.
But maybe they could present about 8 scales of complexity to choose.
For "mp3 export", for example, a full complexity available could be use without fear.

For professional studios, why not a model like that? They will render a  file with full complexity for the final edition.

Hey, some people say that complex games are a industry to make PC uses poor.   Há háaa.
There is always need for a new videoboard, more memory, a nem faster processor etc.  A snowball...

If Modartt create a scalable piano model, some hardware manufactures (suposed "conspiracy mafia") will offer patronage support. Háa háaaaa.

While playing ??????  What for?

I admit to ASSUMING that Philippe has at least a 3 year vision, probably a 5 and perhaps 10 year vision, although uncertainty of all things increases the farther out we try to look.

As I said earlier, I believe that the current implementation of the model is constrained by available CPU performance and the "simplifications" allow a complete and very desirable product.  One that WORKS in currently available processors by removing some of the processor intensive calculations.

It is only my interpretation from what I have read and viewed, but I believe the DESIGN of the model that will take advantage of the 5GHz CPU is already "there".   
{"there" being undefined (-:}
I suspect that there will not be a need for a whole new math model for PTQ 5 - but of course my crystal ball is as cloudy as anyone else's when looking out that far.

Which then begs the question.  Like computer animation, video and image processing, do the Pianoteq computations lend themselves well to using multiple cores?

(this maybe answered in a FAQ already, but my lunch is over and I have to get back to work)  ;-)

My guess is "Yes".

That is based partly on my perception that Philippe et al are anticipating the technology curve(s) and have some design ahead of it and waiting.
Also based on my guesses as to what the model actually IS and how it has been implemented... don't put much faith in this part though (-:

Hmmm, I s'pose I could monitor my dual core CPU usage to see if PTQ hammers both cores approx the same.
In a LOT of apps it hardly gets around to using the second one, but that might be more about windoze and 32 bit legacy issues than anything else.

Yes, I worded that bit about 5GHz machines poorly, but I no longer think in MIPs and was in a hurry.
I probably had "about twice today's 2.x GHz dual core machines" in my mind.

The sound tone quality of actual models do not change with higher CPU.

You just get more polyphony, can use and feel safer that will not get cracks with high pollyphony moments.

This is what COULD change in v4, that is, the scalable model.

Exactly my point;
Which "simplifications" will be coded for execution in real time as processors become faster (and more numerous) ?

Sound boards aren't "Simple", look up Vigdorchik tap tones some time to see what violin makers worry about in the physical realm (-:
I think it can be modeled very well, the model just can't (yet) be run fast enough for anything resembling a live performance.

Thank you, Julien!  That's what I was thinking you would say.  Core clock speed still trumps a plethora (gaggle?  flock?  herd?) of cores.

Does a difference exist between coding for "virtual" cores (i.e., "hyperthreading") and real cores?

Managing the allocation of processors in a parallel computer have always been a nightmare for hardware designers and software developers ! Despite thousands of studies, papers, theses ad developments... since the '70s, this problem has never been completely solved - I mean on a general basis!

One thing is sure, however : a polyphonic sound generator is naturally appropriate to parallelism. But software resources in present OS of personal computers might not be smart enough to manage it efficently in a real-time context.

Alain