New hardware?

[MBTC]

CUDA can do insanely complex stuff. Think PhysX calculations and fluid dynamics. Stuff that makes audio algorithms look like childs play, comparatively speaking.

http://www.youtube.com/watch?v=EeblWU0pV5E

[EvilDragon]

Really, physics and audio DSP are very different beasts to tackle. What looks visually impressive and "complex" there in that demo doesn't mean that CUDA can do equally complex audio DSP algorithms at the same level of facility...

[MBTC]

Perhaps, but the fact that there are audio plugins like reverb that exist proves it can be done. That demo is just one example. Physics calculations such as fluid dynamics ARE complex, period... we don't need eye candy to establish that.

My personal theory is that the real barrier is the learning curve of the CUDA SDK, or perhaps that NVidia has not given proper attention to documenting and/or accommodating certain features, but that the processing capability is fully there. There is a huge amount of unutilized potential.

[TweakHead]

now if I may make an uneducated guess, I think the point is we've got this cards with some processing on them already in our computers and we're debating here if any use could be found for that little extra that usually just takes a nap while we're making music and audio related stuff. TC PowerCore or Universal Audio have specialized dsp cards, we all know that. But since Apple, for example, sells Mac Pros with a lot of audio people in mind (not only, but also a lot) how come hasn't something like the two examples I mentioned come with such computers as some kind of open dsp card for any developer to make use of? why wouldn't this benefit the industry? it would instantly create a new market for better plug-ins while boosting the performance at the same time. This is a recurrent thought I have. Maybe it conflicts with the market interests, where everyone's trying to pull the profits on their own, I guess. But if the user experience was the priority, such a thing would make perfect sense. We have the super sound cards - even non gamers - in our computers, we could just as well have something dedicated to proper dsp usage, it's just my logical conclusion.

We're seeing more and more technology being implemented as CPUs begin to offer the capacity to work their way into copping with it. Like zero feedback latency filters, which are very demanding to process - I take that for granted, even after some experiments with Reaktor I've done myself. So what I'm saying is: more quality is possible, we're not getting the best we could unless we're paying the BIG buck for it and that renders us totally dependable on some brands and their support and dedication for updates. TC has died, plain and simple. Universal Audio doesn't extend their product's range to instruments, which is somewhat weird for my mind, but, there's this big whole on the market today which could be filled with something else. And the lack of a standard is something we've talked about here, this could all be history with this so called "no brand dsp dedicated cards". my 2 cents

[MBTC]

Well in a way I guess we're discussing the same thing -- but that DSP card already exists in the form of a graphics card that isn't used during most folk's music making.

It's not a "no-brand" solution in the sense that one vendor supplies the graphics cards, but as far as I know the software SDK for CUDA is completely free/open and available to anyone... I see lots of hobbyists working with it... example here: http://www.theover.org/Cuda

There's also a tremendously flexible range of power to choose from among GPUs. You can pay $50 for a graphics card or you can pay $1000 depending on how much processing power you want (i.e. think plugin instances). You can buy one of them today and add a second or third later and scale linearly. The guy above is working with low-end (by today's standard) GPUs and getting results.

Now EvilDragon's position is that there are technical reasons that some types of audio applications won't work on CUDA, and I'm not completely denying the possibility, I'm just looking for answers to what they are that I can digest, because the evidence I see seems contrary.

As far as someone coming out with a card that is not proprietary and is designed for open use.... well the problem there is the level of R&D required to produce the hardware and software to do something like that is insane... tens of... no probably hundreds of millions of dollars. If someone invests that kind of money, there needs to be some return in it for them, they cannot just invest it as an act of goodwill to give to the community.

So we end up back at the folks like TC Powercore... someone there bit the bullet and invested some money and took a gamble. Apparently there wasn't enough money in it for them to sustain.

So I think what we are more likely to see are less-ambitious, specialized devices utilizing DSP. Actually that's all the Ultranova is, a DSP with an audio interface and software plugin.... but for a few more bucks why not add keyboards and knobs and call it a synth. Much cheaper to support a single-purpose hardware synth than a complete computing platform like CUDA.

[TweakHead]

You're saying that this card would cost a whole lot of money and possibly produce no returns. That's possibly true, but only until it becomes a standard - made an integral part of a high end computer. It doesn't seem totally impossible to my mind. Some years ago it wasn't even standard for computers to come with a sound card at all, for example. As soon as they show up, more applications came to fill this increased multimedia potential in computers, not only that but the OS themselves have grown more mature with that move. So I think that, naturally, only big companies could pull this off, like Apple which I mentioned above. They can make the whole world use touch phones, why couldn't they do something like this? And sell the products for their all to special Logic X (more like XI) in their dearest App Store, making everyone make some bucks. I think this open card doesn't mean there can't be competition and profit, that's only relegated to software perhaps. But making such a thing become as ordinary as RAM memory or the graphics card would be great. It could also, I think, serve the gaming industry. We don't often talk about that, but rendering real audio in games in real time still has a long way to go to, there's room for improvement. And the games have proven useful for pushing the industry into greater innovation and for financing it even. Just a thought of course.

But no one would deny this would be useful for the pro user. I mean, the new mac pro they're about to sell looks like a turbine from a space ship, has all the looks and specs of a great machine, except... well, you really need to buy a decent sound card, an Universal Audio card to, maybe an Apogee Rosetta pcie (if you're running a bit shinny biz), so forth and so on. So, let me think it through here... Mac what? Pro? Reaaallyyy? Just an ordinary computer with nothing more then your laptop on it, except more of the same (literaly). If this is what the market economy has to offer us, then I say that innovation is being halted more then helped by it, that's all.

And if we're honest, that's happening every single day. Have you seen an Imac on the inside? They could easily made them more resistant and better constructed but what's the point in making a great product that would last forever? No point! Except, of course, the user. This is were we're at at this point in evolution, but it makes me sick. I like to think that more resources for computing within a computer so as to enable creativity is worth it, simply that. What the economy needs to pull it off is really not the point. I mean, it's done a lousy job at making people keep their jobs in my country, it's out sourcing our most successful stuff to the third world bypassing all sorts of laws that give people their rights, including care for the environment, children in factories, so forth and so on... Screw the economy, we're talking evolution here. We just need to stop pretending that Mac Pros really live up to their name, because they don't.

[MBTC]

But Apple and companies like them are not interested in creating open-standards, they are interested in highly proprietary devices and defending their design patents like pit bulls.

Its hard for me to understand your position from your message, it seems like it starts off that you'd rather Apple be the holder of a standard like CUDA but then later in the message it seems you're dissatisfied with Apple?

Anyway one of the things that helped Apple get back on their feet, aside from a major influx of money from Microsoft, was when Jobs was put back in charge and he trimmed their product portfolio down to focus on only a few things. In other words I doubt we're going to see them expand into music hardware.

[TweakHead]

You're right. I was saying that only those companies would be capable of achieving such a treat, but they're focus seems to be else where these days. I'm not totally unsatisfied with Apple. I really like the software and the overall reliability of it. However, that doesn't stop me from judging their current priorities - which is the portable devices, namely phones and tablets.

I just read on the news that Microsoft will release another windows 8 version later this year for the same reasons: most of their costumer base weren't happy with the more oriented towards touch devices interface. They're actually being punished for doing what Apple's been doing, successfully, so to speak.

I think this companies, like you say, are interested in creating "highly proprietary devices" and that doesn't help the user community in many cases. I was also stressing the point that there's a big marketing hype surrounding some products like "high end computer workstations". You need those, plus all the other equipment to release the computer from the "inside the box" form - like another jail - so as to be able to use it creatively.

For graphics you need a graphic pen, for audio you need plenty of stuff. For such a big buck, you'd expect a system with a proper audio card at least. And this card I keep talking about is just another way of saying: I feel there's room for more innovation that would hopefully become a standard. Standards are good for one thing: they mean the vast majority of people gets to use these. The music industry in general fears this massive expansion of people using more serious tools to give life to their own creations and being able to present those online with the same level of presentation and to the same standards that only a few studios could a just a few decades ago. I think of that as evolution, it's good that more people are able to be creative. I don't even care that some people feel they can't make revenues like the rock starts used to. So less ego, and more community. While we see that where you have a community working you get good results: think of how MAX has expanded the usability of Live to the point they've decided to integrate it fully to their software. They know, just as we do, that some hobbyists can actually bring more value to their product. Same goes for Reaktor.

I remember you saying these are not "native" languages. But implementing new hardware with an open language that could be used by any developer out there - and the hobbyists - for bringing more demanding audio (maybe not just audio...) applications to life would be great. If this was shared across all the digital audio workstations out there, a new standard like midi, perhaps, they could all through a big part of their plug-ins processing power to that board and thus produce a big performance boost to all audio workstations. What's wrong with that?

Read on the latest issue of Music Tech's mag - dedicated to Logic Pro - that they'll update their Audio Unit format in ways VST3 already has. This is a double effort for the same thing taking place, another of the sub products of this ego/brand centered economy we live in that also doesn't help the user much. Moving on, another point mentioned is they're aiming for better thread distribution among the cores. That's good of course!

You developers go ahead and tell me. I think our CPUs aren't being used to their full potential just as well, right? It's all good and great if you listen to the marketing hype: 4x more performance in everything gets you thrilled quite fast. But how do these new features in CPU technology being translated into actual performance for the user? Many times it takes some time. One of the reasons I don't jump on the new OS as soon as it shows up: they have this nasty tendency to dive to new features while leaving others behind them, just after they got to the point where they're working properly for the first time ever. I'm not basing this on any detailed draft of information, just on my own subjective experience with computers.

[MBTC]

I personally think CPU technology has hit a ceiling (actually its been hovering around the ceiling for many years now). They are adding more cores, which is good for certain types of applications, but as has been discussed here before some computing tasks are serial in nature and do not lend themselves completely to parallelism. Thermal limitations, among other things, prevent CPUs from getting much faster.

GPUs on the other hand still seem to be enjoying big performance gains, generation over generation, while finding ways to do so with lower temperatures and less power consumption. Exciting times in GPU-land, not so much on the CPU side of things.

All of that aside for a moment, and taking into consideration what you've brought up about Apple's diminished focus on desktops: Lots of the buyers of Apple computers for music making purposes go with laptops or iMacs, partially for value but more likely for mobility. DJs and musicians are on the go more than ever. That presents a problem for makers of, for example a DSP card like the TC Powercore since there's no where to put the card. What's more, high-end GPUs need a desktop as well for thermal reasons, better airflow is required. You mentioned the possibility of an external box, but then you we the benefit of a bus directly on the motherboard, and its back to the streaming over USB/Firewire etc. Of course, this can be done, but we've seen it can be flaky and come with drawbacks of its own like latency.

Did you ever look at the Openlabs stuff? They still sell the Meko, although I can't say I know of anyone that uses one. It's kind of like what you've described, a separate box (just happens to have a keyboard) with its own DSP... well actually better than a DSP -- rather a full blown PC running Windows and lots of soft synths http://openlabs.com/LxdPage

[TweakHead]

Yep, I've seen them listed on music stores. These are dedicated audio workstations, right? I think even Roland/Cakewalk sells some complete solutions as well. But I think it would be a lot better for high end computers - the mac pro is just an example of this, of course - should incorporate some of this high end features within their main configurations. A better, more reliable sound card is just the obvious improvement I'd like to see implemented. I mean, there's great internal sound cards that can be expanded with external interfaces for extra connectivity from vendors like RME or Apogee for example. This two offer really high end low latency, very impressive word clocks for syncing hardware devices and top notch audio quality.

To some extent, what I've been ranting about here is that the standards should be placed higher. It's easy to convince people to buy a new computer based on the looks alone (like the majority of Apple's computers), but if we're serious about making computers better suited for demanding applications, they could just as well come packed with higher quality components overall. My idea of introducing a new component at this "basic configuration level" is another way of saying that we should be getting more for granted based on the price tag of such computers and even that the industry should focus on setting new standards - because ultimately it would help the programmers (not having to translate the same plug-in to different formats) and the users because we'd be getting better software with access to more resources and ultimately a performance and quality boost.

It's easy enough to say this, but if we take into account how the market works it seems a daunting task. I'm going to stress again that it isn't positive when the interests of particular brands surpass the interests of the user or even go against him. Making profit is not only an objective here, but also a surviving necessity - that's granted. But to which extent are we willing to go before we start making more compromises that would allow better solutions to be achieved? It's like politics, diplomacy is much needed here! No one can win alone, and if no one steps back just a little bit, we'll be seeing this plenty of standards, protecting trade secrets versus open attitude - even for users and hobbyists - stretch beyond reason.

I agree with you about CPU, even though the performance has been getting higher non the less.

Cheers