On 2006-05-23 07:03, organix wrote:
Native AU/VST devices will become better in latency, better in sound and really better in functionality. Maybe in a few years there will be no reason to use SFP, because native plugins will better with the same latency.
I'll believe it when I see it =] A few reason why it'll never happen:
- VST doesn't have any infrastructure to support oversampling, so you end up with plugins that eat up a whole 2GHz CPU for 8 voices, which can be accomplished on a 200MHz CPU. need more power? add another 2GHz CPU. huge huge waste.
- Dual-core won't help that much, since even though there are 2 cores on the CPU, there is still just *a single interface*, i.e. both cores share the same throughput as a single core CPU. So you get better performance, but nowhere near 2x the performance. As people add more cores to the CPUs, this throughput problem will only get bigger.
- Lower latencies??? For that to happen, the motherboard architectures and operating system software would have to be *simplified*, and that REALLY isn't what is happening. With a DSP architecture/hardware, you have the DSPs pretty much tagged straight onto the DACs, so software plugins will ALWAYS have higher latencies. a DSP can easily handle the throughput necessary for audio processing with almost no buffering, while with a computer/software based solution, the data has to travel thru CPU/memory interfaces, interconnect buses, PCI/whatever buses, with a fair bit of buffering in between all those parts to handle the multitasking, only to reach the DACs. Yeah yeah, lower latencies, I'll believe it when I see it =D
- VST is complete sh*t (sorry) for hardware integration. UAD/PoCo pull it off because they are pure DSP accelerators and don't deal with live/real-time signals at ALL. They are unusable live. This doesn't matter to the automation-only crowd, but seeing how there are still actual proper live musicians around, this is an issue. If you take a look at all the systems that tried to mix live/realtime signals, they all had huge problems. Waldrof tried it with the AFB16, and it killed them. Creamware tried it with XTC mode, and they completely stopped supporting it. Access tried it with the TI, with some success, but the Virus mailing lists are FULL of people who have huge problems and are pretty much saying what you are saying about Creamware, they have lost it, they s*ck, blah blah. Starting to see a pattern. I seriously doubt anyone else is going to try it at this point. It's possible to get it to work, of course, but the almost-impossible thing to do is to get it to work *reliably*, *all the time*, on *all platforms/computers*.
- With Vista, you will very likely need 2x the amount of processing to get exactly the same performances you get under XP right now, if the last 15 years of Microsoft software is used as reference.
- If you think 64-bit processing is all the rage and is going to double your performance, think again, and look at all the released benchmarks up to this point for confirmations. POVRay64 running under Linux64, all optimised for 64-bit AMD processors, barely get a 20% speedup. Maybe in 5-10 years you can expect compilers to be properly optimized for 64-bit.
- The WHOLE rest of the computer industry is moving in EXACTLY the opposite way of trying to do everything in software. Most of the graphics/3D used to be handled in software (on PCs I mean, Amiga and other platforms had the multiple-DSP-like-architecture down 20 years ago (sadly they had other issues)), then they added 3D accelerators (== DSP), then the 3D accelerators were integrated to the graphics card, then they let you put several (still DSP-accelerated) graphics card together with SLI and PCI-E, now there are Physics Processing Units-based card to off-load all (most/some) of the physics calculation for games OFF the CPU. Given all this, why anyone would think that trying to cram ALL audio processing back onto the CPU is a good idea, or likely to happen, is completely beyond me.
- The ONLY reason a CPU can do any sort of useful processing is *THERE ARE DSPs EMBEDDED INTO THE CORES* (that's what altivec, MMX, SSE and all those "extensions" are.) The difference, though, the distance between those embedded-in-core DSPs and the DACs is HUGE, plus they have to share the CPU's interface and throughput with the rest of the sytem, i.e. system software, application software, etc etc, while if they are outside the CPU, they are 100% dedicated to the audio processing, AND they don't have to share their throughput with anything. A plugins-running-off-CPU architecture will NEVER be able to compete with this, forget it. Plus, with a CPU-only architecture, if you need to add processing power, you need to add a whole other CPU running at 2-3GHz (== MUCH bigger power consumption, much more complex circuits needed on the motherboard to handle the dual CPUs, they will still share a single interconnect/PCI-like bus, etc,) while with a DSP-not-on-the-CPU architecture you can just add another card and be done with it, and that same card will easily eat 10x less power than the CPU equivalent (compare a 15-DSP card 15W power consompution with modern CPUs that often go over 150W for the Intel ones.) This translates into lower costs, less heat, MORE reliable systems (since the warmer a circuit is, the less reliable it tends to get), and you don't need 7 fans that sound like airplanes to cool down your audio machine.
- I've heard this "DSP is dying! Software plugins will make hardware obselete" litany for years and years, and it hasn't gotten anywhere near tangible yet. Plus the whole modeling-hardware-with-software thing doesn't work well for some analog systems, since the mathematical modeling makes a couple assumptions on the Linearity and Time-Invariance side of things (i.e. LTI systems), and modeling BOTH at the same time still isn't anywhere near a solved problem. You can kinda model Time Variance with multiple frequency-response curves for different input levels/types, and you can kinda model Non-Linearity with Volterra series (or something), but it's for from perfect and all-encompassing yet. Plus there's the whole computer-as-instrument thing which really isn't too hot if you are not part of the automation crowd. I much prefer to have a stand-alone synth than a controller-keyboard-plus-computer/laptop combo, and that's just one example. Sorry, hardware isn't going anywhere.
To give you a proper analogy on the latency side of things, what you are saying is pretty much like saying that a factory having to ship all its products across town to get them package will save time versus just moving them next door, or packaging them in the same building as they are built in. And this is assuming a town getting MUCH bigger, with more people, more traffic, and MUCH more complex traffic lanes and traffic patterns. Sorry, but even in the BEST of cases, it'll be *the same speed*, but the next-door architecture will always have the lower cost.
Now if you could convince hardware manufacturer and operating-systems builders to trim down and optimize their systems, then you might eventually have some sort of a point, but right now they are making WAY too much money forcing everyone to upgrade the hardware and the software every 5 years, so I can only wish good luck =].
As for the specific Creamware vs UAD/PoCo type of debate, they both cater to completely different crowds. If you do pure automation and never work with realtime signals, then yes, of course, an UAD/PoCo type of system will be fine for you, while the Creamware system is MUCH nicer if you do any sort of recording/realtime tracking type of work. It is also a VERY dumb debate, because *you can easily use both type of cards in the same system, AND software plugins!!!* at the same time.
This being said, I'm not against software plugins, I'm 100% in favor of using all available tools and processing power in the course of audio production, software definitely have their place (and I use some and I love them, even though elementalaudio just sold out to roger f**king nichols and now sell their plugins for 2x 3x the price grrrrr (sorry, random rant =D)), and the nice thing about software plugins is that, as much as piracy is an issue, it also put awesome tools in the hand of people who could otherwise not necessairly afford them (kids, students, poor people, 3rd world developing countries, stranded alien spaceship pilots, etc), which is really great considering the completely messed up stat of most endoctrination I MEAN education systems around the planet.
It really isn't a sound-quality issue at all. Sound quality, as far as audio is concerned, is mostly a function of the algorithm developers, and the ADC/DAC part of the system. Some software plugins sound awesome, some sound like crap, some hardware/DSP plugins sound awesome, some sound like crap, and one person's crap is another's gold.
At this point, there is more than enough information available for any given individual to make an informed decision about computer hardware and software, and it's very much a function of how you prefer to work. Not everyone is a pure-automation VSTi type of person, and I, for one, would never want to let go of Creamware's routing system, even though I am considering buying a UAD for the Neve (and other) plugins. I also have a Virus and love every tiny bit of it (and it loves me back!) but I also love the Creamware synths (and the mixers. OMG. secret weapon!)
I find integrating outboard hardware to be much nicer and much easier with Scope than it is with a purely softsynth/sequencer type of system, i.e. I can load any of my hardware synths/outboard gear as a module in Scope! I can also do the same with software. Other people will prefer just VST/VSTi, and that's their right, and it's completely fine, and I don't see why one should exclude the other.
If you don't like the way Scope works, just get something else, no need to get all gloomy and negative and apocalyptic about the eventual future of the platform, just get something you like.
As for the whole no-upgrades-in-years, I wonder if you realize it's a good thing? On a *properly setup* XP system, Scope is definitely the most stable software I have. Some devices have glitches, but the software itself is pretty solid.
People who have crashes generally have improperly setup boxes or conflicting hardware, and it's indeed a HUGE challenge at this point to build and properly setup a functional DAW, and alot of people mistake Creamware-stuff instability with just standard Windows-crashes-all-the-time. Sorry, but just take a look at any forum pertaining to audio hardware/interfaces and software, and you will find exactly the same kind of discussions you will find here. Scope has glitches, for sure, my Virus also has glitches, Logic also has glitches, and hahah just look at all the Cubase SX3 threads!!! It's far from perfect. Windows is also full, FULL of glitches (and some serious bugs, which come from design flaws and can't be easily fixed -- and no Vista doesn't look like it'll be anywhere near better on this front =])
And I still don't get the drivers stuff. Did the XP driver infrastructure change in the last 5 years?? I don't think it did. It's still pretty crappy also, seriously the WDM drivers infrastructure isn't anything to write home about, and if you've done any DirectX/etc audio programming, you should know it's really on the painful/unreliable side. I've had a couple issues with a couple (not ALL) games and Scope drivers, but no show stopper, PLUS you can easily put another standard cheap soundcard for gaming if it's an issue. I only use Scope right now for my audio, and it works fine for music, videos, the games I use, etc. Can't imagine what the big fuss over drivers is, and drivers for a different operating systems are USELESS without the whole software to go with it.
Seriously, once 64bit OS are established and common (they aren't yet) and there is a demand, I don't doubt Creamware will port it. THIS BEING SAID, even if there's no upgrade to it ever, I can't say that would worry me very much. People talk about software-plugins as an all-in-one solution, but I know alot of people doing game music and audio contracts for movies and things, and they *all* end up having several computers (think 5 or 6) to run all the ultra-heavy sample colletions/gigastudio libraries they use for their contracts. So keeping a computer around for Scope really isn't an issue for me, and seeing how it's still easy and trivial to buy computer parts for PDP-11 computers that are older than me (thanks eBay!) I don't see any reason to freak out yet. Maybe in 20 years I'll start worrying/shopping around for something else, for the moment I'm so happy, I run full audio projects at like 20% CPU load, with only a single outboard synth. Ultra minimal setup, ultra maximal results. Audio production never has been as good as it is now!
Sorry for the endless rant =P
).
