What's the future of audio PCs gonna be like

[Theo Verelst]

Not as disclaimer, but I come from the field where designing computer circuitry is a professional target, so I'm like either biased or more savvy, or what have you.

 

I made designs for virtual synthesis myself (for instance and I know about PC architecture and where to get good Ips/Flops (instructions per second/floating point operations per second, usually measured in Mega or Giga, nowadays even Tera quantities). Apart from bandwidth matters like 500 48k/16 bits tracks costing your hard drive only 500*48000*2 ~= 46 Mega Byte / Sec which fits over a standard USB 2.0 wire, there's storage amount which for an hour of 500 DAT quality tracks is about 160 Giga Byte. Not very expensive for a whole CD worth of a ridiculous number of tracks, but going to 32 bit floating poimt and a more professional 96 kilo samples /s that might require glass fiber to do fun stuff on the internet with, like putting your multi-track effort on the cloud. I recently did an small experiment with single track at the time (how are you going to record those tracks in practice ?) on the AWS cloud which I made a little video of in this computer lab thread:

A simple cloud audio processing example

 

The point being backups and processing loads are probably an important future cloud thing, for instance when needing an immense amount of plugin compute power that is available on demand, with great reliability and professional specs. For instance you can rent time on a 150 Tops/sec FPGA node, or run a thousand virtual thread processors to do your favorite processing on each individual track, or farm out some reverb algorithm to 10 high powered graphics card compute facilities.

 

Not important? Well it is technically quite possible to make ROMplers on dedicated computer devices with a few samples of (fixed !) latency time, which appears still to be a gap in the market, however, if you have to do quality sample reconstruction, mathematics dictate a high latency oversampling step. So unless you samples and the processing of your plugins is going to take place in very significant, like an order of magnitude higher rates you might want to prepare your samples for more mundane DACs, which as I've worked on takes considerable processing capacity (as it is won't fit on even the fastest Intel in one pass).

 

It is thinkable that a seriously set up programmable logic device can help the common PC architecture to overcome latency issues not befitting the general speed of the modern PC. Whether or not nobody will do nothing with that is another matter altogether!

 

Theo