• Registered: 2014-06-09
  • Posts: 19

Topic: Future of low-latency audio

Just an idle question. RME user for 4 years, complete convert.
So we all moan about latency - often it's what's led us here, getting stable low-latency gear. I just wondered, from a technical standpoint, what future changes might happen to PC architecture that will help with latency.
I'll happily admit I've only a rough understanding of how a USB ASIO driver actually works, but I'd guess that the future isn't there. I thought perhaps a low-level access system like Thunderbolt might help (though I daresay it'd be a nightmare to navigate the various implementations!).
Also, is ASIO itself a problem, and is there a way of making CPUs that optimises them for audio streams, so they're more real-time oriented than buffer-oriented?
I'm just dreaming here, not asking for concrete solutions - just interested to know what the future of 10, 20 even 30 years might hold for latency. And thought this a good place to ask!

  • ramses
  • DAW Mastermind
  • Offline
  • From: UTC+1
  • Registered: 2012-10-14
  • Posts: 13,194

Re: Future of low-latency audio

You need smth like a realtime kernel. Under Windows it goes like this roughly (further details on LatencyMon webpage):

http://www.resplendence.com/latencymon

If a device generates an interrupt, an interrupt service routine gets called which is a low level kernel routine and thus has higher prio compared to normal userland processes/threads.

If this part of the driver is written in a bad way, it can spend too much time and blocks the CPU from doing other things.

During the execution of this low level kernel routine its uninterruptable and can't be freed from the CPU.

From my understanding parts of Windows and Device Driver would need to be rewritten somehow.

But certain low level routines, like writing data to a disk maybe can't be interrupted at any time to prevent loss of data.

Its a complex topic I think ... I only know that for Linux there are patches which bring realtime functionality into the kernel.

For Windows I doubt that something like this will happen, as the vendors take more care of stylistic changes as to really innovate the product.

BR Ramses - UFX III, 12Mic, XTC, ADI-2 Pro FS R BE, RayDAT, X10SRi-F, E5-1680v4, Win10Pro22H2, Cub14

3 Reply by Potscrubber (edited by Potscrubber 2014-06-12 18:56:42)

  • From: Aotearoa
  • Registered: 2007-10-23
  • Posts: 255

Re: Future of low-latency audio

Without getting too technical and not addressing Ramses very good explanation above.. I've always thought Pyramix's masscore functionality was so obvious. It hides a certain number of CPU cores from Windows, and uses them for it's own low latency dsp. Apparently it's very powerful. Of course, this means software built with windows libraries can no longer run on those cores within Pyramix. Like vsts.

Madiface XT, Madiface, 3x Micstasy, ADI8QS
Sequoia 17, W10 x64
https://bsound.co.nz/tools-nix
  • Registered: 2014-06-09
  • Posts: 19

Re: Future of low-latency audio

Interesting. I was thinking the same about MassCore - seems almost too good to be true, to reserve CPU cores like that, but potentially extraordinarily powerful.
Agreed that Windows etc aren't interested in making the change - I mean, they can't even be bothered to do USB Audio 2! - but I suppose I had wondered if us audio users might be able to piggyback other technical developments, especially those with larger budgets in them. If servers needed this functionality, we'd find ourselves using it pretty quick!
Until then I guess MassCore style implementations are our best bet - I understand that they claim a possible 1.3ms latency, which I'll admit is as near zero as we'd ever practically need.