LinuxMusicians

Just some thoughts - corrections, suggestions, idle trolling most welcome.

Much of our contemporary PC architecture rests on standards laid down in the 80s. Digital audio and "multimedia" systems came later and still today it's like digital audio was an afterthought.

My first PC was built by a mum and dad shop in 1996. The CPU clock speed was 133MHz. It had a soundblaster clone 16 bit PCI sound card. It had a "Turbo" button on the front of the beige case. I looped and overlaid samples for some terrible sounding protosongs in CoolEdit95, downloaded from Tucows. Today we disable Turbo and limit our clocks to a lousy /sarcasm 3.7GHz.

My past 2 systems have both performed best at the same JACK settings: 256 frames, 3 periods, 48kHz. In that time the clock speed has risen a lousy (not /s) 700MHz. That's the same internal latency for 10 years.

It seems that digital audio is still an afterthought when it comes to what our DAWs are doing at the hardware level and this is a legacy within contemporary x86-64 builds that harks back to the IBM days. Before the Soundblaster!

So, I'm wondering if something like RISC-V and open source hardware will change the scene. Will it ever be possible to build a DAW around the audio I/Os and processing, rather than around a monolithic, all seeing, almighty, holy CPU? A CPU which cares as much for your audio as anything else it's doing.

Will I still be on 256/3 in ten years? Does it matter? What improvements do you see ahead?

It's been possible for years. I work at a popular venue with Digico SD10 mixing consoles which are over 10 years old. These have separate FPGA and SHARC® processors for audio and DSP, and run sub 2ms latency for 48 channels at 48kHz over MADI. 96 channels at 96kHz is possible with different AD/DA converter racks. These desks run Windows Embedded, based on XP IIRC.
I've actually just tested some tweaks on my system, a lowly i5 6500 with M-Audio 2496. Basic tests with xruncounter utility showed good results - first xrun at 96% DSP load at 128 frames, 2 periods.

My understanding is audio latency issues are rarely due to CPU speed but rather the motherboard design, as CPU speed has been more than capable of keeping up with audio streams for about 15 years. Case in point, I own two laptops with basically the same CPU. One is a workstation laptop and can do 64/3 with no xruns, while the other is a consumer-grade laptop that can only reliably do 256/3 (128/3 mostly works but still produces occasional xruns). The same USB audio interface is used on both. Here is a video that explains the reasoning:

https://youtu.be/GUsLLEkswzE

Motherboards can certainly introduce bottlenecks, however the generic CPU used in a desktop or laptop cannot match a processor that is specifically designed for low latency audio processing like a SHARC.