Native vs. DSP (low latency)

[stangroom]

Hey DUC!

I have a friendly wager with a co-worker of mine about the importance of low latency DSP plugin processing.

I said it wasn’t important anymore… new computers are fast enough to run native plugs just fine.

He said nope… still necessary for tracking and some post work flows.

We need you to settle this bet! There is a fiery burrito in store for the looser of this question….

Are they still important to you and why???

[nednednerb]

Depends on workflow and workload.

Complex studios, large sessions, so many plugins.... What happens exactly when the following arises:

1) your hardware insert is limited to real-time. You MUST stream an audio signal to it played off an audio track read off an SSD.

That means, for one, your fast computer cannot actually "lookahead" meaningfully. The mix playback for the next few milliseconds therefore depends on taking the hardware output back into a PT analog input.

2) your star vocalist needs to hear the Auto-Tune AND long reverb while tracking an overdub, but you have 15 seconds to press record, NOT enough time to bounce out and commit tracks and make a new session for overdubbing.

That means, for two, since you must work faster because it's the star vocalist who is the one whose business actually can afford the whole recording session ($$$).....

But all I need for a fiery burrito is to eat one... Save your moolah lol.

I think there is certainly a reason for more power in certain spots in the DSP workflow for very intensive and complex work.

Personally playing with 10-30 track sessions and with my own simple music and overdubbing my own things, sure probably nothing to worry about.

Sent from my Pixel 6 Pro using Tapatalk

[albee1952]

My own personal experience:
I ran a commercial studio in downtown Nashville for several years. My normal band recording template had roughly 60 tracks(including 12 AUX tracks and a MASTER). On those tracks were several dozen plugins(6x channel strip, 12x EQ III 7-band, 2x IK reverbs, 2x D-verbs, 1-2x Sparkverbs, 1 amp sim, 4x limiters, 8x compressors(SMACK and BF76), 2x INSTRUMENT tracks (DB-33, MiniGrand) and 4x Mute-o-Matic. In addition to sends for the reverbs and drum bus comp, all tracks had 6 stereo sends feeding my headphone amps for musicians. I used an HP z820 tower with an HD/Native card and pairs of HD IO and 192 IO(32 total preamps available for tracking). I worked at 24 bit/48K with a buffer setting of 64. Max plugin ADC was 11 samples(the amp sim). I never had any mention about delay in the phones and things worked beautifully, in native. The HP was a 12 core @3.5GHz with 64GB of RAM. SSD system drive and spinner recording drives(sample libraries on their own spinning drive) with an nVidia graphics card pushing 3 screens. Did I want HDX? Yep. Did I NEED HDX? Not for my workflow or my clients.

[BScout]

Originally Posted by stangroom

Hey DUC!

I have a friendly wager with a co-worker of mine about the importance of low latency DSP plugin processing.

I said it wasn’t important anymore… new computers are fast enough to run native plugs just fine.

He said nope… still necessary for tracking and some post work flows.

We need you to settle this bet! There is a fiery burrito in store for the looser of this question….

Are they still important to you and why???

Take the loss. Eat the burrito. You are wrong.

In tracking, low latency is king and CPU power doesn't make a difference because the delay happens on crossing the buffers to the CPU and back out. Not whether or not the CPU can crunch the numbers once it is "inside" the operating system's domain.

On an FPGA mixer (like HDX), audio doesn't have to go through those. All on chip.

[LDS]

The use case for DSP has definitely lessened over the years with the improvements in native processing and removal of limits in Pro Tools, but it definitely still has a place. The main benefit of it is the dual buffers.

Just as an example with the HDX hybrid engine on my 10-core iMac Pro - a dozen or so virtual instruments, 30 to 40 audio tracks, with bussing and processing in place will pretty easily throw playback errors if I try and punch in new recordings at 64 samples using just native processing. Flipping the 30 or 40 audio tracks and bussing to DSP drops the CPU usage by a factor of 3 or 4. Everything runs smoothly. I can punch in. I can add more VIs. I can toss HDX into classic mode and I can do weird and wonderful routing right across the studio in a fashion that is remarkably similar to an analogue desk. Buss VIs, send them to outboard synth modules for mangling, process them with plugins on the return, bus them together and process them in the mix, add in more sounds on the fly from anywhere, really.

It had a weird few years there in the middle, with things like voice limits getting in the way but HDX has become quite a different, yet incredibly succinct sound creating tool here with all the recent changes like Hybrid mode. I personally wouldn't bother with it for straight forward band tracking, or if I were just mixing. But for unifying a studio and all of its available sound making and mangling bits and bobs, there still remains few systems out there that can do it as well has HDX.

[ejsongs]

In addition to the many things I currently do, I have been and still am a professional session guitarist in the Los Angeles music scene. I have been for over 20 years.



As a player I can feel the difference and much prefer tracking on protools DSP based systems. It affects the way that I play. I am very sensitive to such things. The following is a true story:


I did a session last week and the guitars that I tracked were flown into the mix session. It was a sample replay. Anyway, the producer hit me up and said that something was wrong as the magic of the ref sample and the replay that I did was gone somehow in the mix session. I heard the record and agreed that something was amiss. I had them send the mix session to me. I immediately heard what was wrong. I pocket of how I laid the gtrs didn't feel like what I had played or sent them...it simply didn't sound like me. I opened the session and figured out the issue. The flown gtrs were 10 samples earlier than they were supposed to be...yes only10 samples off at 48K. I simply nudged the grts back by 10 samples and Voila!!! The vibe was back. I sent the tracks back and now everyone is happy again...true story.



I say this to say that subtle latency does make a difference. i can feel the difference when tracking with dsp vs native. Imn my own productions, I have recut stuff in pt hdx simply because i couldn't the feel i was after on my native rig. Native simply isn't tite enough for me. I've run a focusrite interface at a 16 buffer at 48k in s1 and it still wasn't as tight as pt hdx. M1 chip or not, I will always have a dsp system for this reason.


Ej