I'm finding the latency of my Lynx Hilo over USB to be rather system dependant, back when I was using an i7 3930k and Asus P9x79 to drive it 128 samples buffer was the lowest I could use glitch free.

Since I upgraded to a Ryzen 3800x and MSI X570 Tomahawk motherboard, 64 samples at 44.1khz (Lowest setting in driver) and 128 samples at 96khtz is easy and glitch free.

My Motu 424 PCIe card interfacing two 2408 mk3 and a 24IO should be even better

I'd be really interested to know if it is. That's what I'm using on my main stationary system and you described my upgrade plans.

I hear the Ryzen 5000 series have even better latency, which I am planning to upgrade to a 5900x later on when prices drop a bit

Just hooked up my Motu PCIe 424 card with one of my 2408mk3 soundcards, on MSI X570 Tomahawk, Ryzen 3800x, 3600mhz CL14 ram, put it at 96khz, 32 samples buffer, fired up Pianoteq6

0.3ms latency, and glitch free! :D
But at max pedal sustain some minor pops

At 64 samples, it's fine, 0.6ms latency is still pretty good...

Tried the same in Reaper, the same
0.8/0.9ms latency in and out
3.14ms/0.67ms hold, RT Longest block according to Reaper's performance meter.
Had to set Pianoteq to 48khz internal sample rate and limit polyphony to 24, otherwise on max sustain and 24 notes, CPU can't keep up, even thouh Pianoteq only takes 0.8% CPU at these settings,
but at 128 samples it's fine

My Lynx Hilo over USB tops out at 128 samples at 96khz, but can do 48 note poly of Pianoteq max sustain at 96khz internal rate
2.3/2.3ms latency in and out
3.14ms/1.33ms hold
MOTU 424 PCIe can handle 128 notes at the same settings
1.5ms/1.5ms latency in and out
5.13ms/1.33ms hold - RT Longest block

So if the Ryzen 5900x is better here, I have no complaints

Can't beat PCIe, but these MOTU drivers have not been updated since Windows 7 days...

Here is a video of small project running in real time on both of my Sound cards running under ASIO4all driver at the same time, in Reaper 6, 24 bit 96khz, at 64 samples buffer, 0.6ms latency

Ain’t ASIO4All a wrapper of the WDM drivers, which have nowhere near the latences ASIO drivers have. ?
IIRC the 64 sample buffer ASIO4All provides is just a facade (or the tree masking the forest, to take Vin’s image ; )), ASIO4All takes 64 sample buffers but it has to cope with the WDM drivers behind it, that have all kinds of buffers and way higher latences; and ASAIC I’m pretty sure the when you aggregate several WDM drivers, like in your case, it adds some more latency.

You’d have to make some measurements with the RTL utility and see if you really/actually have 0.6ms output latency or supposedly 1.2ms RTL

Not sure either way
But the post before that using the native ASIO drivers for each sound card is legit though

Just that I can't use both setups at the same time with the native Asio drivers....

edit - although I can just feed my MOTU cards into the Hilo via ADAT and digital IO
that way I can have lwo latency and then mastering quality audio monitoring
I'm already clocking them off the Hilo

edit 2- just setup the MOTU 424/2408 mk3 as the main ASIO device in reaper, 96khz 64 samples works fine.
ADAT out to Hilo for monitoring.
HILO analog in for recording and output back into ADAT in on 2408 mk3

This also allows me to use the Hilo for playing music in Windows at the same time and sending that into the 2408 mk3 for recording as well
Much more flexible this way as ASIO grabs the whole card

Hello,

Because I haven't seen this interface in this post and just got it. Here are my results for the VRS8:

On a pc with windows 10, core i7 3770 (pcie card with hdmi cable). I don't know why but I can't go lower than 48 buffer in windows (this PC is dual boot with OSX).

Thanks for the heads up, nice to finally see a full RTL report on Windows.

Why is the Reported RTL double the measured RTL ?

What is the reported In / Out in the DAW's ?

Measured numbers are in the same range of the better TB devices on the Database , no idea re why they tapped it at 048 , that would be a question for their support team.

That driver has been a very long work in progress, the last piece of the puzzle is the actual efficiency and stability at respective latencies.

Do you have another interface to do some comparative tests against ?

I don't know why the Reported RTL double the measure. I hadn't noticed that.

Same problem in the DAW (Cakewalk). At 96 000 and 128 buffer I got 5.5 ms.

I have only an audient id4 to compare. I'am more a Logic user and I will check the latency on OS X with the same PC.

Now for the audient id4 (windows) :

So, do these number's mean it's good or bad?

Has anyone done a thread on ASIO vs CoreAudio?

Just throwing this out there as it doesn't really make any sense to me.

I was on a Mac OS X system for the longest time and using CoreAudio with my interface. The Mac was getting long in the tooth, so I recently switched to a Windows based DAW. It's all the exact same software applications, same interface and same speakers and DAW. The settings of the projects are also all the same -- 48k 24bit. The only thing that has really changes is that I am on Windows and using ASIO drivers and the RAM is more and a different CPU.

For whatever reason -- it doesn't sound as good as it sounded on the old Mac. New projects I am starting as well as projects I copied over from the old setup. I keep getting the impression it's more trebbly and more high endy sounding -- everything I am hearing now. It doesn't sound as warm as it used to.

Is this possible or does one need to adjust and compensate when using the same hardware/software on Windows?

https://docs.microsoft.com/en-us/win...s-in-windows-7

From that link.

Honest question. What's that mean?

I'd call it a living definition of 'aspirational'....with absolutely no proof to back the assertion

I can guarantee if you've set everything up properly, i.e. no sample rate missmatch etc then it will sound exactly the same.

You have what's known in the trade as 'perception bias'


regarding ASIO and Core audio. ASIO is more efficient at a kernel level as it requires fewer kernel calls per sample. It's PCM digial audio it sounds the same.

if you're in any doubt:

with the same hardware, bounce a file from the mac ( don't use any time based FX) do the same from windows and then do a null test.


M

RTL wise, the VRS8 mesures better than any RME interface, but less than the PreSonus Quantum, and is on par with the Lynx Aurora LT/TB.

That's what I suspect as well. It has been about a week or so. I will give it another week I guess. But I think everything is set the same. The interface is set to 48k. The project settings are set to 48k. The plugins I set to 48k. I can't think of where else to check.

The dreaded confirmation bias can do awful things to our perception. I'd find ways to carefully/fairly/immediately compare the two DAW/playback environments before pursuing the concept that ASIO vs CoreAudio is responsible for perceived differences.

I did a post a few months back that may be helpful with respect to latency differences. I used the same interface in both Win10 and macOS. Results are in chart form; all done at 48k. Link is below.

Audio Interface - Low Latency Performance Data Base

Very Strange , I have seen the opposite from drivers with the same heritage ( ESI) , where they report nominal and have far higher I/O and RTL , this one is flipped.

Something is obviously off there regards its reporting via the ASIO driver, that would be sending the delay compensation algorithms on the DAW's out by significant amounts for not only recording, but also PDC, etc.

If you can ask the question via support why its reporting incorrectly, and give us a heads up what they respond with, that would be informative.

Cheers, the full OSX results for the VRS8 would be good.

re comparative performance, I was referring more to the driver efficiency at the respective latencies , which is equally if not more important.

The Audient isnt really going to give it a decent head to head run unfortunately , but thats cool, the complete measured RTLs for OSX and windows is valuable to know.

Hopefully I can get one across the bench at some point to do the final efficiency and stability testing.

I wrote to the support this morning. I asked for the 32 buffer size and why its reporting incorrectly.

Now for the OSX side (but it's not a real mac !) with the pcie card, the performance (rtl) is a little bit worse. But I'am happy that it just work.

I want to build a new PC this year or wait a driver update and buy the new mac with the M1 chip. The performance of the VRS8 seem very good (until now) and with the ML-1, I think it was a good deal.

Asio does not use Windows Audio APIs, it is specifically built to circumvent them. So this does not apply here

DAWs can sound different in how they process sound. Usually, though, that's subtle at best, and modern DAWs are more similar than different. The same brand and version of DAW, across two operating systems, is unlikely to have any differences at all. But some claim that, say, Cubase sounds different from Pro Tools sounds different from Ableton Live. I've never performed a comparison myself.

Otherwise, if using the same audio interface, with ASIO, there's no way for anything to interfere with the sound on Windows. There's nothing to compensate for. It's all digital with no processing. ASIO is a thin a layer as you can get over the audio interface, so you're getting about as pure a stream as you can get from the interface. CPU and memory will not impact this.

CoreAudio on macOS does do other processing under the hood, depending on resampling, interface aggregation, etc. But with a single interface and everything running at the same bit rate/depth, I wouldn't expect to have any real processing, and certainly nothing that would act as a low-pass filter (what most people mean when they say "warm"). That said, if you were using the built-in audio chip on any laptop (including macOS and Windows), you'll hit all sorts of processing to compress the sound, boost bass, Dolby, etc.

Do you have renderings to compare between the two? You could load each uncompressed render into a spectrum analyzer. If not, and I mean this in the most respectful way possible, it may just be in your head. We're all guilty of a little confirmation bias from time to time.

Pete

Rendering an audio from a DAW won’t tell you the whole truth. The whole truth would come from capturing a PCM stream to the hardware and comparing -that-!

Well, someone did! Same mix, same plugins, same settings: https://youtu.be/eJfe-A9Ebrc

there are two places OP needs to look if they think there are differences.

1. What the DAW is doing on that OS. A rendering should sort that.

2. The actual output to the audio interface: this is what you recommend. Only real way to capture this would be to grab a digital audio output (if available) from the audio interface. There's no reliable way to intercept the ASIO stream and in CoreAudio at the same point in the process. Not that I know of, anyway.

You could record an analog output, but that introduces more variables.

Pete

The only scenario where PC vs Mac or different DAWs could sound different is when something was broken. The math is always the same, consequently the results are the same.

There are people who have tested this, over and over, myself included.

Why are we still talking about this in 2021.

Computer science ain’t maths. Computer science is application of maths in a realm where only binary numbers exist and everything’s limited — as contrary to maths — and where it exists a million ways to skin a cat.
If you’re not a software engineer and don’t have a suffissent experience in this field, you cannot grasp the behind the scenes of a DAW. Even then, since you don’t have access to the source code, you don’t even know what’s exactly happening behind the scene.

Hear for yourself, same mix, same plugins, same settings: https://youtu.be/eJfe-A9Ebrc

In 2021 you’d better off putting more truth in your believes!

Pulling this thread back on topic !

Cool, it will be interesting what they come back with.

Notice the reported and measured are within range of each other ( only 5-6 samples variance ) , which just makes the Windows reported vs measured even more pronounced.