USB audio drivers and async transfer: what your DAC is actually doing with your computer

Every few months someone emails me a variation of the same question: "I bought a decent USB DAC and it still sounds harsh — did I waste my money?" Nine times out of ten the answer isn't the DAC. It's how USB audio actually works, and why the default setup on most computers quietly undermines it.

I've spent enough time listening through DACs connected via USB — everything from a Chord Electronics Mojo 2 on a laptop to full desktop rigs pushing an Audeze LCD-X — to have strong opinions about this. And I'll admit I got some of it wrong for years. The cable is almost never the problem. The driver and transfer mode usually are.

What USB audio actually is (and isn't)

USB wasn't designed for audio. It was designed for keyboards, storage drives and printers. The audio class was grafted on later, which means it carries some architectural baggage that doesn't suit low-latency, jitter-sensitive playback particularly well.

When your computer sends audio over USB, it doesn't push a continuous stream of samples. It packages audio data into small chunks — called packets — and sends them at intervals governed by a clock. The question of whose clock controls timing is the crux of everything. The host (your computer), the device (your DAC), or some negotiated middle ground: that's what transfer modes are about.

USB audio has three transfer modes: adaptive, synchronous, and asynchronous. Synchronous is essentially unused in modern audio hardware and I'll leave it there. The meaningful split is adaptive versus asynchronous.

Adaptive transfer: why it matters less than you think

In adaptive mode, the DAC takes its timing cues from the host. The computer is the master clock and the DAC adjusts its internal sample rate to match whatever the host sends. The problem: computers are extraordinarily noisy electrical environments. Switching regulators, GPU bursts, background OS processes — all of that activity creates timing irregularities that end up at the DAC's clock input.

The technical term is jitter: timing variation in the digital signal. Jitter at the DAC's converter stage doesn't produce distortion in a simple, audible way. It manifests as low-level noise and sideband artefacts that can make the treble sound smeared or harsh rather than clean and extended. This is precisely the difference I'm listening for — real high-frequency detail versus brightness that's actually artefact.

Adaptive USB isn't inherently terrible. A well-implemented adaptive receiver with a good VCXO and a tight PLL loop can manage adequately. But it's asking the DAC to do extra work compensating for a noisy host, and cheaper implementations don't always manage it convincingly.

Asynchronous USB: the DAC takes charge

Asynchronous transfer mode flips the relationship. Here, the DAC's own clock is the master. The DAC tells the host how fast to send data, rather than the other way around. The host buffers audio and delivers packets on the DAC's schedule. Because the DAC's clock is isolated from the PC's electrical environment — in good implementations, it's an ultra-low-jitter oscillator running independently — the digital-to-analogue conversion happens with far tighter timing control.

This is why essentially every serious USB DAC released in the last decade specifies asynchronous USB. The Chord Mojo 2, Cambridge Audio CXN100, and most standalone DACs at $500 and above all run async. It's not marketing language; it's a measurable architectural advantage. You can see it on a spectrum analyser as lower jitter sidebands around a test tone.

The short version is: async puts a high-quality clock inside the DAC and makes it the authority. Adaptive puts a noisy PC clock in charge. Given the choice, always prefer async.

USB Audio Class 1 versus Class 2: what the driver question is really about

Here's where Windows and macOS users have different experiences, and it's a source of real confusion.

USB Audio Class 1 (UAC1) is supported natively by Windows, macOS and Linux without any additional drivers. It's limited to 24-bit/96kHz and two channels. Many simpler DACs, including portable ones designed for phone use, operate as UAC1 devices. Plug them in and they work, no installer required.

USB Audio Class 2 (UAC2) extends this to 24-bit/384kHz and DSD, among other capabilities. macOS has supported UAC2 natively since around 2012. Windows does not support UAC2 natively below Windows 11 22H2, and even on current Windows the native driver implementation is limited enough that most manufacturers still ship their own ASIO or WDM driver. If you're running a Windows machine and you haven't installed the manufacturer's driver for your DAC, you're almost certainly not getting the best out of it — and you may be running at a lower sample rate than you think.

I've watched this trip people up in listening sessions. Someone will bring a nice DAC to a meetup running on Windows with no driver installed and wonder why their hi-res files sound no different from CD. Check Device Manager. Check what sample rate Windows thinks the device is running at. Nine times out of ten, it's sitting at 48kHz because the OS has defaulted it there.

ASIO, WASAPI and why the audio path through the OS matters

Even with the right driver installed, Windows has a habit of mixing all audio through a software mixer called the Windows Audio Session API, or WASAPI, and resampling everything to a single "preferred" sample rate before it reaches the driver. If your DAC is set to 192kHz in Device Manager but Windows is internally running everything at 48kHz because Spotify or a browser tab is also open, your hi-res audio is being upsampled from 48kHz. That's the opposite of what you want.

The solutions are WASAPI Exclusive mode — where your playback application bypasses the Windows mixer and writes directly to the driver — or ASIO, which most manufacturers implement via the ASIO4ALL wrapper or their own proprietary driver. Roon, Foobar2000, JRiver and most serious playback apps support both. If you're listening through Windows Media Player or Spotify's Windows client without WASAPI Exclusive enabled, you're not in control of your signal path.

macOS is cleaner by default, largely because Apple maintains tighter control of the audio HAL (Hardware Abstraction Layer). Integer mode in applications like Audirvana bypasses CoreAudio's volume control processing, which can introduce dithering. It matters at the margins. Worth enabling if the option exists.

Linux users, particularly those running Roon or HQPlayer, often have the cleanest path of all — direct ALSA access with no resampling — though "clean" comes at the cost of setup complexity that most people aren't here for.

USB cables: I'm going to disappoint some people

Right. I know some readers will push back on this, and fair enough. But here's where I land after years of listening and reading the measurements: in an asynchronous USB implementation, the cable's electrical properties — resistance, capacitance, shielding — affect signal integrity on the data lines, but the async architecture is specifically designed to re-clock and correct for exactly these kinds of variations. The DAC's internal clock doesn't care about cable capacitance because it's not taking its timing from the cable.

What you do want from a USB cable is reliable data transfer — no dropouts, correct shielding to reduce RF ingress, adequate conductor quality. A decent cable from a reputable brand does all of that. Spending $300 on a USB cable for marginal electrical differences that async transfer largely corrects for is, in my honest view, not the most productive use of your budget. Spend that on a better DAC or a better headphone amp.

There are scenarios where cable quality matters more: if you're running very long runs (over 3 metres, where signal degradation is real), or if you're on an adaptive USB device, or if your PC's USB implementation is genuinely problematic. In those cases, a USB isolator or a dedicated USB audio interface card can make a meaningful difference. But a boutique cable between a competent async DAC and a modern computer? I'd rather you save the money.

Practical setup: getting USB audio right on a real computer

Let me walk through what I actually do when setting up a USB DAC for serious listening. I run an Audiolab M-DAC+ off a MacBook Pro when I'm working at my desk in Adelaide, and the same chain occasionally off a Windows laptop when I'm travelling.

On Windows: install the manufacturer's driver first. Reboot. Open Device Manager and verify the DAC is listed under "Sound, video and game controllers" with its actual name, not "USB Audio Device". Open Sound settings, select the DAC as default, click Properties and set the default format to the highest sample rate your DAC supports — typically 24-bit/192kHz or 24-bit/384kHz. Then in your playback application, enable WASAPI Exclusive or ASIO. In Roon, that's under Settings > Audio, selecting the device and enabling Exclusive Mode. In Foobar2000, install the ASIO component and select your DAC's ASIO driver. Verify no other application has audio open when you press play — Exclusive mode will refuse to start if something else holds the device.

On macOS: open Audio MIDI Setup, select your DAC, and set the format to match your source material — 24-bit/96kHz for most streaming hi-res, 24-bit/192kHz if your DAC and music library support it. Enable Integer Mode in your playback app if available. That's largely it; macOS doesn't fight you the way Windows does.

One thing worth doing on both platforms: check that your playback app is reporting the correct sample rate back to you. Roon shows this in the signal path display. If you're feeding 96kHz material and the chain shows a resampling step you didn't ask for, something upstream is converting. Fix that before you buy anything.

For headphone listeners building a more complete desktop rig, the USB DAC is one piece of a larger chain — you can see how I think about the full setup in our guide to the best DACs and network streamers, and there's more on desktop rig building in the best audiophile headphones guide where source quality comes up repeatedly. If you're also running a standalone headphone amp, the output impedance and voltage swing of the DAC matter — worth cross-referencing with the Sennheiser HD 660S2 and Focal Clear MG (check price) reviews where I go into source matching in some depth.

Does any of this actually change what you hear?

Getting the driver and transfer mode right isn't an audiophile ritual. It's basic signal hygiene. The cleaner and more controlled the digital handoff from computer to DAC, the more the DAC can do what it was designed to do. Whether you're running a Chord Mojo 2 (check price) off a laptop or a more ambitious desktop stack, the USB configuration is the unsexy prerequisite that either validates or undermines everything downstream.

I've heard people declare that expensive DACs are indistinguishable from cheap ones, usually while running both in adaptive mode through a Windows default driver with the OS mixer wide open. That's not a fair test of the hardware. It's a test of whether the computer's audio path is broken, and it often is.

Sort the driver first. Get async transfer confirmed. Bypass the OS mixer. Then listen. The improvements won't be dramatic — this isn't the difference between a $200 and a $2,000 headphone — but the treble will be cleaner, stereo images will be a bit more stable, and you'll stop wondering whether the harshness you're hearing is the headphone or the source. Spoiler: it's usually the source.