What is audio buffer size?

The short answer

Digital audio is a stream of tiny measurements called samples, taken tens of thousands of times a second (44,100 or 48,000 times a second is typical). Your computer does not handle those samples one at a time, because stopping to process every single one would be hopelessly inefficient. Instead it collects them into small batches and works on a whole batch at once. The buffer size is simply how many samples go in each batch.

Buffer size is measured in samples, and you will see values like 64, 128, 256, or 512. A buffer of 128 means the system fills up a container with 128 samples, processes all of them together, sends them on, and then starts filling the next container. Smaller number, smaller batch, more often. Larger number, bigger batch, less often. That one setting quietly governs how responsive and how stable your audio feels.

How buffer size works

Picture a conveyor belt feeding a worker. Rather than hand over one item at a time, the belt loads a tray and the worker deals with the full tray in a single motion. A small tray gets carried over quickly, so the worker reacts fast, but they are back and forth constantly and can get overwhelmed. A big tray means fewer trips and a calmer pace, but each item waits longer for the tray to fill before anything moves.

Buffer size behaves exactly like that tray. The batch has to fill before the system can process and release it, so the buffer directly sets a floor on latency, the delay between sound going in and coming out. At a 48,000 samples-per-second rate, a 128-sample buffer holds about 2.7 milliseconds of audio, and a 512-sample buffer holds about 10.7 milliseconds. Every stage in the path waits on that fill time, which is why the buffer is the single biggest lever you have over responsiveness.

So why not always pick the smallest buffer? Because the batches also arrive faster than the system can always keep up with. With a tiny buffer, your computer has to wake up, process, and deliver a batch far more often, which raises CPU load. If the machine is ever late finishing a batch, even for an instant, the audio stream runs dry and you get a dropout: a click, a pop, or a burst of crackle. A larger buffer gives the system more slack to finish each batch on time, which is why it is the safer, steadier choice.

Buffer size versus latency

The two are joined at the hip, but they are not the same thing. Buffer size is the setting you choose. Latency is the delay you get as a result, and it is only one part of that delay. Buffer size sets the largest, most controllable chunk of monitoring latency, but the full round trip also includes converter time, driver overhead, and any DSP processing in the path.

• Smaller buffer. Lower latency and a more immediate, play-in-real-time feel. The cost is higher CPU load and a greater chance of dropouts if anything on the system hiccups.
• Larger buffer. Rock-solid, glitch-resistant playback that tolerates a busy machine. The cost is added delay, which you notice most when you are monitoring your own voice or playing an instrument live.

Because the buffer sets the floor and everything else stacks on top, a clean, efficient audio path matters just as much as a small number. For the full picture of what adds up between input and output, see what is monitoring latency.

Which buffer size should I use?

There is no single right answer, because it depends on what your machine can comfortably sustain and how sensitive your task is to delay. The honest approach is to match the buffer to the job rather than chase the lowest number for its own sake.

• Live monitoring, voice, or performance. Reach for a smaller buffer, often 128 or 256, so you hear yourself with minimal lag. Drop it as low as your system stays clean at, and back off the moment you hear crackle.
• Mixing, editing, or heavy effects. A larger buffer like 512 or higher trades responsiveness you do not need for the stability you do, letting a loaded project run without glitches.
• General listening or calls. A middle value is usually fine. If you are not playing in time with the audio, a few extra milliseconds is nothing you will ever notice.

The reliable method is to start a little high, then lower the buffer one step at a time until you hear the first dropout, and settle one step above that. The best buffer is the smallest one your specific machine runs cleanly, and that varies with your CPU, your other software, and the driver doing the work.

Buffer size in patchd

In patchd, a Windows virtual audio mixer, the buffer size comes from your audio interface's own control panel, and patchd rides whatever you set there while reporting the resulting latency live in its master clock readout, so the tradeoff stays yours rather than something fixed for you. The engine is real time, and it runs over ASIO and WASAPI Exclusive, the low-overhead Windows audio paths, so the work happens with as little added delay as possible on top of whatever buffer you pick.

At a small ASIO buffer, the engine adds about 10.7 ms routing your audio through a bus at a 512 buffer, and smaller ASIO buffers take it lower, low enough to talk and monitor yourself live. That headroom is what lets you run a full effects rack, a DSP chain of a noise gate, EQ, compressor, and more, without the delay stacking up into something you can feel. A bus in patchd is a virtual output that other apps pick up as a microphone, and routing is color coded to stay readable, with hardware outputs in green, amber, and coral and buses in cyan, violet, and magenta.

The takeaway

Buffer size is the number of samples your system processes in one pass, and it is the main dial between two goods you cannot fully have at once: low delay and rock-steady playback. Smaller buffers respond faster but ask more of your CPU and risk dropouts. Larger buffers add delay but stay smooth under load. Pick the smallest buffer your machine runs cleanly for the task in front of you, and let a low-overhead audio path do the rest.

patchd is pre-launch, with a free effects rack on every channel and an engine that follows whatever buffer your interface is set to, showing the resulting latency live. If you want a Windows mixer that keeps latency honest while your audio stays clean, join the waitlist and get notified when it ships. The paid Studio tier is $39.99 per year, and the low-latency engine is identical on both tiers.