UR 824 ASIO Buffer Size


When I set the Asio buffer size to less than 192 samples the sound is more or less distorted.
Im running Windows 7 64 bit on a laptop with Core i7 CPU Q 820 @ 1,73 GHz and 8GB ram. I must admit I did exspect much better performance. Any advice how to get better performance?

Best regards

Just by comparison: on my MacBook Pro Early 2008 (Dual Core Duo 2.4, CoreAudio driver) I can play Mussorjski’s Baba Yaga on Kontakt 4 at 32 samples (5ms) (Vienna Piano). I had a couple breakups when playing some extreme notes and the samples had to load. Then everything went perfectly fine.

The CPU on this slow machine was peaking at 50% on denser passages with pedal down. Most of the time it was around 25%. When increasing the buffer size to 256 the CPU indicator barely moves.

At the same time, I must say that I noticed sort of a “frying” noise when idle, and the character of this sound changed with varying the buffer size. It totally disappeared when relaunching Kontakt. Maybe some “dirt” remained from my furious hammering at 32 samples?

Am I wrong, or this noise means bad clocking?

EDIT: The noise reappeared after I had to force-quit Logic 8 (while trying to open an older project that Logic can no longer open: thank you, Apple! @#!!)


I get this noise when idle as well. Performance is not great on these. Even though they sound good, I’m looking at a UFX in 2012.

I restarted the Mac, and the noise is still there, with no application open. I cannot but suspect clocking problems. Is the original poster calling this kind of noise “distortion”?


The distortion I get is a cracling digital noise when sound passes through the interface. Så when I play a virtual instrument i get a distorted signal when setting the Asio buffer size below 192 samples. The lower buffer size the more distortion.

So now I can play my virtual instrument with an output latency of 11,5 ms which is much more than I expected and much more than my previus RME usb audio interface

Any advice on how to get better performance?


With a little better performance, I’d be happy with the unit. Not sure if a software update will make any difference however.

So, no, this is not the same noise. Yours appears to be depending on the buffer not filling in time.
How many notes are your trying to play, when this problem appears?

So now I can play my virtual instrument with an output latency of 11,5 ms

It should be a low enough latency to play (hearing resolution should be around 20ms, if I’m not wrong).


It all depends on specific requirements. A semi-pro or good home studio setup is looking for less than 10ms of latency for live recording. See the link below for a quick summary (disclaimer - no connection):

Apparently even 3ms of latency can be too much for vocals and drummers want about 6ms or less. Yes, keyboards can get away with around 20ms although, (according to SOS article), most players seem to prefer about 10ms; but the UR series is not marketed as a keyboard specific interface.

I can easily understand that some people find the latency of the UR series either acceptable or unacceptable depending on their specific requirements but what I find disturbing is the apparent lack of input from Steinberg on this subject of latency.

As a user of Cubase, I am understandably very interested in acquiring the UR824, but not at all interested in purchasing one if it has a lower equivalent latency than my RME card currently in use (i.e. when using the same PC or Mac). Why? - because recording vocals, live recording and direct monitoring are important.

A I/O latency of about 6/8 ms with a buffer of 256 samples is a common ballpark performance for many interfaces.

Is it not possible to obtain some official input from Steinberg on this subject of latency? It is a subject that can be found on more than one thread now.

I’m holding back on the purchase of a UR824 for the moment, that is, until I can see that improved integration with Cubase is not going to come with reduced usability.

Since this interface has zero-latency monitoring, the recording delay decribed by SOS is not (or might not, depending on your setup) be concerned. Theoretically, if you choose hardware monitoring instead of software monitoring, the input signal goes directly to the output, with a 2ms-delay. Add to this compression, EQ and reverb, and you have one (or a few) ms of delay more.

The beauty of having effects on the input channel is that you can avoid using software monitoring for adding effects to the voice/instrument during recording. You can simply rely on the good effects supplied by the audio interface, and have the minimal delay that direct monitoring allows.

Concerns may arise when you use software monitoring, for example when you want to record with a plugin effect. I’ve not tried this, but I can say that software monitoring my (dry) guitar with Logic set to 256 samples (@44.1) did not result in any perceivable delay. But a drummer friend of mines could easily perceive delays in the MOTUs and Digi’s we tried, that I could not hear…

As for virtual instruments, as I reported in another thread I could play a sampled piano live with a 32 sample buffer with just a couple initial “note stealing”. Playing at 256 samples is perfectly fine, and causes no harm to the flow of notes. Playing a full orchestra (Sibelius 6/Kontakt 4) was perfectly fine at 1024 samples, while I suspect a setting of 512 samples caused a short break in a busy passage. I’ve no comparison, since I always left Sibelius at the default setting of 1024 samples.

Effect plugins may be another matter. Just for my personal curiosity, I tried with a modified version of the DawBench test file, by using eight Logic’s Multipressor effects per channel. (Multipressor is a heavy multi-band compressor, but I don’t know how it compares to the plugins used for the original test). With my old MacBook Pro Early 2008 2.4 I could play 12 ‘Sine’ tracks before audio breakout. But I don’t see myself using 96 multiband compressor anytime soon…


Thanks for the information Paolo.

I can see that you are happy with the UR824, and there is nothing wrong with that, it obviously does what you want with little or no problems. I know the sound quality of the UR824 and MR816 is very good and if you can use the “zero-latency” function then they are obviously a good buy.

However, the “zero-latency” setting is not good enough for most of my needs (electric guitar and keyboards with plug-ins, vocals through headphones with software reverb etc…) so the latency performance when software monitoring is crucial for me. “Zero-latency” doesn’t mean anything when you have to monitor through software plug-ins.

Guess I’ll stay with the inherently faster PCIe setup with analogue effects (frequently more expensive but often a better sound as well) until the driver/controller/software industry catches up and we begin to see affordable USB interfaces with comparable latencies.

Following to your latest message, I did another test (thank you for driving me through this stage of the new device’s exploration!). I connected my electric guitar and semi-acoustic guitar to the Hi-Zi inputs of the UR824. Then I fired up Native Instruments’ Guitar Rig 4 to test some effects.

At the reasonable sample buffer of 256 samples (@44.1), Guitar Rig’s control panel reports the following values (in ms): Input 5.9, Processing 2.9, Output 4.3, Overall 13.1.
At a more dangerous sample buffer of 32 samples, Guitar Rig reported: Input 3.7, Processing 0.7, Output 4.3, Overall 8.7.

In both cases, playing through software monitoring is absolutely perfect. I can hear no delay, and the sound never degrades (actually, it was an opportunity to hear how good my inexpensive semi-acoustic is, how time to restring my electric guitar has come, and how good Guitar Rig sounds, and how much I need to practice!)

At a certain point, I also had to fight against horrible hum. Just with a cable connected. The problem was worse with longer (despite better) cables. It suddenly disappeared, and I cannot understand how and why. This problem did not show yesterday, when I spent some quality time with my semi-acoustic. It did not show again during the tests.

Another test I hope to do soon is connecting external processing units (my RNC and my Electrix MoFx, for example, are there collecting dust, and they deserve a party). Since there are no inserts, I must rely on software routing. This will be a nice stress test for the UR.


Just installed my UR 824 and am getting the popping associated with the buffer size if I go below 192 samples. No better than my 5 year Mackie Onyx which cost 1/3 the price.

Yes - very dissapointing :frowning:

Seems linke v 1.0.4 has solved ny latency problem. I Can now have buffer sine as low as 64 samples without cracling noise. Great :slight_smile:
Tanks for the Update Steinberg !

Thanks for letting us know! :slight_smile: