Topic: High Latency problem with new Windows 10 machine

Good evening.

I am very impressed with the trial demo that I ran on my old Windows7 notebook from ASUS, but that processor was so slow, that I kept overloading the CPU if I tried to play more than 4 or 5 notes at a time.  Today I purchased a new Microsoft Surface4 Tablet which gives a Performance Index in Pianoteq v5 of 14, but the latency is about 1/4 second.  I have tried running it via Windows Audio, Windows Audio (Direct), ASIO4ALL, and also running VirtualMIDISynth.  I have reduced my buffer to 256 and to even less.  I have tried both Bluetooth audio (where I first noticed the problem) and via the laptop's internal speaker.  I have also tried the demo program on my Dad's Sony Vaio, also running Windows 10.

Nothing reduces my latency below about 1/4 or 1/5 second.  I found ways to make it almost one second long, but nothing to reduce it to a 'playable' latency.

What can I do to reduce my latency in Windows 10 on this machine?  I would like to take advantage of the Black Friday 30% discount on PianoteqPro, but can't see spending anything if I can't reduce the latency.

Thanks,
David

- David

Re: High Latency problem with new Windows 10 machine

The tablet PC might be too slow for Pianoteq. I found latency on Intel Core 2 Duo machines to be borderline usable but performance still lacking. Anything from i3 up should be okay. An i5 quad core machine works great. Contact Pianoteq/Modartt support. Their customer service is excellent IME.

3/2 = 5

Re: High Latency problem with new Windows 10 machine

dklein wrote:

Good evening.

I am very impressed with the trial demo that I ran on my old Windows7 notebook from ASUS, but that processor was so slow, that I kept overloading the CPU if I tried to play more than 4 or 5 notes at a time.  Today I purchased a new Microsoft Surface4 Tablet which gives a Performance Index in Pianoteq v5 of 14, but the latency is about 1/4 second.  I have tried running it via Windows Audio, Windows Audio (Direct), ASIO4ALL, and also running VirtualMIDISynth.  I have reduced my buffer to 256 and to even less.  I have tried both Bluetooth audio (where I first noticed the problem) and via the laptop's internal speaker.  I have also tried the demo program on my Dad's Sony Vaio, also running Windows 10.

Nothing reduces my latency below about 1/4 or 1/5 second.  I found ways to make it almost one second long, but nothing to reduce it to a 'playable' latency.

What can I do to reduce my latency in Windows 10 on this machine?  I would like to take advantage of the Black Friday 30% discount on PianoteqPro, but can't see spending anything if I can't reduce the latency.

Thanks,
David

Try to update your audio drivers. I use Pianoteq with my old Core 2 Duo and new Windows 10. There is no problems with latency - very playable. I don't even use ASIO, just Win Audio. Try to increase sample rate to maximum through both audio device settings (windows) and Pianoteq settings.

My PT settings, that are fine:
Dev type: Win Audio
Host sample rate: 192000 Hz
Buffer size: 1920
Internal Sample rate: 48000 Hz
Max polyphony: 48

Windows audio sample rate (audiocard settings): 192000 Hz

Latency: is shown as 10ms (but, I guess, it is even smaller)

I also used Pianoteq with acer tablet. But in this case I decrease max polyphony

Last edited by Kridlatec (29-11-2015 10:09)
Pianoteq 6 Pro (D4, K2, Blüthner, Model B, Grotrian, Ant.Petrof)
Studiologic SL88Grand, Steinberg UR22mkII

Re: High Latency problem with new Windows 10 machine

Kridlatec wrote:

Latency: is shown as 10ms (but, I guess, it is even smaller)

Nope, it's higher. Driver reported latency is just one small part of the whole "total round-trip latency" equation.

https://www.presonus.com/community/Lear...io-Latency


I see no reason to run your stuff at 192k... waste of bandwidth and CPU time - and humans are not bats. You could probably get lower latency if you used 44.1/48k and, say, buffer size of 128 samples.

Last edited by EvilDragon (29-11-2015 11:52)
Hard work and guts!

Re: High Latency problem with new Windows 10 machine

EvilDragon wrote:
Kridlatec wrote:

Latency: is shown as 10ms (but, I guess, it is even smaller)

Nope, it's higher. Driver reported latency is just one small part of the whole "total round-trip latency" equation.

https://www.presonus.com/community/Lear...io-Latency


I see no reason to run your stuff at 192k... waste of bandwidth and CPU time - and humans are not bats. You could probably get lower latency if you used 44.1/48k and, say, buffer size of 128 samples.

In my case higher rates
1. give lower latency (with the same buffer size)
2. produce richer sound
And don't tell me about bats and "that is impossible", I've heard it many times. Fact is a fact: 192KHz sounds for me noticeably better. May be it is due to drivers or my audio device, but I have this effect

Last edited by Kridlatec (29-11-2015 17:53)
Pianoteq 6 Pro (D4, K2, Blüthner, Model B, Grotrian, Ant.Petrof)
Studiologic SL88Grand, Steinberg UR22mkII

Re: High Latency problem with new Windows 10 machine

Kridlatec wrote:
EvilDragon wrote:
Kridlatec wrote:

Latency: is shown as 10ms (but, I guess, it is even smaller)

Nope, it's higher. Driver reported latency is just one small part of the whole "total round-trip latency" equation.

https://www.presonus.com/community/Lear...io-Latency


I see no reason to run your stuff at 192k... waste of bandwidth and CPU time - and humans are not bats. You could probably get lower latency if you used 44.1/48k and, say, buffer size of 128 samples.

In my case higher rates
1. give lower latency (with the same buffer size)
2. produce richer sound
And don't tell me about bats and "that is impossible", I've heard it many times. Fact is a fact: 192KHz sounds for me noticeably better. May be it is due to drivers or my audio device, but I have this effect

As I see it with 192kHz (or rather hear), sure, it makes it possible to produce frequencies up to 96kHz, which is far beyond what we can hear. However,
the upper range that we do hear is much more detailed because there are, of course, more samples, more points, to shape the waveform rather than only having enough samples to have one positive point and one negative point in alternation for the waveform. Whether your transducers are high enough quality to produce this effect is another question.

Windows 10 1511 | Intel Core i5 4690K 3.5GHz | 16GB RAM
Focusrite Scarlett 2i2 | JBL LSR305
Casio Privia PX-150 | M-Audio Oxygen 49 (Gen 3)
Pianoteq Standard 5.5.1 | D4, K2, U4, KIViR, Bells

Re: High Latency problem with new Windows 10 machine

Kridlatec wrote:

In my case higher rates
1. give lower latency (with the same buffer size)
2. produce richer sound
And don't tell me about bats and "that is impossible", I've heard it many times. Fact is a fact: 192KHz sounds for me noticeably better. May be it is due to drivers or my audio device, but I have this effect

That's not a fact, that's a subjective experience/opinion of yours. I hear no difference between 48k and 192k personally, so I don't waste CPU power on it. Also it's been mentioned by Pianoteq developers that pianos don't have a lot (if at all) spectral energy above 15k (or even less than that), which IMHO completely removes the need for processing Pianoteq at 192k.

Last edited by EvilDragon (29-11-2015 22:01)
Hard work and guts!

Re: High Latency problem with new Windows 10 machine

EvilDragon wrote:

Also it's been mentioned by Pianoteq developers that pianos don't have a lot (if at all) spectral energy above 15k (or even less than that)

This is a spectogram for the Blues Demo generated by Pianoteq standard trial at 44100Hz (wav export with high quality settings) that seems to confirm that.
I would like to do a blind test to check if I can hear differences between Pianoteq generated files at 192KHz and 44.1KHz, can someone supply me those files?

http://www.pictureshack.us/images/46207_pianoteq_wav.png

Re: High Latency problem with new Windows 10 machine

The internal sampling rate of Pianoteq maxes out at 48kHz in any case, as far as I can gather from the UI (someone please correct me if I'm wrong). In that case I'd assume that Pianoteq engineers being the clever folk they are would make sure that the signal generated is properly bandwidth limited to the Nyquist limit of 24kHz. Then there is no point going beyond 48kHz sampling rate for output as far as sound quality goes.

The main benefits of oversampling beyond the hearing range are getting rid of aliasing artefacts (when the signal being sampled contains frequencies above the audible limit, i.e., the signal is not properly bandwidth limited), or as a brute-force workaround for less than optimal digital-to-analog conversion hardware (rarely an issue these days).

For compatibility with other hardware/software that uses higher sampling rates it may be useful to have the option of Pianoteq producing a matching output. But as far as quality of the sound goes it should not make any difference.

https://www.youtube.com/embed/cIQ9IXSUzuM

Last edited by SteveLy (30-11-2015 01:39)
3/2 = 5

Re: High Latency problem with new Windows 10 machine

EvilDragon wrote:

I hear no difference between 48k and 192k personally, so I don't waste CPU power on it.

That is enough: you don't hear, I hear. My own subjective experince is much  important for me, then someones else, when I am training at home with headphones You may also play, as you want. I am not claiming that I hear the difference between 48 and 192 kHz. I claim, that on my equipment and soft, when I switch between this modes - there is difference. Why this happens? Drivers, hardware, anatomy, psychology? I don't know and frankly - not interested.

Last edited by Kridlatec (30-11-2015 09:49)
Pianoteq 6 Pro (D4, K2, Blüthner, Model B, Grotrian, Ant.Petrof)
Studiologic SL88Grand, Steinberg UR22mkII

Re: High Latency problem with new Windows 10 machine

Kridlatec wrote:
EvilDragon wrote:

I hear no difference between 48k and 192k personally, so I don't waste CPU power on it.

That is enough: you don't hear, I hear. My own subjective experince is much  important for me, then someones else, when I am training at home with headphones You may also play, as you want. I am not claiming that I hear the difference between 48 and 192 kHz. I claim, that on my equipment and soft, when I switch between this modes - there is difference. Why this happens? Drivers, hardware, anatomy, psychology? I don't know and frankly - not interested.

You may not be interested but others might be. There are valid reasons for Pianoteq to offer such high sampling rates (other than marketing gimmickry). If your setup sounds better at 192kHz then more power to you. There often are good reasons to oversample, and the maths and engineering behind it all is a bit involved (and more subtle than the way most lay or even some professional folk understand the problem). But in practice all you need to do is try out what works best with your equipment, as you @Kridlatec and @EvilDragon have both done.

PS. Sorry @OP @dklein for going off the rails, i.e., the topic of this thread: getting Pianoteq working on a Windows 10 machine.

Last edited by SteveLy (30-11-2015 10:59)
3/2 = 5

Re: High Latency problem with new Windows 10 machine

Thanks, EvilDragon, for recognizing that things are off-topic.  I'm the one stewing frustrated with a currently unplayable system, despite ASIO4ALL on a new Windows 10 core i5 machine - and, after a bit of hopeful suggestions that didn't help, I would get my expectations elevated by emails saying that this thread had more responses – only to have them dashed by reading theoretical and actual arguments of whether humans can tell anything about the presence of 192 kHz harmonic tones…

In any case, while the Pianoteq on-screen buffer size prediction tells me that I have a 4 ms latency, my ears tell me something entirely different (especially when I have local sound production on my keyboard turned on, so I get to hear the keyboard first, and then the computer speaker as a notably late echo, likely about 1/5 second.  I do not think that it is either the keyboard or the MIDI cables, as when these two were plugged into an old windows 7 notebook (whose CPU can't play more than four or five notes polyphony before it bogs down), my latency was perceptibly zero.  I have updated all of my windows 10 components, as well as making sure that I am running updated sound card drivers (for the onboard Realtek sound circuit).  My MIDI cable, a Manhattan 179171 standard MIDI to USB cable, does not have any drivers available outside of the standard windows driver – there is nothing on their website as they explain that only generic drivers are useful.  The keyboard is a Casio WK-3000, an older keyboard that hooks up through the MIDI cables only, and does not have a USB connection to the computer (other than through the MIDI to USB conversion).  And, as I noted above, it ran with nearly zero latency on a windows 7 machine.

Anybody else have any useful suggestions for me?

- David

Re: High Latency problem with new Windows 10 machine

First of all, ASIO4ALL is a kludge - but likely your only choice if you're using only onboard audio chip. But I would definitely recommend to invest in a proper audio interface that has its own low latency ASIO drivers (and its own MIDI ports, so you don't have to use a MIDI-USB cable, which is also a kludge of sorts...).

You could try to use Windows Audio (WASAPI) for Pianoteq's audio output instead of ASIO, perhaps. See if that helps. I can get reasonably low latencies with it on my now 4 years old laptop (also an i5)... But in any case, for real low latencies, investing in a good audio interface is the way to go.

Also, "near zero latency" is impossible (except in case of direct monitoring, but this doesn't work for virtual instruments). Refer to the Presonus article above. Even if the driver reports 1 ms latency, there's still at least 10 ms of additional latency that goes into DA conversion, USB buffers, etc.

Last edited by EvilDragon (30-11-2015 11:37)
Hard work and guts!

Re: High Latency problem with new Windows 10 machine

It sounds like the problem is specific to your machine: something about the way it's set up. ASIO4ALL on an i5 Windows box should work a treat normally but obviously something is amiss. Make sure you have no other application hogging the sound card. Then, if problem persists, at that point I'd contact Pianoteq/Modartt support directly. They are very helpful IME.

Last edited by SteveLy (30-11-2015 13:21)
3/2 = 5

Re: High Latency problem with new Windows 10 machine

Thanks again, evildragon, for your most recent reply.  I am starting to think that you are correct about my is one a dedicated MIDI processor – if nothing else, this will help a lot when it comes to plugging in some monitor speakers rather than just trying to convert from the mini plug headphone out on the tablet.

Well, as I better shower and get off to work, I will report when I did this morning: I went back to using the windows drivers rather than ASIO.  Curiously enough, they were faster.  And, for reasons I don't understand, using the windows driver rather than windows driver in "exclusive mode" was faster.  Still, even with my buffer turned way down (the lowest possible for each), there is still a perceptible latency, although things are playable unless I try to hit sixteenth notes – then my ear and my fingers get out of sync.  Still, for slow things, it's okay.  Also, another thing that I noticed had to do with the speed of the calculations – at 44 kHz, my relative latency was increased, and with a low buffer, there was a lot of clicks, pops, and trash.  When I changed it to 48 kHz, the latency decreased to its current lower level, and the trash went away, even at the lowest buffer setting.  Perhaps this is because the processing speed is running at approximately 24 kHz, half of forty-eight rather than an odd fraction of forty-four.

At least things are improving, though slowly…

- David

Re: High Latency problem with new Windows 10 machine

I wouldn't say ASIO4all is a kludge, but yes, sometimes it works really fine, sometimes not. At all! I happen to use PianoTeq on a i3 laptop with ASIO4all and get very low latencies, whille on another, better i5 laptop, I just get problems. Go figure.

Re: High Latency problem with new Windows 10 machine

dklein wrote:

Perhaps this is because the processing speed is running at approximately 24 kHz, half of forty-eight rather than an odd fraction of forty-four.

Wrong. When sample rate is 44100 Hz, then processing is running at 44100 times in a second - not half. Half is the maximum frequency that can be reproduced using that sample rate without aliasing artifacts (see Nyquist theorem).

But yes - higher sample rates automatically mean shorter buffers. It's simple math. However the end result is the same: having a lower sample rate with lower buffer size is more or less the same strain on the CPU as having a higher sample rate with a larger buffer. They still need to process the same amount of samples within the same time period. Example: a buffer of 256 samples at 44100 Hz will have the same number of samples to process as a buffer of 512 samples at 88200 Hz. Simples.

There's also one thing called "native sample rate" at which your audio interface is working. For Realtek I think this is 48k. This means that you get somewhat better performance when using THAT sample rate (and its multiples) instead of others, because its DA converters are finetuned for working best at that sample rate.

Luc Henrion wrote:

I wouldn't say ASIO4all is a kludge, but yes, sometimes it works really fine, sometimes not. At all! I happen to use PianoTeq on a i3 laptop with ASIO4all and get very low latencies, whille on another, better i5 laptop, I just get problems. Go figure.

It really depends on the audio chipset variant used, and other factors. What A4A is is basically a wrapper that wraps WDM driver for use with ASIO-enabled programs. We can compare this to bit-bridging programs from 32-bit to 64-bit, in a way.

Last edited by EvilDragon (30-11-2015 15:17)
Hard work and guts!