Hi all,
Are there anyone who tried running Pianoteq on both of Linux and Windows?
I am facing some ploblem that discontinue noise come out only on Linux.
My pc is not that great pc, and I am thinking this is a performance problem.
Someone advised me to use rt patched kernel, However It was even worse than stock kernel.
Could you share me your tries?
I am using asio4all on Windows and Jack2 on archlinux with internal sound chip.
Thank you.
Using linux here, and everything works perfectly....though i've never tried the windows version.
My first guess would be the default levels for your sound chip: try muting stuff like line-in, aux, etc
Linux works great but you may need some tweaks to get good realtime performance. Check this for some tips: http://wiki.linuxaudio.org/wiki/system_configuration
Which is better Linux and Windows?
Windows according to Crumar/BGMI: ;^) http://www.pianoworld.com/forum/ubbthre...ost1420518
(alright, they are using Windows XP embedded, so it's not really a fair comparison. Also a rather old OS now)
Greg.
I have been playing around with trial 4.5 version on Linux. I'm running Fedora 19 32-bit version -- binaries ran without me needing to install any software but I already had a bunch of ALSA/MIDI libraries installed for PianoBooster.
I'm using a spare laptop -- a pretty old one from 7+ years ago. The CPU in it is a Pentium T2310 @ 1.45ghz -- roughly a Core 2 Duo without hardware virtualization -- so really it doesn't have enough power to run Pianoteq at maximum settings. In fact, the first time I started testing it, sound cutoffs and distorted sounds was 25% of the notes.
A few things I've been able to tune to get it working better under Linux:
* Do not use PulseAudio or JACK. If you don't need JACK (to connect to other MIDI software), it's better to output sound directly to the hardware devices. (On this laptop, it shows up as "HDA Intel, 7.1 Surround Sound" -- look for something similar to choose.) Otherwise, the audio servers themselves take up a lot of CPU power.
* Ignore the recommendations for RT kernels. For MIDI latency, regular kernels work fine and don't take the RT kernel performance hit.
* Minimize the window when playing -- that reduces cpu usage by roughly 5%.
* Use crontab to renice the priority if your user account does not have sudo rights to renice. Here's what I have:
* * * * * ps -ef | grep -v grep | grep Pianoteq | awk '{print $2}' | xargs renice -10 -p > /dev/null 2>&1
* Set hardware sampling rate to 48KHz, internal sampling rate to 24KHz.
* Use the "intimate/player" models with less reverb.
* Set polyphone to Auto (pessmistic)
* Set buffer size to smallest amount for lower latency.
With the above options, Pianoteq is very enjoyable to play. It makes me want to hit the ivories/ebonies all the time for practice.
* Use crontab to renice the priority if your user account does not have sudo rights to renice. Here's what I have:
* * * * * ps -ef | grep -v grep | grep Pianoteq | awk '{print $2}' | xargs renice -10 -p > /dev/null 2>&1
Instead, you can simply setup Pianoteq so it has permission to run realtime. To do this, use the following command:
setcap cap_ipc_lock,cap_sys_nice=ep "/wherever/you/installed/Pianoteq 4/Pianoteq 4"
Thanks for the hint. I'll update my scripts.
Another item I've run into -- newer CPUs take a nice with on-demand power saving. I have a wide variety of CPUs running Linux servers at work and newer CPUs are much more aggressive at throttling down. Option 1 would be to shutdown throttling whenever PTQ runs:
service cpuspeed stop
/.../Pianoteq\ 4
service cpuspeed start
Or if you prefer to leave PTQ up all the time, set the OS to scale more aggressively using:
echo 11 > /sys/devices/system/cpu/cpufreq/ondemand/up_threshold
Increasing the samples buffer also reduces CPU demand. 128-384 are good compromises -- at 512 though, I start feeling the latency.