Piano World Home Page Forums Our Most Popular Forums Digital Pianos - Electronic Pianos - Synths & Keyboards Does Pianoteq benefit from more cores

Or only from individual core score? These specs are always too general.

There was a conversation on the Pianoteq forum, how to install Pianoteq on a Raspberry Pi 4...
Most important things to do are enabling all cpu cores for Pianoteq and set the cpu frequency to maximum.

Pianoteq definitely benefits from more cores.

I didn't find the conversation, but I found the instructions:
https://drive.google.com/file/d/1uYW5p3675RZheZQqyLuTiGStu3xxcNAQ/view

There is meanwhile an easier way to install Pianoteq on Raspberry Pi 4 though...

Yes Pianoteq takes full advantage of additional cores, but also yes, it's not so obvious what's going on.

I'm running Pianoteq on the Odroid N2+ where I can assign the 6 cores to Pianoteq and I was able to play around with the core assignments to get a better idea of how it worked. For reference, the N2+ has 6x cores: 2x low-performance and 4x high-performance cores. Here's what I found out.

Assigning Pianoteq to a single core.
The Pianoteq performance index reflects the performance of a single core. It doesn't matter if I assign Pianoteq to 1 or all cores, I get the same Performance Index number. In the case of the N2+ if I assign Pianoteq to 1 of the low-performance cores I get a ptq index of 20, and for a high-performance core 34 (38 if overclocked). If I assign Pianoteq to all 6 cores, the ptq index will show 20...34. You may also see dual ptq indexes on computers with hyper-threading, with the virtual hyper-threaded cores showing a lower ptq index.

Assigning Pianoteq to more cores will do two things:
1. It will stabilize/peg the Pianoteq Index (blue line) to the top of the graph so it doesn't fluctuate. The blue line represents the CPU power available to Pianoteq. You want this blue line pegged to the top of the graph - and you can get that by either assigning more cores to Pianoteq or getting rid of unnecessary background processes that eat CPU cycles. If you see wide fluctuation in the blue line, this means that the estimated available CPU power / overhead is being used up by other programs. Mainly you don't want the blue PTQ index to drop below the audio load (orange graph) - causing dropouts.

2. Extra cores will also help keep the orange audio load graph in check when there are spikes in CPU usage (e.g. lots of notes/sustain/bass/dynamics etc.). When I'm only running PTQ on 1 core, the orange audio load will easily spike under normal conditions causing cpu overload/dropouts - and by spike I mean the orange graph will spike above the blue ptq index line. But when I assign Pianoteq to more cores, it is able to do a better job of keep the orange audio load graph from spiking - even if I'm forcing the dynamics, sustain polyphony etc. For example, on the N2+, using 1 high-performance core, it will easily spike at polyphony of around 32-64. With 4x cores I can easily push the polyphony up to 128++ without any dropouts/cpu overloads - the orange audio graph will easily stay low and level.

ARM little.BIG cores
On some ARM computers with little.BIG cores (low and high performance cores) - the audio load and ptq index blue line can become very fragile/brittle when the OS switches between the low and high performance cores. It results in the blue line fluctuating and the orange audio load graph quickly spiking. On those computers where this is a problem - it helps to assign just the high-performance cores to Pianoteq and leave the low-performance cores for the OS.

How many cores is the limit?
I'm not sure how many cores are the point of diminishing returns. I can tell you that on the N2+ with 6 cores, Pianoteq will utilize all the cores. On the system monitor (htop) - I can see the main Pianoteq application has a single dedicated thread on 1x core and then 1-2 child audio threads assigned to each of the other cores.

Keep in mind that the Raspberry Pi and N2+ are at the bottom of the Pianoteq CPU threshold (so it's a good place to see the results of using more cores). But even my old dual core 2015 MacBook Air has more than enough single-core power to easily handle anything I can throw at it without any issues on playback.

If only it were true that using more cores could make Pianoteq sound more like a piano!

Maybe more cores allow to press the pedal and layer more notes...

You want to play a Liszt piece perhaps? That makes sense then. For average piano beginner or intermediate level, 2 cores with freq set to moderate value is fine and lowers power consumption too.

Something doesn't add up. It doesn't seem like you could get that much improvement with a single-core without spending a fortune:

https://www.cpubenchmark.net/singleThread_value.html

I just been trying out the U4 upright.
Yamaha do not make a U4, so I assume it is an extrapolation of that which would transpire should they do, in the forseeable future.
It's really really good! Better than the Steinway, better than the Bechstein, or their latest, the Petrof 284 or whatever.
Better even, than the B1 silent I am/was thinking of buying.
It is a complex sound.
But what do I know?(shrug) me, a lowly Brit from the wrong end o' town . .:) .

With the B1 you can live in, and smoothly move between, two worlds. I can recommend a silent piano. Have been playing one for more than a year so far.

Lowly Brit? I didn't know that was possible. I thought all Brits were Great

My belief about this matter has always been that Pianoteq does ‘better’ with single core @ higher clock speeds instead of multiple cores @ lower clock speeds, but things might’ve changed since I last checked.

What I do know for certain is that no sampled piano performs better than Pianoteq under stress (lowly spec’d computer). Pianoteq, if you tweak a bit, can run on a 1980’s computer; can VSL do that? No!


Disclaimer: “1980’s computer” is used symbolically; please be advised that I cannot be held responsible if you go and actually buy a “1980’s” computer and things don’t go as planned.

What's all the fuss? Pianoteq runs on just about anything.

If it doesn't run properly (because you're using a computer from the days of George Washington), then you need a new computer. And not just because of Pianoteq.

I never saw an answer to what's the recommended specs for Garritan and Pianoteq.

Pianoteq: Any computer.
Garritan: Any computer with an SSD.

I think this is technically still true. There is a small performance cost associated with distributing threads across multiple cores. But modern CPU performance far exceeds what Pianoteq needs, so you'll never notice the difference in real life.

Intel's bottom-of-the-barrel, entry-level 4-core CPU, the Celeron J4125, runs Pianoteq fine. You can buy a Celeron J4125 computer for US$200. If you're playing advanced pieces, maybe go a little higher up the CPU ladder to an i3 or i5 (or equivalent AMD cpu). Many people are running Pianoteq on a Raspberry Pi and the J4125 is easily 2-3x faster.

Thank you.

I just noticed a 4-core Celeron J4125 running @2.7Ghz, 6GB RAM, and SSD. The ‘computer’ is tiny (about the size of a Rubik’s cube), and can be mounted behind a monitor (VESA); all this for under $200.

Would this computer suffice for Pianoteq7 running at full throttle (high polyphony, multiple mics, etc.)?

Both companies have "recommended specs" in their websites. If you don't want to believe their recommendations, why would you believe the recommendations of others?

I have already told you a couple of times that - for windows machines and my small experience - specs are meaningless because some hardware/drivers are just bad... Both PT and CFX work very well in my 2015 low spec laptop, and very bad in my 2020 mid spec laptop.

I tried a very old celeron (2 or 4 gb ram, cannot remember) and it struggled. But it was a cpu well worse than the j4125. If you can order and return the computer, it is worth a shot. My feeling is that it will be good enough for low polyphony. For pianoteq SSD is not necessary.

The Scots Welsh and Irish might be, but nobody likes the English!
Don't much care for 'em myself either . . .

I would avoid the computers with this form factor (photo below). Reviews mentioned that the heat-sink was only ok and many times didn't work effectively, so the fan was constantly going on/off and the CPU would throttle down to avoid overheating. Plus the boot disk is eMMC - similar in speed to a microSD card you might find in a digital camera - so it's much slower in terms of boot time.



I'd take a look at the J4125 computers with this form factor (below). Per reviews - the heat sink actually does it's job and it boots from an SSD which is much faster. This particular form factor is marketed under the model number GK3 or GK3V. Some factory in China probably makes the base model - but you'll find the GK3/GK3V model sold under different brand names (for example this one is branded under Terryza). It is also small enough to VESA mount on the back of a monitor.