I wonder if this could be generalized to other CPUs to say that for Pianoteq, a lower memory latency has a significantly higher impact on performance than raw computing speed?

Interesting hypothesis.

Just another benchmark comparison of the Odroid's Celeron J4105 with the Pentium N4200.

https://cpu.userbenchmark.com/Compare/I...8vsm444211

The J4105 is newer and slightly better in most of the benchmarks.

But the Avg. Memory Latency is on par (43.5 Pts vs. 42.9 Pts).

PTQ's performance index is 37 for both CPU in turbo mode, as mentioned a few posts above.

But nobody knows, how PTQ's performance index is calculated, so it is not more than another snapshot of your hypothesis.

PS: Could be interesting, if changing a RAM-module with a different CAS-latency has influence on Pianoteq's performance index.

Groove On wrote:

If you set Pianoteq to 48kHz / 48kHz, polyphony Auto-Optimistic, what sample rate gives you acceptable performance?

With 48 kHz sample rate / internal sample rate, auto optimistic polyphony, I hear no crackling at and above an audio buffer size of 192 samples (4.0 ms delay) on the blues demo, when pressing the sustain pedal constantly.

That is excellent news! It sounds like the Odroid H2 has the horsepower to run Pianoteq at the standard out-of-the-box default settings - all in a tiny package. What Performance Index number are you getting?

If you set Pianoteq to 48kHz / 48kHz, polyphony Auto-Optimistic, what sample rate gives you acceptable performance?

With 48 kHz sample rate / internal sample rate, auto optimistic polyphony, I hear no crackling at and above an audio buffer size of 192 samples (4.0 ms delay) on the blues demo, when pressing the sustain pedal constantly.

Per online benchmarks your i3-3217U has lower memory latency, a “smart cache” for the CPU and supports hyper-threading vs. the Celeron J4105.

You're right. According to this benchmark, the two CPUs differ greatly on memory latency and raw computing speed.

Interestingly enough, the difference in Pianoteq performance (+75%) is almost exactly the difference in memory latency score (+73%), completely dismissing the difference in single (-36%) and multi core (-95%!!!) speed.

I wonder if this could be generalized to other CPUs to say that for Pianoteq, a lower memory latency has a significantly higher impact on performance than raw computing speed?

... PianoTeq 6 is running with 192 samples (4.4ms delay), at 44kHz in 48 tone polyphony, and I never hear any crackling Question?

Thank you for posting your results. Your setup is very encouraging!

Question
If you set Pianoteq to 48kHz / 48kHz, polyphony Auto-Optimistic, what sample rate gives you acceptable performance?

... don't understand why I get a significantly higher performance index with a CPU that other benchmarks say it should be comparable (or worse) than these CPUs. My CPU is an i3-3217U ...

Per online benchmarks your i3-3217U has lower memory latency, a “smart cache” for the CPU and supports hyper-threading vs. the Celeron J4105.

But this is all good news. Sounds like Pianoteq is running decently on these two older & entry level CPUs. The J4105 is a 3-year old budget CPU, and the 3217U is a 3rd gen. mobile CPU!

ClosetMusician wrote:

If disable dynamic CPU throttling, and set to max 2.4 Ghz, I get a performance index of constant 37.

Interesting. The same index I had on an Intel N4200 based fanless laptop:

https://forum.modartt.com/viewtopic.php...19#p961319

I don't understand why I get a significantly higher performance index with a CPU that other benchmarks say it should be comparable (or worse) than these CPUs.
My CPU is an i3-3217U, running in an old Vaio Duo 11 ultrabook.

If I disable CPU throttling, the performance index is a constant 65. I never disable throttling when playing because there's no overload anyway (the performance index varies a lot in this case, but the sound remains perfect).

The case has a fan, but it runs so slow (thanks to throttling) that it is almost inaudible. Unfortunately the monitor is almost broken, so I was looking for equally silent alternatives with a touchscreen, like the Teclast X4, but these performance results for the CPU are very disappointing.

I use the low-latency linux kernel, could that be the reason for this huge difference?

If disable dynamic CPU throttling, and set to max 2.4 Ghz, I get a performance index of constant 37.

Interesting. The same index I had on an Intel N4200 based fanless laptop:

https://forum.modartt.com/viewtopic.php...19#p961319

Do you have JACK audio running for the MIDI?

No, standard was set to ALSA. It worked. I never changed it.

Interesting board.
What performance index do you get?

Interface says 19...31 when playing the demo blues song. CPU freq stays around 1.4Ghz (with max at 2.4 Ghz)

If disable dynamic CPU throttling, and set to max 2.4 Ghz, I get a performance index of constant 37.

Does your unit use the A15 or A7 cpu

These seem to be ARM's.

Odroid H2 uses Intel. Spec sheet says: Intel Celeron J4105 Processor (14nm Quad-Core, up to 2.3GHz, TDP 10W), 4MB L2 Cache, UHD Graphics 600 (Gen 9.5)

Since it's a 64bit CPU you can run any modern software on it.

Hey everyone,
I created a new PianoTeq/Midikeyboard set-up with a new mini/onboard computer "Odroid H2". I couldn't be happier. I found a lot of posts with setups including Raspberry Pis, but couldn't find a post for the more beefy Odroid H2. I'd like to share:

This will be very interesting to investigate. Thank you very much for posting this topic and information.