Pianoteq on Intel Atom 330

"Maybe a USB soundmodule with specific ASIO will be better, but I doubt it." Why? I guess it should be much better instead.

1.6 GHz isn't nearly enough for the complex computations inside Pianoteq, I presume. And I think hyperthreading doesn't help much, as it's not a "real" multicore then.

Everything works here on a 2.0 GHz Core2Duo (E4400) cpu, though, even up to 256 poly.

It's really a low-end machine, you can't expect miracles from it

The 330 is actually a dual core, the 230 is single, as I understand it. The hyperthreading is per core on the 330, showing 4 cpu's though only two physical cores are present.

Meanwhile I found out that there are no linux drivers available for my M-Audio Quattro, kind of spoiling the idea, as the flash memory I considered gets quite expensive if I am to buy the 8 GB I assume I will need for a Windows installation - plus the Windows license on top of that.

I know for sure that hyperthreading helps pianoteq when running on a pentium 4 on windows xp.

There was another thread of someone using pianoteq on an atom 270 with some success here:
http://www.forum-pianoteq.com/viewtopic.php?id=671

Comparing the performance of the 32-bit version on windows xp and the 64-bit version on linux would certainly be interesting.

Note that the initial v3.0.5 released had a performance regression which has been fixed with version v3.0.5/20090805 , so be sure to test with that version !

Here's some benchmark results for the Atom 330:
http://forum.eeeuser.com/viewtopic.php?id=43085

I ran the same benchmark program on my 2.4GHz Pentium 4, and it is slower than the 330:
Intel Atom 330 1.6GHz Dual Core – ALU: 8032 MIPS iSSE3: 6724 MFLOPS (from above link)
My own 2.4GHz P4: ALU: 4870 MIPS iSSE3: 5260 MFLOPS

I'm not sure how well these benchmarks track actual Pianoteq performance though.  Also, my processor is a single core without hyperthreading, whereas the Atom results presumably exploit the dual cores & hyperthreading.

Note that I can run Pianoteq on the P4 with a polyphony of 32 with low latency - it works fine. (the default C3 Solo Recording preset)  32 sounds like a small number these days, but in Pianoteq, it's actually very usable.

Greg.

Never got the Linux to work as I wanted, and found a reasonable offer on a 330 with XP Home SP3.

Seems I can make ptq3 work fine with 64 polyphony, but above that I can saturate the CPU.

I still need a new power supply for my M-Audio Quattro (only have a 500 mA, that is NOT enough).

When I get the power supply I will post my findings, and more details about the hardware (and software) I considered for the setup.

After running Pianteq successfully live on stage with a HP NC6000 laptop (1.6GHz Intel Centrino, 512MHz, WinXP, 2G Transcend compact flash disk, external USB/audio) I decided to try an Intel Atom N330. Also for fun but foremost to realize a more rugged version of my standalone pianoPC running pianoteq.  I bought an Intel D945GCLF2 with a 2Gbyte CF card and an external USB/audio module.
I have no problems running it glitchfree @ 41kHz with 24 polyphony and a latency under 2msec. I must say I spent a fair amount of time and effort to shrink winXP to its basics: disabling all unnecessary services and other processes that might interfere with the basic task.  Also to avoid excessive wear on the solid state CF card I needed to use some tricks like the Enhanced Write Filter from Xp embedded. This completely disables any OS writes to disk so that the system is always in excact the same condition after power up.  Lots of information can be found on the MP3Car.com forum.  They face the same problems (a reliable PC with dedicated audio tasks in a harsh environment)

After some tweaking I'm totally happy with the performance right now. Actually I'm running a full rack of additional VST instruments.  (PTQ, B4, Lounge Lizard, FM7, etc)
The only thing remaining is to construct the case and take it on stage!

I use an M-Audio Quattro.

Nope, but I'd be happy to. Will keep you posted (seems easier to test than Kees suggestions for XP optimising).

Finally got the setup together, now running PianoTeq 3.5 on the Atom 330. I am using a M-Audio Quattro external sound module, and the settings for the Quattro is set to Very Low Latency (0.8 ms) and a sample rate of 96000 Hz.

To avoid cracks I have set the internal sample rate in PianoTeq to 24000, otherwise quick runs and slides may cause cracks. Polyphony is at 64, I might be able to increase it (I just adjusted the internal sample rate down from 32000).

I have tried to see if I can tell any difference from setting the sample rate and polyphony at higher values, and through my headset (rather cheap AKG set) I am not sure. I'm usually quite observant and skilled in sound evaluation, but reading the other post about how much was percieved/imagined, and how much of the sound quality that was "real", I am so far happy with this setup.

As a conclusion I am not sure I can recommend the setup on the Atom dual core. My goal was to have a setup that I could bring around together with my midi motherboard, and found out I needed the PC, a sound module and a touch screen (occasionally the PC gets stuck in a message, I need to be able to se that and fix it :-( ). The cost and ending up with three units might be compared to e.g. a Clavia Nord Electro Rack, and playing mainly rock/blues in a rock band it might have been sufficient.

So in the end I guess I am waiting for an opportunity to get better hardware, and then have a dedicated well-equipped PC to use just for playing piano at home, and then find a solution for when we play a gig.

Still very happy with PianoTeq, though - just so much wanted the easy and portable stand-alone solution.

Reply to nickfielibert: If you have components and PianoTeq already, then it is an obvioius choice. I started from scratch, and had to weight cost and trouble against the final outcome.

Good luck, looking forward to hear about your experience.

Sounds great - looking forward to try it as soon as I get home from work.

Julien,
Where are these settings under Windows 7?

Looking at the profiles I have it currently set on "High Performance".

Best,
Simeon

Success!

I also turned off the CPU overload detection.

I have ASIO 4 All Set at 192 Samples with Allow FullMode (WaveRT) enabled.  Also set Force WDM to 16 bit.

The performance index on the Audio Load meter is reading about 9.

I think this is the closest to getting exactly what Netbook users have been looking for.

Great work.
Best
Simeon

Hello chaps,

May I ask if the Atom CPU optimisations also apply to the Mac version?

I use Mac OS X 10.5 on a Dell Mini 9, and recall being rather impressed with the performance of the Pianoteq trial in the past (v3.0?).

Will I notice any significant improvement with the latest release?

(Yes, I realise that Apple never released any Atom CPU powered systems).

Cheers,
James
x

I recently set up Pianoteq 3.6.2 on my Eeepc 1000, which has the Atom 270N processor.  I am using Eeebuntu Linux, and I can run it with no problems at a 44100 hz internal sampling rate with a buffer size of 128 and 5.8ms latency.  Here are the most important things that made this possible:

1) Use fluxbox, or a similarly low-profile window manager.

2) Turn off wifi.  This actually made a huge difference.  Turn off the actual wifi interface, such that the led is unlit, not merely disconnecting from the internet.

3) Put the CPU on Maximum Performance mode.  This is obvious, but also necessary.

I never thought I would be able to get Pianoteq to run on this netbook, but it runs perfectly with no xruns.