1

Topic: iOS version please...

Dear Modartt team,

I play live using MacBook Pro + RME FF400 and running HALion for Neo Soul pianos and Pianoteq 4.5. The setup sound great and is a very stable. However, I need time to connect all gears, dongles, run programs etc. I wish to use iPad for playing live. I've found that Neo Soul is available for iOS already - great news! I hope that CPU power of iPad 3/4/Air is enough for processing Pianoteq modelling. Using the keyboard (Studiologic Acuna 88) with iPad only is my dream setup.

Any plans for iOS?

Kindest regards,
Jacek

2

Re: iOS version please...

I am Android fan myself, but if you introduce PT for iPad I will get the new 12,2" iPad as it would be also suitable for PDF piano sheets. 10" version (I tried on my 10" Google Nexus) is too small for me.
It would really be dream come true :-)

3

Re: iOS version please...

Won't happen - not enough CPU.

Last edited by EvilDragon (13-01-2014 10:31)
Hard work and guts!

4

Re: iOS version please...

EvilDragon wrote:

Won't happen - not enough CPU.

Is it official from Modartt? Well, It's hard to believe that a version with limited polyphony (64 voices) similar to player cannot run on a quite fast CPUs found I the current tablets.

Regards,
Jacek

5

Re: iOS version please...

It's not official, but I'm pretty certain it won't happen. Modartt is using JUCE framework for Pianoteq which doesn't support compiling to Android nor iOS, as far as I know.


And, personally, I'd hate to see inferior versions of Pianoteq done for tablets taking up precious time that could be better used in perfecting Pianoteq's models for the platforms it's actually best for: music production desktops and laptops, rack-mounted computers for the live performer, etc.

Hard work and guts!

6

Re: iOS version please...

Here is an interesting comparison:
http://www.tomshardware.com/reviews/thi … 229-2.html

7

Re: iOS version please...

EvilDragon wrote:

It's not official, but I'm pretty certain it won't happen. Modartt is using JUCE framework for Pianoteq which doesn't support compiling to Android nor iOS, as far as I know.

According to http://www.juce.com/about-juce iOS and Android are supported by JUCE.

EvilDragon wrote:

And, personally, I'd hate to see inferior versions of Pianoteq done for tablets taking up precious time that could be better used in perfecting Pianoteq's models for the platforms it's actually best for: music production desktops and laptops, rack-mounted computers for the live performer, etc.

Well, new platform means new clients/sales/income, more money for developing new versions. I use desktop PC w/W7 home and MacBook Pro live. But I think that performing with tablet could be more convenient, easier and faster to start on the stage and don't need UPS like rack-mounted computers.

Best regards,
Jacek

8

Re: iOS version please...

JacekH wrote:

But I think that performing with tablet could be more convenient, easier and faster to start on the stage and don't need UPS like rack-mounted computers.

It's all moot if it would have to be limited in sound quality or polyphony because of limited CPU power in tablets. Not being able to reach 256 voices of polyphony is not good at all IMHO. It's best to keep using tried and tested powerful computers and that's it.

Last edited by EvilDragon (13-01-2014 15:29)
Hard work and guts!

9

Re: iOS version please...

Luc Henrion wrote:

Here is an interesting comparison:
http://www.tomshardware.com/reviews/thi … 229-2.html

That about sums things up enough, I think.


Stick to the laptops.

Last edited by EvilDragon (13-01-2014 15:30)
Hard work and guts!

10

Re: iOS version please...

JacekH wrote:

I play live using MacBook Pro + RME FF400 and running HALion for Neo Soul pianos and Pianoteq 4.5. The setup sound great and is a very stable. However, I need time to connect all gears, dongles, run programs etc. I wish to use iPad for playing live.

How does an iPad solve this problem? I'm very curious to understand what you mean.

Greg.

11

Re: iOS version please...

Luc Henrion wrote:

Here is an interesting comparison:
http://www.tomshardware.com/reviews/thi … 229-2.html

...but that are pretty old benchmarks...

A current benchmark shows the new Apple A7 processor to be roughly
comparable to an Intel Core 2 Duo E7500 or an i3-2350M:

http://www.anandtech.com/show/7335/the- … s-review/6
http://browser.primatelabs.com/geekbench3

As iOS is a supported platform in Juce and I would assume that Modartt is
using vDSP for high-performance DSP algorithms on OS X ( which is also available
on iOS),  it is not so far fetched that we might see a version in the near
future...

Tobias

12

Re: iOS version please...

M audio Oxygen 88 Works on ipad USB powerd! And CMP Piano for Ipad  10 velocity layers, no Loops, Works with 500!!! Mono voices (250 stereo Voices 5ms latenz) its like a Dream came true.... Its my live set up. There is enough CPU Power for pianoteq!!!! Iam sure!

13

Re: iOS version please...

Sample playback takes a lot less CPU than very detailed physical model that Pianoteq has.

Hard work and guts!

14

Re: iOS version please...

Well, the way i see it, you can't go wrong with having the functionality, vs. not having it, especially if it's just a matter of recompiling to a new target system. i'd simply let the user choose less cpu-hungry settings on slower tablets, no need for an "inferior" version...

http://soundcloud.com/delt01
Intel i3@3.4GHz, 4GB, Geforce GTX 560Ti • Linux (various distros) 64bit
Pianoteq 5 STD+blüthner, Renoise 3.0 • (modded) M-Audio Oxygen 88

15

Re: iOS version please...

In case anyone is wondering, my 1Ghz laptop can run at 20 polyphony. The current ipad air runs at 1.4Ghz.....which has a polyphony of around 30-40 which could pose problems for people wanting to play flight of the bumblebee on their ipads. The next generation will most likely be in the 2-2.5Ghz range and then it will be feasible to produce the software.

16

Re: iOS version please...

There is a wide variety of x86 CPUs so even if the GHz for a specific ARM matched up to a slow Bay Trail, that does not mean it'll match up to a i3 laptop.

The Anandtech article shows INTEGER benchmarks for an iPhone 5s CPU roughly matching a Bay Trail 1.5ghz.  Well that's about equiv to a 800mhz Core i3.  (See the PTQ benchmark thread @ http://www.forum-pianoteq.com/viewtopic.php?id=3202 -- there is a Bay Trail tested at 2.4ghz and 1.2ghz.)

Second, PTQ in all likelihood uses FPU instead of integer and in this department, ARM is WAY behind x86 -- the gap is more like 5X slower for FPU instructions.

Sample-based instruments don't use much CPU power.  All they need is fast streaming I/O.  Load the samples fast off disk -- stream the samples fast from memory.  You just need a ton of memory for the best libraries.  If you are using a 64GB piano, you'd better have 64GB of memory to precache everything or it's a minute to load the samples from disk before you can start playing.

17

Re: iOS version please...

Mossy wrote:

There is a wide variety of x86 CPUs so even if the GHz for a specific ARM matched up to a slow Bay Trail, that does not mean it'll match up to a i3 laptop.

Obviously

Mossy wrote:

The Anandtech article shows INTEGER benchmarks for an iPhone 5s CPU roughly matching a Bay Trail 1.5ghz.  Well that's about equiv to a 800mhz Core i3.  (See the PTQ benchmark thread @ http://www.forum-pianoteq.com/viewtopic.php?id=3202 -- there is a Bay Trail tested at 2.4ghz and 1.2ghz.)

The anandtech article you reference is using a intel 3770 which starts at 1.33-1.46GHz and turbo clocks to 1.8-2.4GHz and is pretty damn close to an i3. Also, the benchmark for pianoteq that you reference puts the 800MHz i3 "equivalent" a7 performance at around 192 polyphony in pianoteq....

Mossy wrote:

Second, PTQ in all likelihood uses FPU instead of integer and in this department, ARM is WAY behind x86 -- the gap is more like 5X slower for FPU instructions.

You could have said floating point instead of FPU (Who use's that acronym?). I don't know how old your knowledge of arm is, but ARM is not WAY behind x86 in terms of anything. ARM performance per GHz is similar to x86 at this point.

Mossy wrote:

Sample-based instruments don't use much CPU power.  All they need is fast streaming I/O.  Load the samples fast off disk -- stream the samples fast from memory.  You just need a ton of memory for the best libraries.  If you are using a 64GB piano, you'd better have 64GB of memory to precache everything or it's a minute to load the samples from disk before you can start playing.

I don't know why you're bringing up sample based instruments, everyone on these forums knows that they aren't portable friendly, or cpu intensive...

I stand by my statement that the next apple chip will be sufficient to run pianoteq for most people.

Last edited by njaremko (17-03-2014 22:16)

18

Re: iOS version please...

Here is an article that specifically focuses on ARM floating point performance.

http://www.anandtech.com/show/6971/expl … processors

You see the range of FP32 & F64 numbers are in the 1.5x range.  Then jump to the comparison against x86 processors:

For fp32, Ivy Bridge can execute up to 16 fp32 flops/cycle, Haswell can do up to 32 fp32 flops/cycle and AMD's Jaguar can perform 8 fp32 flops/cycle.  Current ARM cores can do up to 8 flops/cycle using NEON instructions. However, ARM NEON instructions are not IEEE 754 compliant, whereas SSE and AVX floating point instructions are IEEE 754 compliant.

So existing ARM CPUs -- mhz to mhz -- are anywhere from 5x slower (vs AMD Jaguar) to 20x slower (vs Intel Haswell) for FPU operations.  I am not disputing ARM has caught up to x86 in the integer department but floating point is another story.  (You can use NEON if you don't care about accuracy I guess.)

Now it's just the general trend that "world simulation" calculations -- whether 3D or physics -- usually depend on floating point processing power.  Nowadays, they're more dependent on GPU performance but we're still sometime away before a common APU/GPU/HSA API is available across platforms to run generic programs.  If Pianoteq bucks this trend by using mostly integer math, then the fastest ARM chips are good enough.  I doubt it but I haven't pulled out the profiling tools to check the exact instruction mix -- up to you to verify that.

19

Re: iOS version please...

Mossy wrote:

Here is an article that specifically focuses on ARM floating point performance.

http://www.anandtech.com/show/6971/expl … processors

You see the range of FP32 & F64 numbers are in the 1.5x range.  Then jump to the comparison against x86 processors:

For fp32, Ivy Bridge can execute up to 16 fp32 flops/cycle, Haswell can do up to 32 fp32 flops/cycle and AMD's Jaguar can perform 8 fp32 flops/cycle.  Current ARM cores can do up to 8 flops/cycle using NEON instructions. However, ARM NEON instructions are not IEEE 754 compliant, whereas SSE and AVX floating point instructions are IEEE 754 compliant.

So existing ARM CPUs -- mhz to mhz -- are anywhere from 5x slower (vs AMD Jaguar) to 20x slower (vs Intel Haswell) for FPU operations.  I am not disputing ARM has caught up to x86 in the integer department but floating point is another story.  (You can use NEON if you don't care about accuracy I guess.)

Now it's just the general trend that "world simulation" calculations -- whether 3D or physics -- usually depend on floating point processing power.  Nowadays, they're more dependent on GPU performance but we're still sometime away before a common APU/GPU/HSA API is available across platforms to run generic programs.  If Pianoteq bucks this trend by using mostly integer math, then the fastest ARM chips are good enough.  I doubt it but I haven't pulled out the profiling tools to check the exact instruction mix -- up to you to verify that.

Nothing you just discussed is relevant, the a7 is a lot faster than all of the chips in that benchmark. You're not accounting for the increase in performance and the new instruction set that came with 64bit.

Also, NEON has been been upgraded to 32 floating point 128 bit registers (IEEE 754-compliant fp32 vector instructions) in the A7.

Do some research on the a7 and you'll see it has better performance than Bay Trail. If apple doubles performance again with the A8, then it would be able to compete with haswell in terms of raw theoretical integer and floating point throughput, but that's not a guarantee because building a 1.5B transistor is nontrivial as yields are inversely proportional to the number of transistors you're using.

Last edited by njaremko (18-03-2014 21:53)

20

Re: iOS version please...

njaremko wrote:

Nothing you just discussed is relevant, the a7 is a lot faster than all of the chips in that benchmark. You're not accounting for the increase in performance and the new instruction set that came with 64bit.

Shrug  ... you should be able to post benchmarks proving this statement then.

I'm happy to change my mind if numbers are posted.

21

Re: iOS version please...

Mossy wrote:
njaremko wrote:

Nothing you just discussed is relevant, the a7 is a lot faster than all of the chips in that benchmark. You're not accounting for the increase in performance and the new instruction set that came with 64bit.

Shrug  ... you should be able to post benchmarks proving this statement then.

I'm happy to change my mind if numbers are posted.

Edit 1: Also, the next powerVR chip has hardware accelerated ray tracing....so that would probably help on the physics simulation side. (Though probably won't see it till next year)

Edit 2: It's also quite likely that the A8 will be similar to the Cortex A57 in terms of features, so you can expect IEEE-754 DP FP in NEON. The chips are apparently 4x as efficient at 25% the space....so looking forward to that smile

http://www.mondaynote.com/2013/09/22/64 … -6-months/

http://www.anandtech.com/show/7335/the- … s-review/6
Read the comments^

http://www.anandtech.com/show/7460/appl … r-review/2

Last edited by njaremko (19-03-2014 04:03)

22

Re: iOS version please...

CPU needs for pianoteq tends to increase, and it was about time since laptop'spower were increaing for quite a long time.

Pianoteq 5 will have reworked version for K1 (K2 now) D4, U4, Bluthner, Yc5, with the new algorithms for soundboard and string moddeling. The new algorithms take more CPU, but the result will be a sound closer to the natural piano sound.

Last edited by Beto-Music (19-03-2014 04:19)

23

Re: iOS version please...

Beto-Music wrote:

CPU needs for pianoteq tends to increase, and it was about time since laptop'spower were increaing for quite a long time.

Pianoteq 5 will have reworked version for K1 (K2 now) D4, U4, Bluthner, Yc5, with the new algorithm for soundboard and string moddeling. The new algorithms take more CPU, but the result wil be a sound closer to the natural piano sound.

Yep, the CPU needs will likely increase, but it will probably be marginal.  Not sure what the second part of that sentence is saying though sad

I to am totally stoked for the improved pianos and sound quality.

24

Re: iOS version please...

Pianoteq use matematical algorithms for the piano sound modeling process, and so they have algorithms for soundboard, for strings, for the harp, lid, all piano elements.

For pianoteq 5 they created improoved algorithms for soundbaord and for strings, so it will deliver a better sound, closer to natural.

But according to the vídeo some models will have the old version and the improoved models wil be called (RC7), unless I interpretated it wrong.

njaremko wrote:

Not sure what the second part of that sentence is saying though sad

Last edited by Beto-Music (19-03-2014 04:16)

25

Re: iOS version please...

No haha, I know all about algorithms tongue

Beto-Music wrote:

, and it was about time since laptop'spower were increaing for quite a long time..

I was referring to that^


And in case you're curious. The RC7 in the video is probably an acronym for release candidate 7 (though that is a pretty large number to be associated with an RC...) indicating that they're in the final phase of development.

Last edited by njaremko (19-03-2014 05:20)

26

Re: iOS version please...

I imagined RC7 was refering to models that would need more CPU power (like U4 had required).  Thanks for explain.   :-)
But would be a good idea create two versions for each model, one for médium CPU Power and other for high CPU power.


Is only me who feels that people (PC users) no longer had the desire for good CPU power advances like had 10 years ago?

Today people are almost sick with things like Facebook and other Social Nets, that no longer wait for great advances in computer power. People playing silly simple graphics games that are created to steel their Money by selling game credits, just to try some competition with Facebook "Friends".

Last edited by Beto-Music (20-03-2014 15:57)

27

Re: iOS version please...

We're past the days of huge yearly performance gains of the 90's. Now it's all about making smaller, more efficient chips that run at the same speed as last years chips.

We still see around a 10% increase in cpu performance every year. (And that will continue until quantum mechanics starts to cause problems due to the transistors being too close....still a while till then).

CPU performance isn't a huge factor in gaming, an average desktop cpu is generally enough because almost everything is handled by the (massively) more powerful GPU.

28

Re: iOS version please...

Yes, I understant the size limits of atoms related to the limits of actual processors.

But the multi core technology was going well, but somehow alsom stopped a bit in the last 2 years.

njaremko wrote:

We're past the days of huge yearly performance gains of the 90's. Now it's all about making smaller, more efficient chips that run at the same speed as last years chips.

We still see around a 10% increase in cpu performance every year. (And that will continue until quantum mechanics starts to cause problems due to the transistors being too close....still a while till then).

CPU performance isn't a huge factor in gaming, an average desktop cpu is generally enough because almost everything is handled by the (massively) more powerful GPU.

29

Re: iOS version please...

At this point (quad core) there isn't any incentive to have more, slower cores because at least 50% of software isn't thread safe, so Intel is sticking with dual/quad core and higher clocks (also dynamic overclocking) because that's the way that they're tackling the cpu problem (and in my opinion, the best way). AMD is going the more cores = more power route (hex cores and octacores), and that works for them.

30

Re: iOS version please...

Ok, the benchmarks in the comments show latest ARMs roughly matching Atom Z3770s in floating point.

The point of confusion is whether the Z3770 runs at the full turbo ~2.4ghz the entire time.  Obviously when all 4 cores are in use, it can only run at ~1.5ghz.  With a single thread without GPU, 2.4ghz should be easy.  But with Pianoteq using 2 cores?  Hard to stay -- no reviews say anything about the various turbo states.  (By comparison, the Avoton server I tested can run all 8-cores at 2.4ghz nonstop.)

So if 2 threads on a Z3770 limit turbo to say 1.8ghz -- that's probably ok performance for a mid-range config.  (I use 24 polyphony, 11hz/44hz config on a laptop roughly at that performance level.)  It's better than nothing at all of course.

For PR reasons, PTQ developers might not be comfortable releasing versions that have to be run at throttled settings on an smartphone/tablet -- you can imagine the reviews on Apple/Google app stores complaining about sound glitches if a user decides to max the config.  Tomorrow's chips will increasingly close the gap though.

Last edited by Mossy (25-03-2014 12:50)

31

Re: iOS version please...

Pianoteq was probably update years ago to deal well with multi cores.

I supose the many features and elements ( virtual soundboard, string, harp) can be set well to diferente cores, when needed.

32

Re: iOS version please...

Hi
I'm a big user of Pianoteq on windows (desktop and laptop), but since a few years, I use my iPad for live concerts or street busking. I plug in a battery powered midi keyboard and a battery powered amp, and I'm ready.
The main benefit in using the iPad is weight and battery life (between 5 and 10 hours). The minus is storage room.

What I really enjoy with Pianoteq is the very light storage space footprint. If we wanted just a quarter of it's expressivity and sound quality, it would use up several Gigabytes of precious space.

For the moment, the offering in piano sounds on iOS isn't really wide. Ik-multimedia has a collection, but they sound very "soundfont'ish" (but less than most apps out there) and takes up quite a bit of space, 
CMP grand piano has probably one of the nicest sound qualities out there, but it's around 1 Gb just for one piano sound, and it doesn't integrate very well with cubasis or other iOS DAW's

I often dream of having a pianoteq app, even if it's a cutdown version (I don't mind not having mic placement or effects other then reverb), that would give me a fantastic expressive piano sound, everywhere I go.

Until then, I'll continue busking with a lower quality piano, but continue recording my perf in midi, and then play it back on pianoteq when I get home.

Otherwise long live my favorite virtual instrument of all time

33

Re: iOS version please...

Reasons why IOS it's not a good idea :

https://www.youtube.com/watch?v=cu1-1zUWU5Q

https://www.youtube.com/watch?v=TAaU8yPXA1A

34

Re: iOS version please...

Lets not forget that Arturia used JUCE for their plugins, but still managed to get iSEM, iProphet, and iMiniV to the iOS platform.  There are several top names with active developments...customers are just waiting for more AUv3 support so we can forget about Audiobus and IAA struggles.  Im in the market for all such apps.