Topic by dhalfen

  • Avatar
  • dhalfen
  • Member
  • Offline
  • From: Baltimore, MD, USA
  • Registered: 25-05-2008
  • Posts: 469

Topic: Mac Processor Performance

Hi!  I just wish to focus on Mac hardware here -- I've seen plenty of posts for Windows systems, but they only tell me so much!

I have a Core 2 Duo 3.06 GHz iMac, running OSX 10.5.6, and the efficiency in version 3 is _significantly_ better than from 2.3 -- whereas I usually saw numbers in the mid-20% to low-30% for multicore rendering in 2.3, I'm now having to really bang fistfuls of notes to get the program into the 20% or higher range -- Bravo!!!

My main question:  any users running Pianoteq on a Mac Pro?  What performance specs do you see from the program?

(As Apple just released the Nehalem Mac Pros, I'm drooling in anticipation -- when I can afford to hemorrhage the money to get one of those suckers!!!)

"Our developers, who art in Toulouse, hallowed be thy physical-models.
Thy version 4 come, thy new instruments be done, in the computer as it is in the wood!"

Reply by BazC

  • Avatar
  • BazC
  • Member
  • Offline
  • Registered: 28-01-2008
  • Posts: 173

Re: Mac Processor Performance

I have an older Mac Pro - two dual core 2.66GHz. Normal playing pushes the processor up to around 12 - 14% If I hold down the sustain pedal and bash as many keys as possible I get it up to around 25% Oddly enabling or disabling multicore support doesn't seem to make any difference to those figures!? Maybe you have to reboot PTQ for them to take effect?

I'm not particularly impressed with the new batch of MAc Pros, they seem to have increased in speed a bit but the price has gone up dramatically at least in Europe, the previous range were much more tempting from a price/performance view IMO.

Reply by dhalfen

  • Avatar
  • dhalfen
  • Member
  • Offline
  • From: Baltimore, MD, USA
  • Registered: 25-05-2008
  • Posts: 469

Re: Mac Processor Performance

The price for Mac Pros has _always_ been ridiculous here in the States -- I know that I'm going to have to sell an arm or the like to get one -- and I always get "the best," stupidly or no...  ;^)

(I don't know what old vs. new in terms of price happens to be, but I know that I shook my head and laughed at the figures.  My old top-of-the-line Quicksilver G4 Dual 1 GHz [oh, from 2002, when those numbers were somewhat impressive] cost a little over US$3,000 [I added memory later, thus saving some money!], with every other feature being the best of available choices then...  If you factor in the two [largely useless] processor upgrades I made later, it probably cost US$4,500 or more.  Plus, it sucked energy like a black hole.)

Anyway, back on topic:  Really?  That performance sounds comparable to my own computer's -- do 8 cores really make that little difference?  I mean, if 25% is the basic _max_ ceiling which you're seeing, then my max of just over 30% with only 2 cores isn't too shabby at all!  However, even _that_ difference is kinda nice.  (And I just went "Sorabji-plus-Charlemagne-Palestine" all over my keyboard, with the pedal completely down throughout!  ;^)

I wonder if processor _speed_ plays a more significant role, making the number of cores of secondary (but useful) importance.

"Our developers, who art in Toulouse, hallowed be thy physical-models.
Thy version 4 come, thy new instruments be done, in the computer as it is in the wood!"

Reply by BazC

  • Avatar
  • BazC
  • Member
  • Offline
  • Registered: 28-01-2008
  • Posts: 173

Re: Mac Processor Performance

dhalfen wrote:

Anyway, back on topic:  Really?  That performance sounds comparable to my own computer's -- do 8 cores really make that little difference?  I mean, if 25% is the basic _max_ ceiling which you're seeing, then my max of just over 30% with only 2 cores isn't too shabby at all!

Well I only have 4 cores (2 dual core processors) but I agree I'm surprised there is so little difference, my guess is Pianoteq doesn't make very efficient use of the extra cores or else that % guide isn't very accurate. As I said earlier enabling multicore rendering makes very little difference to the performance PTQ reports.

I just tried something slightly more scientific - I played the Chopin Waltz Op 70 No1 midi file that comes with Pianoteq with multicore rendering on and off. I rebooted PTQ when I changed settings and everything else was left the same. The highest I saw CPU use go with multicore on was  15% - that was the highest it went with multicore off too! The general spread of CPU use looked about the same for both tests too and the performance index reported by PTQ was 31 for multicore on and 30 for multicore off!

Last edited by BazC (08-03-2009 14:51)

Reply by dhalfen

  • Avatar
  • dhalfen
  • Member
  • Offline
  • From: Baltimore, MD, USA
  • Registered: 25-05-2008
  • Posts: 469

Re: Mac Processor Performance

Aaaa, oops, I read your "dual-core" as "quad-core" --- brilliant of me, I know!  %^)

(Got quad-core on the brain, apparently...)

Oh, that's what I forgot to mention:  multiprocessor support makes a _big_ difference on my computer.  The numbers _easily_ reach 20% or higher, and if I go really nuts, then I approach the 30-40% range very quickly.

My "performance index" ranges from 19-22 -- I wonder how that compares.

"Our developers, who art in Toulouse, hallowed be thy physical-models.
Thy version 4 come, thy new instruments be done, in the computer as it is in the wood!"

Reply by BazC

  • Avatar
  • BazC
  • Member
  • Offline
  • Registered: 28-01-2008
  • Posts: 173

Re: Mac Processor Performance

dhalfen wrote:

Oh, that's what I forgot to mention:  multiprocessor support makes a _big_ difference on my computer.

Really!? That's a bit worrying! We're both running Intel computers with OS10.5.6 I wonder why multicore rendering has little effect on my system!?

Anyone from Moddartt have any ideas?

I'm running PTQ version 3.0.0/20090219 BTW I wonder if there's been a minor update that I missed or something?

Last edited by BazC (08-03-2009 15:05)

Reply by dhalfen

  • Avatar
  • dhalfen
  • Member
  • Offline
  • From: Baltimore, MD, USA
  • Registered: 25-05-2008
  • Posts: 469

Re: Mac Processor Performance

Just tried the Chopin on mine:

Multicore rendering:  18% max, PI of 21-23

Normal rendering:  22% max, PI of 22-23

Interesting!  He doesn't use much pedal (except in the middle lyrical section), so maybe I should create something wild in Sibelius and test it.  We'll see!

(Also, Hugh's a pretty darn good Chopin player!  He should release the first all-Pianoteq set of nocturnes or ballades!)

"Our developers, who art in Toulouse, hallowed be thy physical-models.
Thy version 4 come, thy new instruments be done, in the computer as it is in the wood!"

Reply by BazC

  • Avatar
  • BazC
  • Member
  • Offline
  • Registered: 28-01-2008
  • Posts: 173

Re: Mac Processor Performance

Hmm seems there isn't such a huge difference on your machine with the Chopin - more than mine though! It'll be interesting to see if there's any change with the 301 update.

Reply by julien

  • Avatar
  • julien
  • Admin
  • Offline
  • Registered: 19-04-2007
  • Posts: 1,000

Re: Mac Processor Performance

BazC wrote:

Really!? That's a bit worrying! We're both running Intel computers with OS10.5.6 I wonder why multicore rendering has little effect on my system!?

It makes a real difference on slower computers, but on a fast "core 2" cpu (duo or quad) , you will have to reach a very high polyphony in order to see the difference.

On my "core 1 duo" 1.83GHz, multicore rendering allows me to play midi files that would crackle otherwise. It helps also on old hyperthreaded pentium 4 cpus.

julien's Website

Reply by BazC

  • Avatar
  • BazC
  • Member
  • Offline
  • Registered: 28-01-2008
  • Posts: 173

Re: Mac Processor Performance

That's reassuring thanks Julien!

Reply by dhalfen

  • Avatar
  • dhalfen
  • Member
  • Offline
  • From: Baltimore, MD, USA
  • Registered: 25-05-2008
  • Posts: 469

Re: Mac Processor Performance

Just for the record, I'm noticing _no_ difference with 3.0.1 -- same numbers.  I _had_ to try. 

(The notes say _G5_ optimization, so I expected nothing.  %^)

Ultimately, I'd like to compose pieces along the lines of Simeon Ten Holt's multi-piano works, so using multiple instances of Pianoteq (each with highly-different perspectives and/or qualities!) is _very_ important to me --- in the future, of course! 

Therefore, I want to know what kind of "horsepower" I could get with 2 quad-core processors.  8^D

Yeah, I was a _little_ surprised with the Chopin, but I tend to go crazy with the pedal in my own stuff (I'm all about the resonance!!!), so the waltz may not be the best tester for my question. 

I just tried running a MIDI file of Liszt's "Totentanz" (the complete score, apparently) and the program reaches a _max_ (that I can see) of 46%, with the polyphony easily approaching 200 or more at the busier moments.  Version 2.3 didn't fare so well.  Wow!

"Our developers, who art in Toulouse, hallowed be thy physical-models.
Thy version 4 come, thy new instruments be done, in the computer as it is in the wood!"