Topic by amhosib

  • Avatar
  • amhosib
  • Member
  • Offline
  • Registered: 11-10-2014
  • Posts: 32

Topic: cpu benchmark and pianoteq performance index.

I am idly doodling using the demo mode using a hp6910, t7500, 2.2hz and benchmark 1700 (approx) and getting a PI of 36/37.
Using a i3 1.7ghz, benchmark 2500 (approx) the PI is only 30/31 all other parameters being equal (power etc)
I had the impression that a doubling of the cpu benchmark would result in a a doubling of the PI ( clock speed being equal)
Can I therefore assume that the clock speed may be a more important parameter in this relationship.
How much advantage is there to be gained in overclocking (if your bios allows, laptops tend to be bios locked)

Using either of the following:
i5 dual core 4570t 2.9ghz/benchmark 4800
i5 quad core 4570s 2.9ghz/benchmark 6800
can I expect a PI of at least 90+/130+ respectively.
What is a reasonable PI so that pianoteq runs at its maximum intended performance.

Reply by amhosib

  • Avatar
  • amhosib
  • Member
  • Offline
  • Registered: 11-10-2014
  • Posts: 32

Re: cpu benchmark and pianoteq performance index.

It appears that the relationship between benchmark and PI is not linear.
Testing the i5 quad core 4570s (6800) gave a PI of 46, and I note that a 2nd generation i7 2640m gives a PI of 48.
It might be useful to have a chart available of anticipated PI with a range of processors.

Reply by JerryKnight

  • Avatar
  • JerryKnight
  • Member
  • Offline
  • From: USA
  • Registered: 05-05-2009
  • Posts: 96

Re: cpu benchmark and pianoteq performance index.

As was stated in an earlier thread, the internal Performance Index inside Pianoteq is a very crude estimation, based on internal calculations relevant to Pianoteq. It has also been said to be context sensitive (ie. what notes are being played).

Also, there are several reports that I've found that a PI in the 20-25 range is perfectly usable, even with high polyphony and low latency. I'm not sure about the PI for my ancient Core2Duo Mac mini, but I think it's around 28, and I never hear any choking.

Since the only relationship between performance and sound quality is whether the CPU can keep up, it's just a question of how high of a sample rate, how little latency (e.g. buffer size), and how much polyphony you need. My ancient mini runs fine with the highest settings I want (44.1kHz, ~2-3 ms, max polyphony). Unless you need the computer to do lots of other things while running Pianoteq, anything above ~30 is probably overkill.

Reply by marcos daniel

  • Avatar
  • marcos daniel
  • Member
  • Offline
  • Registered: 02-09-2014
  • Posts: 280

Re: cpu benchmark and pianoteq performance index.

I use Pianoteq with a Pentium 2020M, passmark 2100, perfomance index between 25-30,  the audio load graph shows red bars but I never get a click nor pop. The program works well even if I set the buffer to 'suboptimal sizes'.

Last edited by marcos daniel (26-01-2015 19:16)

Reply by amhosib

  • Avatar
  • amhosib
  • Member
  • Offline
  • Registered: 11-10-2014
  • Posts: 32

Re: cpu benchmark and pianoteq performance index.

I had already gleaned from my doodling that a PI of 36 was perfectly OK for my needs.
I have another HP6910 with a T9500 (1850 benchpass) and 8gb ram that I wish to use solely for either pianoteq or other VST software that I have, and I dont want to spend more money on a less than necessary cpu upgrades.
The 2.6ghz cpu should squeeze a few more points onto the PI.
It works perfectly ok with Alicia's Keys and other NI software via Kontakt 5 player.
I am not sure how it would cope with Ravenscroft 275/Garriton CFX however.
I suppose its the difference between acceptable and exceptional.

Reply by Mossy

  • Avatar
  • Mossy
  • Member
  • Offline
  • Registered: 22-01-2014
  • Posts: 200

Re: cpu benchmark and pianoteq performance index.

There's a previous thread here where I made up a simple benchmarking process using PTQ's command line options and tested the 10 different computers I have access to.

https://soundcloud.com/mossy456/