Specs for IP 10 on ARM-based macOS

Share icon

As we  hear more about folks running IP 10 on ARM-based macOS machines, I am starting a thread for posting spec comparisons.

To set a baseline, this is 21 trials with IP 9 on an 8 core i9 MBP with 32 GB 2667 MHz DDR4, 256 kB L2, and 16 MB L3.

  Create new graph  time: 	166.02ms, relative speed= 	1.04
  big data update  time: 	34.90ms, relative speed= 	1.00
  curve fit  time: 	252.03µs, relative speed= 	0.95
  user curve fit  time: 	2.86ms, relative speed= 	1.01
  double complex fft  time: 	294.03µs, relative speed= 	1.08
  single complex fft  time: 	260.89µs, relative speed= 	1.09
  double real fft  time: 	136.18µs, relative speed= 	0.97
  single real fft  time: 	137.59µs, relative speed= 	0.99
  5 pass smooth  time: 	207.55µs, relative speed= 	0.97
  Sort 8192 points  time: 	5.06ms, relative speed= 	1.00
  WaveStats  time: 	78.35µs, relative speed= 	1.03
  simple eqn  time: 	143.20µs, relative speed= 	0.99
  exp eqn  time: 	185.66µs, relative speed= 	1.00
  sqrt eqn  time: 	161.97µs, relative speed= 	0.99
  sin eqn  time: 	130.14µs, relative speed= 	1.00
  User fit fctn  time: 	102.02µs, relative speed= 	0.98
  MatrixOp eqn  time: 	35.38µs, relative speed= 	1.00
  **** done ****
  total test time=   6.31326

I am especially interested in seeing specs using the M5 chips (when they are available in the Mac-mini computers later this year).

I don't think that the benchmarking suite is particular useful with current machines. The benchmark was developed for a `Dual 2 GHz PowerPC G5 Mac Pro` according to the documentation.

March 17, 2026 at 04:48 pm - Permalink

I can test MBP M1 Pro 32GB RAM. IP10 in W11 Parallels VM, with 2 processors & 6GB RAM allocated to it. I'll get to test M5 when they deliver my new MBP M5 Pro I ordered yesterday. I am not sure if this test is really that meaningful, but why not? I am also interested in seeing what the new M5 improvement is in this case. 

**** test on Microsoft Windows 11 Pro (25H2)10.0.26200.7840 using 10.00 and 21 passes; 
**** 2 processors & 6GB RAM allocated on MBP M1, 32GB RAM. 
  Create new graph  time: 	82.80ms, relative speed= 	3.62
  big data update  time: 	94.71ms, relative speed= 	1.51
  curve fit  time: 	836.57µs, relative speed= 	2.25
  user curve fit  time: 	10.68ms, relative speed= 	3.32
  double complex fft  time: 	212.57µs, relative speed= 	6.93
  single complex fft  time: 	204.52µs, relative speed= 	5.81
  double real fft  time: 	107.64µs, relative speed= 	5.24
  single real fft  time: 	111.06µs, relative speed= 	4.43
  5 pass smooth  time: 	172.94µs, relative speed= 	3.14
  Sort 8192 points  time: 	8.37ms, relative speed= 	4.57
  WaveStats  time: 	258.97µs, relative speed= 	1.03
  simple eqn  time: 	711.66µs, relative speed= 	1.96
  exp eqn  time: 	900.95µs, relative speed= 	2.23
  sqrt eqn  time: 	856.87µs, relative speed= 	2.06
  sin eqn  time: 	675.72µs, relative speed= 	1.72
  User fit fctn  time: 	407.73µs, relative speed= 	3.17
  MatrixOp eqn  time: 	37.79µs, relative speed= 	0.63
  **** done ****
  total test time=   6.67408

March 17, 2026 at 05:27 pm - Permalink

I don't think that the benchmarking suite is particular useful with current machines.

Is there a better benchmark to spec IP 10 in emulation on ARM-based Macs? Especially in comparison to IP 10 on an Intel Windows?

I'll get to test M5 when they deliver my new MBP M5 Pro I ordered yesterday.

I will be curious to see the performance improvements. For various reasons, I'm leaning now to wait for the M5 mini. Same power (or more), lower cost.

March 17, 2026 at 06:11 pm - Permalink