[PD] Pd performance relationship to CPU model

martin brinkmann mnb at martin-brinkmann.de
Fri Mar 13 11:30:40 CET 2015


On 12/03/15 16:25, Roman Haefeli wrote:

> I've been using Pd and having netpd sessions on my 8 year old Laptop
> happily ever since. Recently, I was doing a session on my much newer (1
> year old) work computer and I was hitting the CPU limit much quicker
> than I'd expected. I booted my old laptop and loaded the session there
> and the Pd load on a core was around 55%. The old laptop has Core 2 Duo
> T8300 (2.4GHz) CPU, the newer one a i5-3427U (1.8GHz) CPU.

i would suspect, that the 'dynamic overclocking' of the i5 was not
working for some reason, so that the cpu was actually running at
1.8 ghz. in my experience pd's cpu-load is more or less proportional to
the cpu-frequency, though modern cpus might be a little more efficient
at the same frequency (about 10 percent).

> A thing I'm wondering about is that a CPU benchmarking
> website [1] lists for the i5-3427U a single thread rating of 1435 and
> for the T8300 a rating of 934. So, according to this benchmark, the
> newer model is supposed to be faster, even when running single-threaded
> applications.

i believe what matters most for pd is the singlethreaded floatingpoint
performance.

> What happened to the "newer = faster" rule?

it looks like history is more or less over, concerning x86 singlethread
performance. i have just found this:

http://preshing.com/20120208/a-look-back-at-single-threaded-cpu-performance/

bis denn!
	martin



More information about the Pd-list mailing list