19 Feb
2016
19 Feb
'16
4:46 p.m.
On Fri, 19 Feb 2016, Menno wrote:
Still not sure whether to update the mobo thou. I have
a ASUS that dates
from way back 2008 (!) with Intel COre 2 Duo E8400, 3 GHz.
Can someone tell me if a 4 core Intel CPU will be significantly faster when
it comes to audio processing?
faster is not really the right measure I think. I have an i5 4 core cpu
and my son has a 2 core cpu simmilar to yours (but newer). From a desktop
perspective and in terms of "snappy feel" there is really not much
difference. Mind you he uses his for mostly watch youtube stuff where GPU
makes the most difference anyway.
I do notice that compiling a program is faster on the i5, but that is
hardly a measure for audio use. Raw speed does not give a good measure of
lowlatency, it seems the more tricks the cpu uses to improve throughput,
the higher the latency as well. Latency depends on being able to meet
schedules, tricks like "Boost", "Hyperthreading" and their AMD counter
parts will affect the OS ability to meet short schedules. This is why I
chose the i5 over the i7.
On my atom based netbook, I found that turning HT off and setting my cpu
to run at 800Mhz gave better audio (no xruns) at lower latency than
letting the cpu run faster in ondemand (variable from 800 - 1600mhz). In
theory, ondemand should not make any difference, the speed switching is
very quick. However, any loop time based on cpu clock will go over time if
the cpu speed suddenly lowers during that loop. (that is the only thing I
can think of that gives xruns right when the cpu speed drops) "Boost" is
just more of the same speed change issues.
With audio use, the whole measure is if the DSP does not run out of time
in one period. That is, if jack is set to 32/2, is all the DSP finished in less
than 1.33ms. If 4 cores is better than two depends on the DSP being done.
For example in a use case where there is incoming MIDI that runs a
softsynth to create dirrect sound for stage use, all the processing is
serial and so would all have to be done on one core anyway. (overly
simplified BTW) More complex setups might very well be able to split DSP
over more cores.
Are you using Jack1 or jack2? Jack1 is single core, Jack2 tries to spread
the load over cores. An audio application may be able to spread it's own
processing over cores beyond what jack does too. At least the GUI stuff
does not need to be a part of the DSP thread.
So the answer is... it depends. ;)
--
Len Ovens
www.ovenwerks.net