[PD] [half OT] laptop processors for rt sound

derek holzer derek at x-i.net
Sun Jun 12 18:18:20 CEST 2005


hi Julien,

my (and others') problems with mobile Pentium4 processors are very 
well-documented on this list! I simply cannot recommend them for *any* 
kind of realtime audio processing, especially under Linux! This is due 
to the "denormal numbers" bug in their architecture. When the mantissa 
of a floating point number gets X number of places below the decimal 
point, it kicks into "extremely accurate mode", and there goes your DSP. 
Calculating the tail signal of a reverb (very very small, almost 
infinitely quiet sample level) is a classic situation where this occurs. 
Later generation P4s may have some fixes around this, but the problem 
still exists.

I got rid of my last P4 latop and got a Athlon64 because if this. If you 
don't mind a hot, heavy, loud computer (I could blowdry my hair in the 
morning with mine!) that lasts about an hour (in my case) then you might 
investigate this. For me, besides the fact that I put an extra kink in 
my spine every time I go on tour with this machine from carrying it 
around, the biggest issue now isn't CPU power. I can't remember the last 
time I hit bottom with this laptop. It's the noise, and all those 
desktop CPUs in a small casings will have this problem. So doing any 
recording work with microphones in the same room is completely out!

good luck,
d.






-- 
derek holzer ::: http://www.umatic.nl
---Oblique Strategy # 198:
"Instead of changing the thing,
change the world around it."




More information about the Pd-list mailing list