[PD-ot] L2 cache and Pd

[Miller Puckette]

Hi all,

We'll probably never get a good systematical understanding of all the issues,
but in general, I believe Pd is very modest in its cache needs except for
one particular situation, which is when using Gem's 'pix' operations.  For
that, I think a good rule would be that the cache should be able to hold
two images of the size and resolution needed.  So if it's 640x480, 4 color
planes, 8 bits, you want 2.45 megabytes --> get the T8x00 for sure.  If
you're going to try to work in 1024x768 and use pixel operations, make that
6.29 meg so get the T9x00. 

HOWEVER, I sure doubt any laptop will have the memory and I/O bandwidth
to support high frame-rate pix operations on 1024x768 anyway.  Also,
the more expensice CPU is likely to burn more watts, so your battery life
goes down and your machine may crash occasionally or even catch on fire.
So if I were buying, I would definitely go for the T8x00.

The (desktop) machine I'm using now has 2M, and the one I travel with has 512K.
My personal rule is to stay in the $100-$150 CPU price range which seems to be
a sort of sweet spot for desktop machines.  It's probably possible to find a
similar point-of-diminishing-returns for mobile CPUs; just hit pricewatch.com
and look for the point above which the prices start soaring much faster than
FSB and CPU clock rates.

cheers
Miller

On Tue, Sep 09, 2008 at 10:21:41PM +0900, PSPunch wrote:
> 
> Hi there,
> 
> 
> I've been watching prices of BTO laptops rising and falling for a few
> months now.
> 
> While most parts are way over spec for my usage, still CPUs vary widely
> in price only for that extra 0.1GHz in clock speed and... "L2 cache".
> 
> For general usage of Pd and Gem, how critical is the difference between
> a CPU with 3MB and 6MB of L2 cache? To be specific, I am stressed out
> trying to decide between T8x00 and T9x00 lines of the Intel CPU.
> 
> Has anyone measured performance or "cache miss" rates when running Pd &
> Gem with CPUs of such specs?
> 
> --
> David Shimamoto
> 
> _______________________________________________
> Pd-ot mailing list
> Pd-ot at iem.at
> http://lists.puredata.info/listinfo/pd-ot