Subject: Re: how much does the L2 cache affect performance?
To: Zbigniew Baniewski <zb@ispid.com.pl>
From: Joseph A. Dacuma <jadacuma@ched.gov.ph>
List: netbsd-users
Date: 04/23/2007 22:48:45
Hi Mr. Baniewski!
>
> I think, we can talk about "dramatic" difference in favor of (much)
bigger L2 cache.
>
> I made a comparison between VIA-C3 700 MHz (128 KB L2 cache) and Pentium-II
> 400 Mhz (512 KB L2 cache). My experience is, that with VIA I cannot
watch AVI-files (MPlayer reports: "system is too slow to handle..."),
while with P-II/400 there is about 80-95% CPU utilization, but (mostly)
not any "framedrop". The tests were made under Linux.
>
> So, to me it is kind of "dramatic" difference, when the "much slower"
CPU (almost 2 times slower!) in practice seems to be a bit faster.
>
> Yes, I'm aware, that I cannot say "2 times slower" looking just at the
clock speed, that the direct comparison of two somewhat differently
build CPU-s cannot be precise, that (perhaps) doing integer operations,
VIA could
> be much faster... I didn't made any benchmarks, any "Eratosthenes
sieves"; just looking at it from the user's side: I'm running some
application, and I'm wondering, when it seems to be any faster. Just
because daily I'm not using any "Eratosthenes sieves" (I guess).
> --
> pozdrawiam / regards
>
> Zbigniew Baniewski
>
True, but for this particular case, this site may help you. Its in German
(I hope I understood it very well).
http://www.elektronik-kompendium.de/sites/com/0403031.htm
It clearly says about CPU speed, so 256kB L2 cache running at full speed
ist equivalent to a 512kB running at half speed. And with a higher clock
rating -- with transistors more than twice, its faster. In sum, Mr. Nelson
will be better off with the processor with 256kB L2 Cache.
Also, its not the size of the cache alone that determines (better)
performance, the overall design of a processor will make the difference
like Processor A is more efficient in handling floating-point computations
compared to Processor B and the like.
Cheers!
Joseph