Hardware/performance comments

Hi All,

I've recently been running some cache tests under VSTa on a range of PCs and
thought that the results might be interesting.

I've been running a compilation job, with a couple of recursive directory
listings going on in the background. To make the tests fair I've been using
the same hard drive and setting up configurations that are as close as
possible (all DX/2-66, 256k second level write back cache, Cirrus Logic 5428
video, same Seagate HDD).

In case anyone had any doubts about how hard VSTa works the CPU and memory
subsystems I recently found a fault in a 2nd level cache (lockups within 30
secs to 5mins of starting the tests) that was not evident under DOS, Windows
or OS/2. (The fault was actually verified when I looked at some signal
timings).

What I really wanted to see was how much performance was gained by second
level caching, so I set up some tests running various combinations of CPU
and 2nd level caches. The relative performance benefits on 3 types of PC
were roughly the same. The results showed that a DX/2 with its CPU cache
disabled hardly gains anything from a second level cache (about 5%). The
CPU (1st level) cache gives between 400% and 600% over none at all (about
what I'd expected). The one that surprised me was that VSTa gains 30% from
having the second level and CPU caches enabled over just the CPU cache.
This is much higher than the figures I've seen for DOS and Windows (between
15 and 20%).

I guess this 2nd level caching benefit is mainly down to the nature of the
task switching and memory management in VSTa. Has anyone seen any figures
for the same sort of thing on other OS's?

                Regards,
                Dave
Received on Fri Sep 2 05:37:19 1994