In a message dated 11-06-94, Peter Fitzsimmons said to All:


PF>    - HPFS386 is probably slower for local use.  (This is not new
PF>      either, and I have known it all along,  but people still say they
PF>      want it included in base OS/2 because it will be faster -- it
PF>      isn't).

I can't understand why you say this.  In every benchmark I've seen, did
myself or have ever heard of HPFS386 is faster.  When I read your post
I again wanted to see what the factual answer is.  You may remember that
on the usenet conference a guy asked us to compare compiler performance
using 4 samples most of them from the TKBETA toolkit.

Here are the figures for Cset++ compiles run on a 486DX50 using a 1742
Adaptec SCSI controller.  The system has 32MB of memory and 2048K is
allocated to the cache.

           sample1A   tkclock    hanoi    graphic   Avg. Difference

HPFS 16      10         58        16        23           0%

HPFS386       9         48        13        19          17%

Times are in seconds.

In conclusion, I firmly believe that HPFS386 improves the performance
under most circumstances.  Using my system, a high volume I/O bound
systems improvements are even higher (30-35%).  However, this seems to
be a resonable benchmark and I consider 17% significant enough to
worry about.  The difference is obviously larger on longer compiles.

...Raetus
___
 * MR/2 2.1 #179 * Junk--stuff we throw away.  Stuff--junk we keep.

--- Maximus/2 2.02