On Thu, 3 Apr 1997, Chris G. Demetriou wrote:

> (2) It's slow.  see
> http://infopad.eecs.berkeley.edu/CIC/summary/local/ for a somewhat
> fair comparison of various CPUs.

The Multia is not slow due to CPU problems. My experiments with a
dhrystone benchmark show that at 233 MHz it is comparable to a P133
and at 266 MHz it is comparable to a P150. This would put the
integer CPU performance of a 166 MHz 21066 at the level of about
a P100 or thereabouts.

However, that doesn't explain why my 233 MHz Multia generally is
noticably slower than a P100. Running lmbench shows the problem:
the latency for memory access in both the L2 cache and main memory
is 2-4 times what you get from a decent Pentium system. This is a
big problem, made even bigger by the fact that the 166 MHz Multias
are woefully short on cache. You really want a megabyte or more of
cache on an Alpha system. The 233 MHz Multia has only 512K, and
the 166 MHz Multia has only 256K.

cjs

Curt Sampson    cjs@portal.ca		Info at http://www.portal.ca/
Internet Portal Services, Inc.		`And malt does more than Milton can
Vancouver, BC   (604) 257-9400		 To justify God's ways to man.'