From owner-freebsd-hardware  Fri May 23 14:30:54 1997
Return-Path: <owner-hardware>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id OAA17707
          for hardware-outgoing; Fri, 23 May 1997 14:30:54 -0700 (PDT)
Received: from vader.cs.berkeley.edu (vader.CS.Berkeley.EDU [128.32.38.234])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id OAA17679
          for <hardware@freebsd.org>; Fri, 23 May 1997 14:30:42 -0700 (PDT)
Received: (from asami@localhost) by vader.cs.berkeley.edu (8.8.5/8.7.3) id OAA02725; Fri, 23 May 1997 14:30:41 -0700 (PDT)
Date: Fri, 23 May 1997 14:30:41 -0700 (PDT)
Message-Id: <199705232130.OAA02725@vader.cs.berkeley.edu>
To: jin@george.lbl.gov
CC: rhh@ct.picker.com, hardware@freebsd.org
In-reply-to: <199705231702.KAA29056@george.lbl.gov> (jin@george.lbl.gov)
Subject: Re: Intel Pentium II released
From: asami@vader.cs.berkeley.edu (Satoshi Asami)
Sender: owner-hardware@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

 * If you need more FPU power, then, Pentium-II is a better choice. Why not PPro?
 * Most vendors are use 440FX PCI chipset which has currency with conflict memory
 * issue. Unless you do not need memory bandwith, P6 is not a good choice. 

You are probably right about 440FX, but the only Pentium-II boards out
there (as far as I know) still use 400FX.  In which case, P6 will run
faster than a PII with the same clock rate on most applications (the
on-chip L2 does the trick).  I'd really like to see how much faster
the P6 will be for tight-loop applications (i.e., 90% of time is spent
inside loop that fits (text + data) in 256K cache).

Of course, I run my P6-200 at 233MHz, of course you can run your
PII-233 at 266MHz, but the price/performance ratio still favors the
P6.  By the way, has anyone tried to overclock the PII?

 * 									   Also,
 * Intel may not improve PCI chipset for P6 family in the further. If you tell

Intel wants to drop the P6 line (right now if they can) as far as I
can tell.  They were never able to trim the cost, as the second die
(L2 cache) had to be bonded with the main processor die before being
tested (and both had to be thrown away if one of the two didn't work).
That effectively doubled the failure rate, in other words, separating
the chip into two dies didn't help much (the failure rate wouldn't be
much different if they just made it one big die).

However, the PII (Klamath) is not a replacement for P6, it is a
replacement for P5.  Their next chip (codenamed "Deschutes") is going
to be the next-generation P6.  So I guess the P6 will have at least
another year or so of useful life. (1/2 ;)

Satoshi

P.S. By the way, nobody has sent me the results of:

  dd if=/dev/zero of=/dev/null bs=64k count=16000
  dd if=/dev/zero of=/dev/null bs=1m count=1000

  I'm interested in K6 and PII only.  The FreeBSD memory speed SIG
  (that's Bruce and I) eagerly await your input. :)