home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Light ROM 4
/
Light ROM 4 - Disc 1.iso
/
text
/
maillist
/
1995
/
1095.doc
/
000164_owner-lightwav…mail.webcom.com_Tue Oct 3 14:56:52 1995.msg
< prev
next >
Wrap
Internet Message Format
|
1995-11-07
|
3KB
Received: by mail.webcom.com
(1.37.109.15/16.2) id AA143997411; Tue, 3 Oct 1995 14:56:52 -0700
Return-Path: <owner-lightwave@mail.webcom.com>
Received: from mrpitt.dti.com by mail.webcom.com with SMTP
(1.37.109.15/16.2) id AA143067365; Tue, 3 Oct 1995 14:56:05 -0700
Received: from assman (192.84.116.12) by mrpitt.dti.com id <A19951003211757178@mrpitt.dti.com>; Tue, 03 Oct 1995 16:17:57 -0500
Comments: Authenticated sender is <blaise@mrpitt.dti.com>
From: "Blaise Fanning" <Blaise Fanning@mrpitt.dti.com>
Organization: Deskstation Technology
To: lightwave@mail.webcom.com
Date: Tue, 3 Oct 1995 16:50:15 +0000
Subject: Re: News About Raptor
Reply-To: blaise@mrpitt.dti.com
X-Confirm-Reading-To: blaise@kramer.dti.com
X-Pmrqc: 1
Priority: normal
X-Mailer: Pegasus Mail for Windows (v2.01)
Sender: owner-lightwave@mail.webcom.com
Precedence: bulk
> Date: Tue, 3 Oct 1995 14:32:48 -0700
> From: 03-Oct-1995 1705 <leimberger@marbls.ENET.dec.com>
> To: "cevanim@aol.com"@24580.ENET.dec.com
> Cc: lightwave@marbls.ENET.dec.com
> Subject: Re: News About Raptor
>
> >I've also been told by a supplier at a show that the 21164 would have a data
> >path of 256. Not being technical I didn't have the right questions to ask
> >him. Why are they telling us that at the shows, with the new chip coming out
> >with only 128 it causes us to lose confindence in them?
>
> Well I could probably find out, but won't bother. However I do know that
> Data Paths, and pincount don't always jive. It is very possible a device
> could have a 256 data path internal and a 128 bus external.
>
> bill
Since I've taken on the mantle of the guardian of truth, I'd like to
comment on today's verdict...Doh!
The data path inside the chip between the 96kB SCache and the
processor core is also 128 bit. From page 2-13 of the "Alpha 21164
Microprocessor Hardware Reference Manual":
The second-level cache (Scache) is a 96K-byte, 3-way,
set-associative,physical,write-back,write-allocate cache with
32- or 64-byte blocks. It is a mixed data and instruction
cache. The Scache is fully pipelined; it processes read and
write operations at the rate of one INT16 per CPU cycle and can
alternate between read and write accesses without bubble cycles.
The part about one "INT16" per CPU cycle means that the SCache deals
with one 16 byte quantity per cycle. 16 bytes * 8 bits/byte = 128
bits. Therefore, the significant data path inside the chip is not
256 bits, but 128.
I'll try to stop rambling on, but probably not while people are still
spreading misconceptions about things. (Not that Bill was, but I
thought I'd save him the trouble of finding out...)
Blaise Fanning
Chief Technology Officer
Deskstation Technology
(913) 599-1900
blaise@dti.com
--
"Blaise Fanning" <Blaise Fanning@mrpitt.dti.com> sent this message.
To Post a Message : lightwave@webcom.com
Un/Subscription Requests To : lightwave-request@webcom.com
(DIGEST) or : lightwave-digest-request@webcom.com
Administrative Items To : owner-lightwave@webcom.com