home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!decwrl!pa.dec.com!nntpd2.cxo.dec.com!nntpd.lkg.dec.com!ryn.mro4.dec.com!wrksys.enet.dec.com!bhandarkar
- From: bhandarkar@wrksys.enet.dec.com (Dileep Bhandarkar)
- Subject: Re: DEC Alpha AXP System Performance
- Message-ID: <1992Nov11.153838.29918@ryn.mro4.dec.com>
- Sender: news@ryn.mro4.dec.com (USENET News System)
- Organization: Digital Equipment Corporation
- References: <1992Nov10.153629.27510@ryn.mro4.dec.com> <BxIM38.L9F.2@cs.cmu.edu> <1992Nov11.143555.20306@kpc.com>
- Date: 11 NOV 92 10:38:55
- Lines: 19
-
-
- In article <1992Nov11.143555.20306@kpc.com>, paulk@kpc.com (Paul Kalapathy) writes...
- >In article <BxIM38.L9F.2@cs.cmu.edu> lindsay+@cs.cmu.edu (Donald Lindsay) writes:
- >>So, we now have non-simulated non-prototype-system numbers for the
- >>DEC 21064 chip and the HP 7100 chip. The result is that the HP at 99
- >>MHz has about the SPECs of a 21064 at 160 MHz.
- >
- >I suspect that the main reason for this is that HP has had considerably
- >more time for their compilers to mature on the PA. Remember that the
- >Precision Architecture has been around since the mid-80's whereas the
- >Alpha architecture is new.
- >
- The answer is not that simple. HP has a slower clock, but shorter pipeline, and
- they have a larger first level cache, but no secondary cache. It is very
- difficult to compare different machines without actually running benchmarks and
- applications. What matters is the delivered performance and cost. Don't just
- look at aggregates like SPECmarks; individual benchmark performance varies.
-
- /d
-