home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!ames!sun-barr!news2me.EBay.Sun.COM!exodus.Eng.Sun.COM!flayout.Eng.Sun.COM!tremblay
- From: tremblay@flayout.Eng.Sun.COM (Marc Tremblay)
- Newsgroups: comp.arch
- Subject: Re: Performance of Alpha on SPECfp92 (alvinn and ear)
- Date: 17 Nov 1992 17:48:39 GMT
- Organization: Sun Microsystems, Mt. View, Ca.
- Lines: 26
- Message-ID: <lgic3nINNg04@exodus.Eng.Sun.COM>
- References: <lg54apINNf1@exodus.Eng.Sun.COM> <1992Nov17.012550.7243@oakhill.sps.mot.com>
- NNTP-Posting-Host: flayout
-
- In article <1992Nov17.012550.7243@oakhill.sps.mot.com> dannyj@oakhill.sps.mot.com (Danny K. Jain) writes:
- >I think that the reason that Alpha has more Specmarks that HP, even if both
- >were at the same clock frequency, is because Alpha is a 2-issue machine while
- >the HP is a single issue machine.
-
- Not! HP is a 2-issue machine tuned for floating-point computations.
- It can do among other combinations:
-
- a) ld/st fp and fadd, fsub, fmpyadd, etc.
- b) ALU and fadd, fsum, fmpyadd, etc.
-
- What it does *not* do is:
-
- a) ALU and load/store integer
- b) ALU and ALU
-
- That's the main reason why integer performance in relation
- to the clock frequency is low, i.e. 80 SPECint92 at 99 MHz.
- A full dual-issue machine allowing ALU+ALU and ALU+ld/st
- with the same characteristics as the HP machine (especially
- the cache latency and sizes) would show something around
- a one-to-one ratio between SPECint92 and MHz
- (e.g. 100 SPECint92 at 100 MHz).
-
- - Marc Tremblay.
- Sun Microsystems
-