Re: SGI files for chapter 11

Malcolm Dunnett wrote:
In article <tUioAZzJ41zK@xxxxxxxxxxxxxxxxxxxxxxxx>, koehler@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx (Bob Koehler) writes:

VAX should not have been extended to 64 bits. VAX was already too
damn slow. VAX couldn't compete against 32 bit RISC. The decision
to make Alpha 64 bit was along the line of a good idea as long as DEC
knew they had to replace the VAX anyhow.


The "conventional wisdom" of the time was that CISC architectures
were too complex to be speeded up enough and that RISC was the only
way to go.

True enough, so far.


However that same thinking suggested that the 80x86 architecture could
never keep up either and was doomed.

Indeed - that was supposedly one of the rationales behind Itanic, though the prospect of becoming a monopoly may have been a stronger one.

I'm not a chip designer,

People who *are* chip designers (such as John Mashey of MIPS and SGI fame) do not share the opinion which you voice below, as has been discussed here previously.

but Intel
certainly seems to have found ways to get those speed enhancements without
sacrificing binary compatibility.

Yes, but it took them until the late '90s to do so - well over half a decade after Alpha solved the problem in a different manner for VMS.

I'm certainly not suggesting anyone
would ever do it, but I wonder if it wouldn't be possible to build a
VAX chip today ( using all the latest and greatest tricks and processes )
that would be competetive with the x86 chips of today.

People like Mashey think not. x86 is a very simple CISC compared with VAX, and the pipelining optimizations applicable to x86 and used in modern x86 implementations would not work nearly as well for the VAX ISA due to its complexity (for starters, being a 3-address rather than mostly a 1-address architecture).

- bill
.

Relevant Pages

  • Re: Why was VAX abandonned ?
    ... difference between the VAX and Alpha. ... data alignment from "Migrating an Application from OpenVMS VAX to ... Alpha, like other RISC architectures, does not provide hardware ...
    (comp.os.vms)
  • Re: Why was VAX abandoned ?
    ... You should read the Foreward section Digital Technical Journal, Alpha ... AXP Architecture and Systems, Volume 4 Number 4, page 17. ... Workstations had displaced VAX VMS from its original technical market. ... And advances in RISC ...
    (comp.os.vms)
  • Re: Code density and performance?
    ... the best engineers in the world couldn't have implemented viable VAXen in 1996... ... engineers at DEC did not try to implement a new VAX in 1996; instead, VAX development stopped pretty soon after the release of the Alpha. ... Comparing a RISC and a CISC with Similar Hardware Organization,", ACM SIGARCH CAN, 1991. ...
    (comp.arch)
  • Re: Why was VAX abandoned ?
    ... The VAX 9000, NVAX, and NVAX+ use micropipelining. ... dynamically prefeching operands prior to instruction execution. ... RISC techniques can be found in CISC chip and vise-versa! ... > fast chip - the ALPHA. ...
    (comp.os.vms)
  • Re: Intel and Microsoft provide higher 32-bit applications performance on Itanium
    ... > faster than a much older system of half the word-size, the VAX ... > must have been a dismal failure in the market. ... established as the market leader but that ran x86 software at less than half ... Itanic with relatively little native software and near-zero market ...
    (comp.os.vms)