[Info-vax] Did Ken Olsen kill Alpha?

[Bob Koehler]

In article <j7s566$gtr$1 at solani.org>, Michael Kraemer <M.Kraemer at gsi.de> writes:
> 
> Mips wasn't an SGI product, but an independent fabless chip firm.

   Not yet.

> 
>> I don't think anyone was surprised when SGI bought MIPS.
> 
> Because DEC didn't want it but rather burned their money
> in a fruitless endeavour called Alpha.

   Alpha came a little later.  And porting VMS to MIPS would have been
   significantly harder than porting it to Alpha.

> The people I know of didn't give a damn about what's a "real UNIX".

   I wish I had more customers like that.

> As long as it was *IX and had good price performance.
> That's what DEC's Mips line had, if one put everything together,
> hardware, OS, apps (compilers). Istr it also beat the Sparcs of that
> era in raw performance.

   What I really enjoyed was comparing our little DECstation 3100 to
   a little HP workstation.  We had a "mips" test that contained a large
   number of "i++;" inside a large loop, assumed that each caused a 
   single instruction, and assumed that there were enough "i++" inside
   the loop that the loop overhead could be ignored.

   With no optimization, the HP ran significantly faster than the DEC.
   With -O1 or -O2 the HP retained an advantage.  But the HP compiler
   had no -O3 and the DEC compiler did.  At -O3 the DEC compiler
   realized that the value of i was never referenced after the loop,
   threw away the loop, and the resulting code generated an IEEE +Infinty 
   as the resulting processor speed.

   Compiler technology is not to be overlooked.  I'd like to get more
   +Infinity speeds.