[Info-vax] Dvorak on Itanic

[George Cook]

In article <op.uoish9jahv4qyg at murphus.hsd1.ca.comcast.net>, "Tom Linden" <tom at kednos.company> writes:
> On Wed, 28 Jan 2009 09:10:28 -0800, George Cook <cook at wvnvms.wvnet.edu>  
> wrote:
> 
>> In article <op.uohb97imhv4qyg at murphus.hsd1.ca.comcast.net>, "Tom Linden"  
>> <tom at kednos.company> writes:
>>> On Wed, 28 Jan 2009 09:10:04 -0800, David Mathog <mathog at caltech.edu>
>>> wrote:
>>>
>>>> Richard B. Gilbert wrote:
>>>>> John Smith (not the one @ HP) wrote:
>>>>>> http://www.pcmag.com/article2/0,2817,2339629,00.asp
>>>>
>>>>> Itanic toes seem to have been "sailing under a curse".
>>>>
>>>> Well, in terms of actually getting out the door and working as  
>>>> promised,
>>>>   sure.  But in one sense it was immensely blessed.  Throughout all of
>>>> this nonsense, where all the computer manufacturers were not only
>>>> talking about deemphasizing their own CPUs, but in some cases (MIPS,
>>>> Alpha) actually doing so, and pinning future development on Intel's
>>>> great white hope, there was no anti-trust action whatsoever.  Not even  
>>>> a
>>>> hint of it.  Were not there actions more than a little anticompetitive?
>>>> To me it seems analogous to the hypothetical situation where Lexmark,
>>>> Canon, and HP, announce that henceforth all of their printers would  
>>>> only
>>>> use Epson print cartridges.
>>>>
>>>> The only company that didn't buy into this nonsense was AMD.  Now that
>>>> may have been because Intel wouldn't let them in on the party, but in
>>>> any case, they had to respond, and eventually pushed the Opteron out  
>>>> the
>>>> door, thereby showing everybody just how naked the Emperor really was.
>>>> The rest is history.
>>>>
>>>> Regards,
>>>>
>>>> David Mathog
>>>>
>>>
>>> IIRC this started as an HP poject.  One of Jack Davidson's grad students
>>> got
>>> hired by HP ca. 1992 to work on the VLIW arch, and don't forget that a
>>> couple
>>
>> Well, that explains a lot.  Having once been a grad student, I can
>> understand how one fresh out of grad school with no real world
>> experience could have come up with such an unworkable grandiose pie
>> in the sky idea based on nothing but theory.  I had always assumed
>> it was dreamed up by some extremely incompetent electronics engineers.
> 
> You made a bit of a jump there.  He was hired into a team becaquse of his
> compiler background not to design the chip himself.  This was only  
> mentioned
> as a timeline checkpoint.

It still explains a lot.  One (if not the major) flaw in the whole
EPIC scheme (aka epic disaster) is that various hard to solve problems
would be taken care of by the compilers.  If their compiler expert(s)
had never been out of the ivory towers, then it becomes understandable
how that assumption was made. 

>>> of years ago HP committed to putting in another $3B to the Intel effort,
>>> so in
>>> some ways I think Intel got suckered, although the $3B would mitigate  
>>> the
>>> pain.
>>
>> Suckered?  Maybe, but I suspect Intel also used grad students and
>> incompetent engineers who agreed that it was the "end all be all"
>> of CPU architectures.
>>
>> The part I still don't understand is how the otherwise apparently
>> very competent Alpha CPU architects bought into it.  Must have been
>> the same type of irrational group think that got us into the current
>> economic mess.  Unfortunately there are few people who are able to
>> avoid the trap of "group think", Steve Jobs being one example, which
>> is why he is so critical to Apple's continued success.
>>
> Well, in the beginning the Alpha engineers got a lot of things wrong, but
> that is because there was no serious engineering management to ride heard
> on these cowboys.

True, but they at least started with a sound concept particularly given
that various other RISC architectures had proven the whole RISC concept.
VLIW was not a proven concept; all other attempts at designing a general
purpose VLIW architecture were failures.  Intel's arrogance in announcing
a completely unproven concept as the future of computing was bad enough,
but it was simply dumbfounding when the entire industry bought into it.
My first reaction to EPIC was that Alpha no longer had anything to worry
about as far as competition from Intel, then when Alpha was killed in
favor of EPIC, I just felt sick.

After tens of billions of dollars and a decade+ of effort, all Intel
and HP have is a huge chip that barely equals or slightly exceeds (is
it still faster in FP?) other offerings including Intel's x86 CISC
archiecture.


George Cook