[Info-vax] RealWorldTech on Poulson

[Jan-Erik Soderholm]

Neil Rieck wrote 2011-07-03 13:48:
>
>>
>> Admitting that EPIC did not pan out is one thing.  But transforming an
>> EPIC architecture into an OOE one is another.
>>
> As I understand the current debacle, EPIC relied on advances in
> complier technology which never occurred. Any speed improvements we
> saw in Itanium where the result of other technological improvements
> like DDR3 and geometry shrink which would have eventually made their
> way onto Alpha platforms had the HPQ suits not interfered with
> engineering. IMHO, if Intel began nudging Itanium away from EPIC (but
> in the direction of super scaler RISC), the changes could be
> implemented in the GEM back-end code generators (it would mean
> generating Poulson-specific code but we have seen DEC compliers do
> this already in Alpha and Itanium). For example, here are some tuning
> alternatives for the DEC-C compiler:
>
> http://h71000.www7.hp.com/commercial/c/docs/5492profile_005.html#index_x_146
>
> $cc/optimize=tune=xxx
>
> where xxx can be one of:
> GENERIC
> EV4 - Selects instruction tuning for the 21064, 21064A, 21066, and
> 21068
> EV5 - Selects instruction tuning for the 21164 implementation of the
> Alpha architecture.
> EV56 - Selects instruction tuning for the 21164 chip
> PCA56 - Selects instruction tuning for the 21164PC
> EV6 - Selects instruction tuning for the first-generation 21264
> implementation of the Alpha architecture.
> EV67 - Selects instruction tuning for the second-generation 21264
> implementation of the Alpha architecture.
> ITANIUM2
>
> ###
>
> Personal comment: I think Intel and HP are currently at a critical
> point where the next few decisions will allow the Itanium program to
> "take flight and dominate the enterprise  market" or crash. As far as
> Intel is concerned, they make products for all markets so the death of
> Itanium would not take them down. As far as HP is concerned, PA-RISC
> and Alpha are gone so the death of Itanium would hurt them. EPIC never
> worked as expected but "super-scaler RISC" did so both companies would
> be wise to shift Itanium in that direction.
>
> NSR

About geometry shrinks. Not that it realy matters, but... :-)

I just checked these two pages :
http://en.wikipedia.org/wiki/DEC_Alpha
http://en.wikipedia.org/wiki/Itanium

It's interesting to compare the last Alpha (EV7z) and the
current Itanium (Tukwila).

The EV7z was released in 2004, used 180 nm and runs at 1.3 GHz.
Tukwila was released 2010, uses 65 nm and runs at 1.33-1.73 GHz.
Pulson (next year?) will use 32 nm.

EV8 was ment to use 125 nm and run at 2 GHz.

How would a EV7z/8 shrinked from 180/125 nm to 65 or 32 nm performed ?
And with 24 Mb on-die cache (as Tukwila) instead of 1.75 (3 for EV8) ?

The highest clock speed for an 180 nm Itanium (McKinley) was 1.0 GHz.

Ah well... :-)