[Info-vax] HP wins Oracle Itanium case

[JF Mezei]

Keith Parris wrote:

> Tukwila was in 65 nm process and the contemporary Power7 was in 45 nm 
> process. Poulson is in 32 nm process, same as Power7+ is expected to be 
> when it is released. So the situation you described is no longer the case.

Poulson was not originally going to be 32nm. But because it was
postponed for so long, HP would have been ridiculed for coming out in
2012 with a chip that wasn't at least 32.

However, just because, when they decided to use more moderns masks for
Poulson in when they decided to postpone Tukwila, Poulson and Kittson
does not mean that a delayed chip results in significantly more work
being done to improve the actual chip design.

The point of delaying the releases was to have a longer life for IA64 at
roughly the same cost to HP.

Remember that to HP, keeping IA64 on life support is to save face and
not have to tell customers that IA64 is already dead. It is a PR
exercise, so they need only palatable enough performance out of that
IA64 thing.

Meanwhile, IBM is after bragging rights and wants Power to be world
leading, and they want to catch those lucrative supercomputer deals.
Look at the clockrate on those things. (currently at 4.4GHz, rumoured to
be boosted to over 5 for Power7+). It's incentive isn't to avoid telling
customers it is cancelling Power, it is to be the best of the best.


Lets face it, IA64 is basically in maintenance mode with 2 more releases
left with contractually fixed amount of work done on each. Clock rates
may be improved through use of smaller masks, but it doesn't mean that
the architecture and chip implementations are still significantly being
improved. Adding cores doesn't really improve the processing units.

Had HP paid Intel to ramp up IA64's performance to beat Power7, HP
wouldn't have told investors they have given up on Integrity and banking
on Odyssey to rebuild BCS.




So HP's expectations on IA64 are merely to slow the loss of IA64
customers to better control the migration to the 8086. But if HP thinks
that releasing Poulson is enough to do the job, it is sadly mistaken
because to slow the rate of customer loss, HP now need to be honest with
customers and provide clear migration plans and offer sweet deals.

This is especially true of VMS customers who have been on that platform
since last century and cannot easily port overnight to a Unix system.
The time they need for a proper port will span the remaining lifetime
for IA64 so they need to start now.