Al Hopper wrote:
In general terms, I'd describe your proposal as an RFE for a deskside
Niagra/T1 based system. While we could debate the specifics of a desirable
hardware configuration, a range of products could be offered to target
different price points. All the product offerings could share the same T1
CPU, and be easily differentiated by the number of available memory, PCI,
PCI-X and SATA (disk drive) slots.
Seems like a good way to drive Niagra/T1 server adoption while increasing
the volume of Niagra/T1 CPUs and sharing the total Niagra/T1 NRE costs over
larger volumes.
As has been proven time after time, putting the underlying technology, in
this case the Niagra/T1 processor, into the hands of a developer, at a
reasonable cost, is an effective way to increase deployments in the
datacenter.
Since the Niagra/T1 is all about parallelism, I am diametrically opposed to
offering any form of "crippled" Niagra/T1 based product.
Well, given the nature of semiconductor manufacturing, I'm sure there
are Niagara parts with smaller numbers of working cores; these are of
course currently scrap. Their marginal cost is low, but so is their
performance.
In terms of desktop, an Ultra20 is really unbeatable in terms of
desktop perf/$$$, so perhaps this is an argument for cheaper, smaller
server configs instead. It's not clesr that most desktops have
sufficient available parallelism to make a Niagara an interesting CPU.
Thanks for an inciteful post.
The difference between incite and insight is significant.
- Bart
--
Bart Smaalders Solaris Kernel Performance
[EMAIL PROTECTED] http://blogs.sun.com/barts
_______________________________________________
opensolaris-discuss mailing list
opensolaris-discuss@opensolaris.org
