[SGVLUG] Linux Sonoma (Centrino) Support

[Chris Smith]

On 9/20/05, Dustin <laurence at alice.caltech.edu> wrote:
> On Mon, 19 Sep 2005, Michael Proctor-Smith wrote:
> > On 9/19/05, Dustin <laurence at alice.caltech.edu > wrote:
> > > Worse, that means the Turion chipsets are very new, and since it's a more
> > > open architecture it doesn't have to come with a single chipset as the
> > > Pentium-M generally does (centrino).  I assume that would make it harder
> > > to determine support.
> >
> > Remember that Centrino(R) is a platform not a chipset it is actually
> > the Pentium-M/chipset/wifi  chipset combination.
> 
> I did say generally--but anyway it seems that Linux support for one
> Centrino machine is about the same as the other, at least AFAIK.

Not entirely true. The Sony ones and a few others have some wackiness
in them, and a number of Centrino laptops still include their own
Broadcom based wireless, just to annoy us Linux folks.
 
> > other manufactures that work with the Pentium-M. Remember that Intel
> > has given up on the P4("net burst" architechure), and that every one
> > but Intel was right that making a less clock efficient cpu core that
> > "will scale better", does not work. The next Gen desktop/server cpu
> > will be based on the Pentium-M which in turn is our old friend
> > P3/P2/Ppro(with lots of improvements).
> 
> Yeah, that's as I understand it--sanity only reigned after AMD basically
> knocked them on the head.  AFAICT the Pentium-M is close to a PIII with a
> lot of low-power reworking (frequency scaling, L2 cache that can be shut
> off in banks) plus an improved version of the P4 branch predictor.

I think the big thing, that Intel failed to be properly prepared for,
was the total failure of the Itanium to take market share by now. The
Pentium-4 was meant to be a stop gap, geared specifically to evolve in
such a way that it'd drive up the volume of things that the Itanium
needs to be a good CPU (namely way more bandwidth, and software that
tries really hard to avoid being impacted by latency). When the P-4
was originally announced, the buzz from Intel is that it'd *never*
make it in to laptops, that the P6-core would own that space until
Itanium had enough market share to consider designing an IA64 mobile
chip.

> Apparently some people have gotten a Pentium-M to overclock to the point
> where it's faster than a P4 EE, which is amazing.  I always *hated* Intel
> chips, but I admit the PPro design really seems to have legs.  I guess
> just because the ISA sucks doesn't mean it wasn't executed well in
> silicon.

Hehe. Keep in mind that since the PPro, Intel chips haven't executed
x86 ISA instructions internally. They've always been translate, then
execute (same goes for AMD basically).

I actually have a dual-PPro machine from the "good ol' days". Those
things were fantastic for running Linux. ;-)

> > Just to let you know the Turion is a athlon 64 (using low power
> > transistors) without on-die memory controller and in a socket 754
> > package. In other words most chipsets that support single
> > hypertransport cpus with support Turion.
> 
> I didn't realize that.  I thought that was something else (like a "mobile
> Athlon 64" or something) and the Turion was a more extensive re-working.

The Turion line is supposed to be a better than AMD's original mobile
Athlon64 efforts.

> > ...Remember Athlon 64 were much
> > lower powered to begin with and AMD did not have to work as hard to
> > make them into a mobile processor.
> 
> The few bits of performance information I've found suggest that the Turion
> performance was good, though not clearly better than the Pentium-M, but it
> used more power.  Intel worked pretty hard on Pentium-M power consumption,
> my guess is that AMD hasn't done as much.  They may have started closer,
> but I suspect they still aren't as efficient.

Yeah, anyone who says the Althon64 was "much" lower powered than the
P4 hasn't been in the low-powered chip space much. Sure, it has an
advantage over the P4, but it's a gas guzzler compared to Pentium-M or
its predecessor, the PIII. More importantly Intel's Centrino chipset
does a great job managing power beyond the chip. AMD is catching up in
this space, but they are unfortunately quite a bit behind.

-- 
Chris