[Pc_Support] Re: [LeapList] Re: AMD buys ATI, is this good or bad -- AMD has a core chipset logic already

[Bryan J. Smith]

On Tue, 2006-07-25 at 10:50 -0400, Kyle Gonzales wrote:
> If they can pull it off, and increase the quality of their Linux
> drivers, I'll buy their stuff all day long.  Ya know, this also buys
> AMD a chipset platform, as ATI got into the mobo chipset game
> recently... it was a weakness of the AMD model up to this point.

Actually, AMD already had a series of chipset designs.  But before,
AMD's models was to just "incubate" the new platform.  Once that initial
chipset was out, they left it to their partners.

E.g., the 750 (752+751) for Athlon w/SDR, 760 (762+761) for Athlon XP
w/DDR, 760 (762+766 for 6x64 at 33 or 762+768 for 3x64 at 66) for Athlon MP
w/DDR.

AMD's 8111 LPC (Legacy PC) and peripheral design, based on the proven
750/760 logic, had very mature Linux drivers.  AMD is _still_ rev'ing
those chips -- e.g., the AMD8132 adds PCI-X 2.0-266MHz support over the
original AMD8131 (and its PCI-X 1.0-133MHz support).  So they should
re-use much of that logic, just modernize the 8111 with ATI additions.

The GPU is going on the HyperTransport eXtension (HTX), so only the 8132
(legacy PCI-X) and replacement for 8111 (legacy PC / peripherals) is
needed.

ATI's SB400-series LPC and peripheral support, again, suffered the same
support issues as nVidia's MCP (Media and Communication Processor) until
late in the 2nd generation (nForce2).  ATI actually _dropped_ their
LPC/peripheral design in favor of ULi's more recently.

But nVidia just bought out ULi.  ;->

So AMD's got two-fronts ...
1.  Get the ATI GPU on the 8000 HTX (HyperTransport eXtension -- already
used for InfiniBand I/O) interconnect
2.  Merge the 8000-SB400 logic into the best set of mature peripherals
and legacy PC/PCI/PCI-X support

It's clear AMD wants to target the low-end as much as the high-end
overall.  God knows a _major_ issue with AMD platforms right now is that
Taiwanese vendors "pick'n choose" _too_many_ components.  E.g., nForce
forcedeth MAC driver is very _mature_, there are so many PHY chips used
with it, it's always a game of "wait a few kernel revs" and "load the
proprietary nvnet driver" until then.

Intel _forces_ everyone to use _their_ MAC with only an _approved_ PHY
chip.  That's why their Ethernet drivers work out-of-the-box in most
cases.  At least far less than various AMD platforms.


-- 
Bryan J. Smith          Professional, technical annoyance
mailto:b.j.smith at ieee.org    http://thebs413.blogspot.com
---------------------------------------------------------
The world is in need of solutions.  Unfortunately, people
seem to be more interested in blindly aligning themselves
with one of only two viewponts -- an "us v. them" debate
that has nothing to do with finding an actual solution.