[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Konec PPC za 2-3 roky?
Filip Doksansky wrote:
>
> Petr Krenzelok wrote:
>
> > Zdar,
> >
> > tak jsem zabruslil na novou stranku Indexu (Access innovations)a tam se
> > nekde Tinker zminuje o tom, ze po rozchodu IBM a Motoroly udajne IBM
> > vstoupila do jednani se spolecnosti Compaq (ten koupil Digital), aby tak
> > dojednala moznost pouziti procesoru Alpha v jejich masinach.
> >
> > Mick odhaduje, ze PPC se stane do 2-3 let kvuli male pouzivatelnosti
> > prilis drahe a ze svet CPU uz bude patrit pouze Mercedu a Alphe.
> >
> > Pokud se i tohle stane, tak je tento svet opravdu na h..., protoze i
> > Alpha se muze zahrabat vzhledem k necemu, co je AltiVec (navic pokud to
> > vyrobi ve full-copper technologii, tak to bude jeste lepsi).
> >
> > Bude podle me velka skoda, pokud takova technologie zanikne jen kvuli
> > komerce a masovemu nasazeni .....
> >
> > -pekr-
>
> To nemam strach, porad se budou pouzivat v IBM AIX serverech, ataky v AS/400
> a v MACkach. Rozhodne to nezanikne kvuli nejakymu
> shitu jako je merced, nebo alphe (ta je diky absenci multipipelingu
> odsouzena spis ke zkaze nez PPC)
>
> FiDo of INSANITY
Sorry ze je to delsi, ale objevil jsem to na c.s.a.m. a celkem to svedci
o tom, ze jaksi IBM vubec nemini vzdat vyvoj PPC (alespon jsem to tak
pochopil), ba prave naopak ...
This is a post to answer some qs ppl are having about the PPC. In
particular,
when someone says "Is the PPC dead"? The answer is usually "no way as
IBM use
it in their servers". So here's an article from an IBM AS/400 p.o.v:
IBM Firmly Committed
to PowerPC Despite
Somerset Collapse
Most AS/400 customers don't care what kind
of processor is in their little black boxes so
long as the machines run OS/400 and cost less
than last year's model. This cavalier attitude is
a luxury that comes as a byproduct of the
architecture of the System/38 and AS/400, an
architecture that enables IBM to switch
underlying processor technology with ease--at
least compared to alternative servers.
Nonetheless, as the tumultuous transition
from CISC to RISC AS/400 models has
made abundantly clear, the adoption of new
technology is not always an easy one for IBM
or its customers. Any big change in
architecture is a major cause for concern.
That's why plenty of AS/400 customers were
alarmed to learn that IBM and Motorola (the
two original PowerPC design partners who
decided to take on the Intel hegemony in PC
and server chips back in 1991) got a divorce
two weeks ago. The joint Somerset PowerPC
design center is being shuttered, and all of the
chip designers who worked in the Austin,
Texas, facility are being offered jobs at
Motorola. Does this mean that PowerPC is
kaput?
Not by a long shot. Thanks mainly to the
AS/400 PowerPC chip designers back in
Rochester, IBM will be able to deliver
powerful 64-bit PowerPC server chips for the
foreseeable future. The AS/400 team was the
only one out of many IBM and Motorola
teams that got the 64-bit implementations of
the PowerPC chip off the drawing boards and
into production. Two generations of AS/400s
have been using Cobra4, Muskie, and
Apache processors (and now RS/6000s use
the Apaches, too). The AS/400 team is also
working on the follow-on Northstar and
Pulsar Power3 chips that will be used in future
AS/400 and RS/6000 servers. Neither
Motorola nor IBM could get their 64-bit
PowerPC 620 chips to work--the 620 was
supposed to replace Cobra4 and Muskie in
the AS/400 line and allow PowerPC to
expand into the NT server and workstation
markets. As it stands, the best ideas behind
the 620 chip will be incorporated into
Northstar, although rumor has it that Motorola
is still promising to deliver the Real 620 later
this year. Don't hold your breath.
As the AS/400's use of PowerPC shows,
neither Motorola nor the Somerset partnership
mattered much to the success of the 64-bit
PowerPC processors. As for the RS/6000,
the future belongs to Apache, Northstar and
Pulsar, not the 620. IBM's 9672 CMOS
mainframes also use 32-bit PowerPC cores
and will eventually use 64-bit chips. IBM will
remain firmly committed to PowerPC since
nearly 80 percent of its $13 billion in server
revenues comes from systems that are based
on PowerPC technology. That gives IBM
plenty of profits to cover the $1 billion in
research and development necessary to bring
a new chip generation to market every two
years.
Details of the forthcoming AS/400
processors are just now beginning to trickle
out of IBM. The Northstar processor will
debut in the RS/6000 in the third quarter of
this year and will pack about twice the punch
of the current PowerPC RS-64 chips used in
the RS/6000-S70 Raven servers. The RS-64
chips are identical to the Apache chips used in
high-end AS/400e systems and
servers--although an AS/400e can do 40
percent more work than an equivalent Raven
server can because the AS/400 has more I/O
bandwidth. IBM plans to announce an
upgrade to the Raven that will have two times
the performance and twice the main memory
(up to 32 gigabytes) of the current S70. We
expect that IBM will similarly announce faster
8-way and 12-way AS/400e systems and
servers using Northstar shortly thereafter; this
server will most likely have 32 gigabytes of
main memory. Early next year, IBM will
refresh the entire AS/400e line with
Northstars.
In many ways, Northstar will be a hybrid of
the AS/400 Muskie chipset used in first
generation RISC AS/400 servers and the
Apache chip used in current AS/400e models.
Northstar will contain multiple fixed-point
execution units and fast floating point units
similar to those in Muskie; it will also use the
fast L1 and L2 cache memory controllers of
Apache. Muskie had a lot more MIPS than
the slow memory subsystem of the 530 and
53S models could ever let applications get to;
Northstar will fix that. The fastest Northstars
will likely run at 300 megahertz and will have
about twice the power of the current
125-megahertz Apache processors. (See
table for more information.)
IBM won't stop there. The follow-on Pulsar
chip, due in the second half of 1999, will run
at up to 600 megahertz and will be about
twice as powerful as Northstar-- perhaps
even more with tuning. Pulsar, IBM's fifth
64-bit processor, will beat Merced, Intel's first
64-bit processor, on every benchmark,
including Java performance.
As usual, the big question is whether or not
IBM will price AS/400s using Northstar and
Pulsar aggressively enough to take advantage
of the opportunity presented by the delay in
Merced deliveries, which have been pushed
out nine months into the second half of 2000.
History says IBM won't be aggressive enough
on pricing, but Big Blue could always surprise
us.