[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Sun V PeeCees



Chuck Yerkes <chuck+obsd_(_at_)_2002_(_dot_)_snew_(_dot_)_com> writes:
...
> The x86 world is just a different model. One that Sun just hasn't
> been able to compete with at the non-high end.  And line will rise
> as Sun fails to update chips - the Ultra 3 is 2 years late and was
> competitive (as a chip) 2 years ago.  They likely won't hit 2GHz
> in the next year and will fall behind, as they usually do, until
> the Ultra 4.
...

Uh -- I won't argue that the us3 is late and that sun is
behind on performance.  But I think you're confusing
clock speed and performance.  The last time I made
a serious effort to compare between architectures, I believe
I found that these 3 were *roughly* equivalent:
	270 Mhz ultrasparc II
	200 Mhz powerpc 60rE
	400 Mhz AMD K6-2
Note factor of two clock rate difference between the AMD and the
ultrasparc.

More recently, I tried to compare an intel P4 at exactly 1.7 Ghz, and
an AMD athalon 2100 at about 1.7 Ghz -- despite almost the same
instruction set, compiler & clock rate, the Athalon was about 2-4 times
faster than the P4.  Supposedly intel rushed to market so couldn't iron
out some kinks in the P4.  I've been wondering if microcode updates can
improve P4 speed, or if they're going to be always just plain slower.

I haven't yet had a chance to run anything interesting on an ultrasparc
3 machine.  In *theory*, a 1 Ghz ultrasparc3 ought to be very
competitive with a 2 Ghz pentium 4, and might be competitive against a
2 Ghz athalon.  So I think the main thing hurting sun is they're only
shipping this in a few high-end machines, perhaps in every limited
quantities.

I agree sun has been in a different market than x86.  I don't think
that's going to last.  When the PC was introduced, the market was very
differentiated, with clearly distinguishable market segments including
home game machines, business micros, small minicomputers,
superminicomputers, several sorts of main frames, etc.  Some companies
managed to compete successfully in several.  The players and technical
things were different in each market -- but the formula for success was
not so different.  In any case, most of those market distinctions are
long gone, and the rest will be soon too.  We're forming interesting
new distinctions in the x86 market, but like it or not, sun is going to
have to compete head-on with x86 hardware for the desktop market.  The
x86 market has its own classical examples of NDAism, but I think it's
pretty clear that for both sun & the x86 market, being open about
design is going to be essential to success.  The *only* way the x86
market can work their way out of design corners right now is to adopt
open standards -- IBM's experiment with microchannel proved that.  Any
vendor that adopted a proprietary standard would be chopped off at the
knees.  The original SUN 386 box died for just this reason.  I think
sun realizes this too, in some dim corporate sense -- they're just
panicking and trying to have their cake and eat it too.

				-Marcus Watts