My opinion of Altera vs. Xilinx strengths:
I think the two companies actually sell to very slightly different
marketplaces. Here I am thinking of their leading edge products,
not the small stuff that neither makes much margins on.
Altera:
Best Tools. Quite a bit better than the Xilinx tools, except when
you are trying to maximize performance.
Xilinx:
Lowest Price/Performance. This is only important when you
are going for large scale production.
So of the two, Altera is the natural choice for prototyping full
custom chips. This is a market where the customer cares a
little less about how much the chip costs, because he is going
to replace it with a cheaper full custom or gate array anyway.
It is my opinion that this market is at least half saturated. The
reason is that the parts have already achieved a quite high
market penetration. Total sales are something proportional
to the total number of engineers prototyping systems, and that
is a number that doesn't have a lot of growth in it. The only
room for growth would be in much larger parts with extremely
high prices. Something like a million gates equivalent at a
price of $5000 could be very profitable, given the ease of
the Altera tool set. This is what I would hope for if I was an
Altera stock holder.
Xilinx has the lead when it comes to parts used in mass production.
This comes from their two advantages: Lower prices and
lower level software. The Xilinx low-level software allows
a designer to squeeze about 50% more out of the already
cheaper silicon than an Altera designer. These are overwhelming
advantages for those companies that go into mass production
with Xilinx parts. Unfortunately, there really aren't that many
such companies. There just aren't too many applications for
field programmable gate arrays that can't be replaced with
cheaper full custom or gate array parts. One important
advantage of full custom parts is that they can include mixed
signal parts like converters (A/Ds and D/As), phase locked
loops, op amps, filters, etc. I read that Xilinx is working on
next generation chips that include these features. I posted
a link to a set of field programmable gate array articles for
1997 on the Xilinx web page, link:
exchange2000.com
In particular:
Roelandts also briefly discussed yet another new architecture,
planned for 1999 shipment. This architecture would make further
extensions, including D/A and A/D converters, built-in logic
analysis and a 500-MHz differential I/O scheme.
techweb.com
Given this, it is conceivable to me that Xilinx could double or
even triple their sales. Such parts would be great for all sorts
of development work as well. The last project I worked on
that used XC4000s had to have the A/D and D/A etc., off
chip.
I think the possibilities for growth of the two companies are
not as high as their stock prices. But if I were forced to accept
employee stock options at one of the two companies, I would
not hesitate before choosing Xilinx. At least part of the reason
is that I think Xilinx has more room for improvement. JMHO.
-- Carl |
