Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: ONLY one ref multiplier?: PMA clock reference


As a (soon-to-be) chip supplier, I'd like to try my had at this one.

> Stuart Brorson wrote:
> I would like to hear from the chip people on this one, but my conjecture is
> that the cost of OC-192 chips arises from several interrelated factors:
> 1.  10 Gbps is very fast, the chips represent the cutting edge of IC
> technology, and are hard to design correctly.

Absolutely. However, much simpler physical layers based on Ethernet rather than
SONET and multi-lane signaling approaches such as WDM and parallel optics
significantly reduce for the design complexity and technology requirements of 10
Gbps data transport to the point of being, lets say, only 2-3 times as complex
as GbE chips. Note also that cutting edge is dulled to the point that standard
CMOS is well within reason. A recent announcement from one company says that
they will be sampling 10 Gbps serial output devices later this year! That's
cutting edge!

> 2.  There is only a handful of chip houses which can currently produce the
> chips.  There is consequently less competition in the marketplace.

There are now ~5X more companies claiming to be able to produce 10 Gbps Physical
layer chips then there were at 1 Gbps. In addition, all the companies which were
highly successful at 1 Gbps are in that group. There is no shortage of
competition in this area. Consequently, costs will be reasonable. They already
are if you look at some of the recent announcements which also list pricing.
I'll stay IEEE politically correct and not tell you who they are. Give me a
private call and I'll give you the details.
> 3.  The vendors have just recently started producing the chips; they have not
> yet had the time to slide down the cost curve (e.g. as their processes mature)

This doesn't much matter in the networking world. It may on the Telco side.
Everyone knows and believes that unit volumes will grow astronomically if
history and current demand is any indicator.

> 4.  The market for OC-192 chips is relatively new, and there are only a
> handful of companies buying the chips in volume right now.  Economies of scale
> have not yet driven the costs down.

That's because the OC-192 chips themselves are priced so high in the first
place. Chips with the equivalent functionality for 10 GbE are already being
priced at a level approximately one order of magnitude less, and the pricing
still looks like the chip vendors are going to make a tidy profit.
> Note that none of these factors have to do with SONET itself.  I suspect that
> serial 10 GigE chips will also have the same issues, and the PHY chips will
> consequently cost roughly the same amount.  Moreover, I suspect that if 10GigE
> can piggyback on the OC-192 chips, it will help bring costs down for both
> markets.

Alternatively, 10 GbE multi-channel chips will put the pricing pressure on
serial chips since the end user doesn't really care that his data was rainbow
flavored when it went through that piece of glass connecting his boxes.
Best Regards,

> Stuart

Richard Taborek Sr.                 Phone: 408-845-6102       
Chief Technology Officer             Cell: 408-832-3957
nSerial Corporation                   Fax: 408-845-6114
2500-5 Augustine Dr.        mailto:rtaborek@xxxxxxxxxxx
Santa Clara, CA 95054