_______________________________________________
- THE FRIDAY LETTER -
(emailed weekly,
from Gilder Publishing,
for friends and subscribers)
_______________________________________________
| http://www.gilder.com/ | Issue 315.0/November 2,
2007
SIGN-UP A FRIEND FOR FREE!
HEADLINES:
- The
Week / The Avatar of the Graphics Processor
Paradigm
- Friday Feature / George Gilder on Telecosm’s Trajectory
- Friday Blogger Bonus / Explaining EZchip's Enviable Position
- Readings /
The
Week /
The Avatar of the Graphics Processor Paradigm
Gilder Telecosm Forum
Member (10/21/07): George, Where did you find Jules Urbach? Amazing,
impressive technology. I very much enjoyed his
presentation [at Gilder/Forbes Telecosm 2007, Oct. 16 – 18, Lake George, NY]
but I'm curious as to why you chose him?
George Gilder (10/22/07): Jules Urbach is the avatar of the graphics processor
paradigm that I have been touting for the last year or so. With the fully
unexpected and disruptive emergence of massively parallel processing from the
humble game machine rather than the lordly supercomputer, Jules has taken
Hollywood cognoscenti by storm.
His most visible achievement is real-time rendering of photorealistic 3D
images from a computer program. This means what it says--he can create a
photorealistic scene without using photos. By doing real time rendering at 30
or more frames per second from a computer program, he accelerates the graphics
process by roughly 10,000 times the speed attained by the paladins of Pixar
and Industrial Light & Magic (Lucas Films).
I am now trying to introduce Jules to Silicon Valley and MIT. Anant
Agrawal, the MIT prof multicore king at Tilera, told the Telecosm 2007
audience that the rise of the graphics processor bespeaks the total failure of
computer architects to master the software and tectonics of parallelism. How
Atiq Raza of Raza Micro would have responded remains unknown, since he
did not make it to Telecosm (family emergency). Nvidia (NVDA) and ATI/AMD
(AMD) were also absent without leave. Something is going on out there, and I
don't know what it is, but Jules has a lot of big announcements coming.
Urbach’s breakthrough is an 800-kilobyte runtime engine that can operate
as a virtual machine in any browser for rendering and displaying video images
scalably on any device, from cell phone to supercomputer. It includes the MPEG
and JPEG standards but is chiefly optimized for high-definition and 3D images,
in particular a “Second Life”--like Jules World. Since the virtual glitz
of Second Life’s computer graphics, saliently inferior to what Jules can
do, still accrues some $50 million per month of revenues, this capability is
financially significant.
The runtime engine client, resembling the Java Virtual Machine, is also
written in its own optimized language, called OTOY, which taps the specialized
instruction set of graphics processors from ATI and Nvidia to gain enormous
efficiencies in rendering computer imagery.
As calculated by its Hollywood users, such as James Cameron (“Titanic,”
“Transformers,” and Digital Domain), Jules accelerates the rendering process by
an unbelievable factor of 13 million. (10K, says Jules, or 13M, says Cameron. I
suppose it's just numbers.) In any case it's said to reduce the time to
complete a video frame from five days at Pixar and Industrial Light & Magic
to a real-time, 30 frames a second at Jules World. This technology cuts the
cost of animating a film by between 40 and 70 percent.
It is evidently a new paradigm in computer graphics. It begins with the
efficiencies of a graphics language coupled directly to a graphics instruction
set and continues with the elegance of a runtime engine on a server linked
directly to the client VM, though Jules can also use the server model alone
without the Virtual Machine on the browser. The climax is spectacular
scalability and an object-oriented transactions model. (Qualcomm/QCOM
uses ATI instructions in its cell phone chips.)
For Silicon Valley, the advantages of OTOY span four key elements of the
innovation:
1) The Value Proposition for Users: Fast proliferation and publication of
3D and HD video, with object oriented digital rights management and
transactions.
2) The Market Segment: The first user “bleeding edge” is the film
industry, which is betting the next generation of its supremely profitable
animation business on Jules. But this tool disrupts from above and can be used
by the YouTubular multitudes and allow an eBay model of commerce in digital
video objects.
3) The Joy’s Law Value Chain: It is a tool for creating and distributing
product and tapping complementary assets (artists and programmers) that Silicon
Valley cannot own. However, commanding browsers and the datacenters, thin
clients and the massively parallel servers, leading incumbents Microsoft,
Google and Yahoo do own key elements of the dominant computer architecture of
the epoch (see my Wired piece, “The Information Factories,” 10-06). Google in
particular also dominates the long-tail advertising model. And Silicon Valley
needs Jules to protect the its value chain against disruption from above (3D
games and environments) as broadband spreads and the Jules revolution prevails.
4) Revenue Generation: Silicon
Valley companies can collect micropayments per object, per associated ads and
clicks, and per transaction enabled by the OTOY DRM system. The object oriented
transactions model of OTOY amplifies the online advertising stream with direct
payments where appropriate, currently a significant gap in the Cloud computing
model.
Why Now?
The previous video model emerged during an era when bandwidth was a
thousand times greater within the computer than on the network. All video
processes had to be jammed into the twisty little datapaths of customized
processors, with their proprietary instruction sets and analog cores and with
their multi-tiered memory architectures, treacherous internal buses, and
incompatible languages and operating systems for every platform from the
set-top box to the film projector to the satellite feed to the animation
engine.
Today launch of a new computing paradigm based on the browser and the
datacenter and on new multiprocessing and multithreaded video engines
distributed in every computer, game machine, and cell phone calls for a new
software paradigm.
To me, the most intriguing speech at Telecosm was from Lane Patterson, Equinix
(EQIX) Chief Technologist, who explained that new all-optical and optoelectronic
innovations driven by the new architecture were streamlining the network
infrastructure. The new computer architecture will be accompanied by major new
network breakthroughs that will become manifest over the next few years.
That's the significance of Jules World.
To read more posts by
George Gilder and the Gilder Telecosm Forum members, visit http://www.gildertech.com/ and log on today.
|
The Gilder Telecosm Forum To
learn how to join this powerful network of talented, tech-savvy investors and
thinkers online daily to debate, discuss, and decode new and emerging
technologies and share valuable and actionable investment advice, visit www.Gildertech.com
today. |
Friday Feature / Telecosm’s Trajectory
George Gilder, Gilder
Telecosm Forum (10/18/07): Telecosm
2007 was the most coherent Telecosm, given an ever more richly woven thread by
the defection of several irrelevant companies which were apparently spooked by
the idea of global warming skepticism and which thus opened more time for EZchip
(LNOP) and the other companies.
CEO Eli Fruchter's chief news was elaboration of his access chip family, which
increased the EZ available market threefold. Nothing else he said was anywhere
near as significant. EZ now has some six or seven coherently designed devices
on its path. Eli also observed that no other company had an NPU product family
that ranges from one gigabit per second to 100 gigabits.
He also ran into a siege of criticism for his claim that pad-limits would
make application specific integrated circuits (ASICs) as costly as network
processors (NPUs). He is right that the enormous I-O demands entailed by
several Ether-ports plus content addressable memory (CAM) links (think NetLogic)
added to switch fabric interfaces render network processors a peculiarly I-O
intensive product. Pad limits can be overcome by most chips, as the Broadcom
CTO asserted from the audience floor, but Eli is probably right that they will
not be mastered over the next five years by ASICs.
Andrew Schmitt (Nyquist Capital) made a cogent argument that NPUs devoted
to the first three layers, as he said EZ was, are particularly vulnerable to
ASICs. But even though the current chips are chiefly used in the first three
layers, EZ's architecture is adaptable to seven layers. From the beginning what
appealed to me in EZ was its scalability. This strength applies both in its 7
layer potential and in its potential at 100 gigabits and above.
There was much amazement expressed about EZ's survival in the face of perhaps
100 rivals spending thousands of times more money on NPUs. Intel spent 2
billion on its NPUs and still failed because of architectural flaws. One theme
of the conference was the parallelism imperative. EZ happens to have the best
multicore design in the industry, transcending the law of diminishing returns
that afflicts most multicore architectures.
The fact remains that EZ and Bay Micro (also represented at
Telecosm) are the only survivors and Bay is a vertically integrated supplier to
governmental customers who need the asynchronous transfer mode (ATM)
capabilities that Eli eschewed from the beginning in his visionary assurance
that networks would be based on Ethernet and IP.
What was exciting about the conference was its trajectory, from a rousing
debate on global warming to stirring and hilarious songs by Master of
Ceremonies, Jeff Stambovsky, through a series of fascinating presentations by
such voices as Steven Sprague of Wave Systems (WAVX) accompanied by
Trusted Platform Module (TPM) supporters from Intel (INTC) and other
companies. Alex Dickinson (Luxtera board member) on CMOS optics, Chris
Cooper (Seldon Laboratories) on nanotech, Jules Urbach (OTOY) on
graphics processors, and perhaps thirty other companies that remain a dim
memory through the wine and roses and Carver Mead's climactic speech on the
emergence of the “noosphere” in the form of the Internet.
To read more posts by
George Gilder and the Gilder Telecosm Forum members, visit http://www.gildertech.com/ and log on today.
________________________________________
Friday Blogger Bonus / Explaining EZchip's
Enviable Position
Gilder Telecosm Forum Member (10/22/07): There are a couple of themes and risks that
need to be taken together that better explain EZchip's (LNOP) enviable
position...
1) "The hollowing out of the router"
2) Cisco will never outsource a critical piece of its architecture.
3) Cisco will buy EZchip to "freeze out" Juniper (JNPR)
or its competitors.
4) Cisco will have an internal development effort to in-source the network
processor down the road.
All of these common refrains are based on the perception that Cisco (CSCO)
and the networking companies view themselves as hardware companies. They do
not, they view themselves as software companies. Ask them. Ask them where their
competitive advantage lies or where the point of differentiation is... every
company will say it’s in their operating system. Ask the point blank question:
"Are you concerned that the increasing role of merchant silicon in your
products will cause value to migrate out of your company to your supply chain
companies?" The answer is an unequivocal "no" and the margin
structure and trends of the industry support this.
This is why in my opinion the router will never be hollowed out... the
CESR market has no Microsoft (MSFT) equivalent. If you look at the
margins of the PC industry you will see the PC is in fact hollowed out as PC
manufacturers live on razor thin margins while Intel (INTC) and MSFT live on
fat margins. Cisco's margins are more akin to MSFT's margins than they are to
Dell's margins and the more merchant suppliers added to the Cisco supply chain
the fatter those margins get. If the router was truly going to get hollowed out
Cisco and JNPR would be amazing shorts as their 65% gross margins plummeted to
10%.
In this vein you can see why Cisco does not view the network processor as
"critical piece of its architecture" but another bit of hardware that
should be outsourced. That's why there is no chance that Cisco will buy EZ or
build its own processor.
The analogy that I believe is appropriate is that EZ is the De Beers
of the networking business. No jewelry retailer would think it imperative to go
out and buy and own a diamond mine to establish a point of differentiation with
its competitors. Rather its the setting, design, customer service, warranties,
pricing, store positioning etc. that provides a point of differentiation. Just
as buying from De Beers is a logical supply chain decision for jewelry
retailers so is buying processors from EZchip, CAMs from NetLogic
(NETL), and optics from Finisar (FNSR) logical supply chain decisions
for the networkers.
George Gilder (10/22/07): Thanks for another superb post.
I think Cisco is analogous to an IBM (IBM) that refrained from
selling its operating system rights to Microsoft, but instead held them, as IBM
did for the 360 in previous generations.
That is why I believe that Larry Boucher of Alacritech provided the
most intriguing insight of Telecosm. (I would have intro-ed him as the
legendary Larry Boucher, inventor of the SCSI drive.) He declared that Cisco
would be the dominant computer company of the next generation. It integrates
the backplane for a computer on a planet.
But there is still easily room for an Intel amid the hollowed hardware.
To read more
posts by George Gilder and the Gilder Telecosm Forum members, visit http://www.gildertech.com/ and log on today.
__________________________________________
Readings /
Forum
Looks at Chip of Future
http://timesunion.com/AspStories/story.asp?storyID=631048&category=BUSINESS&newsdate=10/18/2007
Weekly
GTI Index
http://www.gtindex.com/
MySpace, Bebo and
SixApart To Join Google OpenSocial
http://www.techcrunch.com/2007/11/01/confirmed-myspace-to-join-google-opensocial/
The
Next Billion (Forbes.com)
http://members.forbes.com/forbes/2007/1112/048.html
Sprint
Weighs
WiMax Unit Spinoff
http://online.wsj.com/article/SB119394350910379407.html?mod=technology_main_whats_news
Lenovo
to Drop the IBM Logo
http://online.wsj.com/article/SB119391251123478937.html?mod=OATE
__________________________________________
FRIDAY LETTER STAFF
Editor: Mary Collins George / mcollins@gilder.com
Research: Sandy Fleischmann / sfleischmann@gilder.com
ADVERTISING INFORMATION
The Friday Letter is mailed each week to more than 40,000-plus
subscribers and friends of Gilder Publishing, including industry leaders,
financial professionals and individual investors. For information about
advertising, contact Linda Bentley at lbentley@forbes.com.
PLEASE NOTE: The appearance of an advertisement in the Friday Letter
does not indicate an endorsement for the product and/or service by George
Gilder, Gilder Publishing LLC, or the Friday Letter staff.
FEEDBACK AND PROBLEMS
For technical problems, or to send letters to the editor, please
e-mail info@gilder.com.
MAILING ADDRESS
Gilder Publishing, LLC
ATTN: Friday Letter
291A Main Street
Great Barrington, MA 01230
_______________________________________________
The Friday Letter is published weekly for subscribers and
friends of Gilder Publishing. If someone you know would enjoy it, please feel
free to forward a copy.
Gilder Publishing makes the Friday Letter available for free. To
help defray some of the costs of producing this information on a weekly basis,
we will from time to time be sending you offers from companies we think you'll
be interested in. These offers will not come more than once a week. If you do
not wish to receive this related information, please opt out of this process at
the link below and we will not share your name with companies outside of Gilder
Publishing.
To SUBSCRIBE please visit http://www.gilder.com/
To UNSUBSCRIBE please go to http://www.gilder.com/fridayletter/unsubscribe.php
Trouble subscribing or unsubscribing?
Email info@gilder.com
http://www.gilder.com/unsubscribe/specialproducts.php
To SUBSCRIBE please visit http://www.gilder.com/
To UNSUBSCRIBE please go to http://www.gilder.com/fridayletter/unsubscribe.php
Trouble subscribing or unsubscribing?
Email info@gilder.com
_______________________________________________
Copyright 2007 Gilder Publishing LLC