AMD CC "transscipt":
-------
SANDERS:
Recap 99: produced close to 20mil units. fc package. high point Athlon, intro in slot... later in the year began 180nm produciton
approach. 30 mil this year
tilt HEAVILY to Athlon in Q4
completed conversion of all austin Athl. to 180nm in 1st half
Austin pilot line... ?
transition of Athlon from slot to socket... more compact and cost effective.
exiting part of 1.2GHz part is that it comes from fab30.
--> Over 80% of Dresden output at 1GHz+
6 layer dual inlaid flow, all copper back-end. (re: 180nm)
90nm transistor gate lengths currently!
(gate which controls switching speed of the device)
Outlook for 2001: Fab25 fully loaded all year @ 180nm,
[something about 130nm]
are installing SOI process flow in Dresden, will be used for first hammer products.
The Alliance effort continues, where the 130nm flow will be qualified this year, with work on 100nm continuing next year.
193nm wavelength required for 100nm, may also see some utilization in the 130nm node.
We're also pushing the front of scaling the transistor down, dielectric size....
SOI: blah blah explanation of SOI manufacturing. "significant improvement in the power of the device"
Memory:
name of the game = capacity
99: 250mil units.
completion of ramping of JV2 fab (FASL)
completion of some other fab
second fujitsu fab in oregon, qualified last Q, ramping now.
in june we approved JV3 with shell space for JV4. on "fast track"
assembly and test:
Malay and Tai
fbga package size, compact, space efficient.
deployed in process line bringing some batch processing to the back line.
99 brought online newest assembly plant in china, now ramping.
JV2 and JV3 will intro 180nm flash, will occur in market place in 2001.
130nm R&D under way. Late 2002 introduction (memory)
expanded package versatility.
-MCP
-ability to ship bare chip to customer so they can do "their own thing"
Flash reliability:
- endurance (# of cycles > 1mil)
- data retention (10 years @ ..degC)
Outlook for 2001:
350M units
intro of 180nm productions
production from JV3, 15 months from start of fab build.
---
Try to explain flash market.
99 recap: Flash market has become the fastest growing market.
AMD believes the forcasts have been very conservative.
AMD believes that the shortage will continue.
some company: world shortage of flash will be the worst "for 5 years"
We expect it to remain for 2001 and 2002.
graph showing parabolic growth, devided up by sector (e.g. cell phone).
Looking closely at cell phone:
overall growth is slowing. for the year 2000 415mil units (handsets). 47% growth. Expected to decline to about 10% - considered conservative in the industry. Cell phone industry is changing. GPS and increased man-machine interface will increase the requirements for flash in each handset. Expect double demand each year.
since beg. of this year, AMD has added 3 fabs (JV2 and 2 others). JV3 expected to come online 2nd half of 2001.
Bit shipment increasing exponentially, but still very far below demand. AMD @ 77% compound annually, but demand @ 102%.
WSTS has reported at Q3 that ave. flash price has continued to rise, AMD expects this to continue through 2001.
FLASH contracts are mostly through 2002, several have been extended through 2003. All carry cancelation penalties.
Based on customer demand, we see exceptionally strong flash market. Supply / demand imbalance continue through 2002.
Few suppliers can produce: AMD, Fujitsu and Intel are the only REAL players. [The rest don't have good enough technology]
AMD innovations in flash markets:
Cell: lov voltage down to 1.8 and 1.65. zero power during standby.
network: 20year data retention
automotive: burst.... 1mil cycle endurance
"We are the only company in the world with such [a wide] product portfolio."
"Because of our relationship with Fujitsu" .... able to cut cost of development in half. Bring fab from groundbreaking to production in less than 15 months.
Highest yield and ...
AMD's focus:
25% for internet infrastructure. Market 8%
cell phone, market 48%, AMD 36% (not 100% positive about numbers)
This is by design.
Because of this marketing strategy we have been able to establish very strong relationsships with our customers - able to pick the market leaders.
practically the exclusive supplier of e.g. cisco, erikson... blah blah
win-win partnership with LSI and SDI.
They will enable us to service the ever increasing requirements of the ... ncp package that may contain dsp or ...
will turn telephone into multimedia terminal with full access to the internet.
spot market prices are still several times market prices.
--------
DIRK MEYERS
Past year: blah blah blah leading in both clock speed and performance blah blah blah
"pretty good support" among third party chipset makers.
2001-2002:
- continue leadership.
- intro AMD athlon and duron in mobile segment.
-
-
Athlon roadmap:
soon introing 2 chips mustang core products:
Palomino
- first palomino intro this q, 1.5GHz at end of Q2
- 130nm in Q4
- mobile palo in Q1
Morgan
- value space, roughly a quarter behind palo.
Sustaining performance leadership on desktop:
Performance made up of 2 parts, clock speed and effeciency
Based on publicly available info we believe AMD athlon 1.2GHz performance exceeds P4 @ 1.5GHz.
Coupled with platform and cost (die size) advantage we believe we have the best solution ....
K8 further enhance this position... well ahead of the historic trend... 3x performance of 1GHz Athlon.
Intro of DDR: blah blah
SMP: production shipments in 1H01. entry level servers.. esp. appliance server space.. often in rack mount form factors.
Platform innovations:
LDT: available starting next year from AMD and others.
leverage 4 and 8 processor systems for hammer family
The multiprocessor problem:
shared bus topology. problem is it doesn't scale well. memory bandwidth and bus load problems. AMD will intro NUMA architecture.
Observations on 64-bit computing:
there are not now nor will there ever be a majority of programs that require 64-bit.
x86-64 extentions fully support 32bit - with industry leading performance.
expect OS to come in time for our products.
advantages for businesses blah blah blah
K8:
goals: sustain and increase position on performance desktop.
- > new architectual innovations.
- > pipeline changes
2 products:
clawhammer: Performance desktop: sampling in Q4, proction shipments in Q1, 02. This product will be less than 100mm2 in 130nm.
sledge: 4-8way server class processor.
Hammer family will heavily leverage LDT and DDR.
-----
ROB HERB
ARG! just lost a huge portion of my "transcript". BAH! Oh well.... I don't feel like listening to it again right now.
later, maybe.
-fyo |
