FROM
ORG.
To
cc
Bcc
DATE
TIME
SUBJECT
FOLDER
ATTACHMENTS
ATTACHMENT
HEADER
MESsAGEID
BODY
Geol Tate
Rambus, ine 7 :
rorrait997
160831
41 with pat gelsinger (and pete macwillams) in portand 10/13/97
\Personal FolsersiRambus imports
gelsinger 1-0n-1 971013 ppt
ATT00226 ot
ATTACHOO0O\am ret 0 32, 19871013141354 00Gedd70€ 198,147 244 4.#1-qelsinaer 1-on-1
TTA ss bdmarket9s3 0 22.19971019141: 198,147 244 4.92.81
Path: sun rambus com'gateway
From: glate@rambus.com (Geof Tate)
"Newsgroups: rambus bamarket
‘Subject. 1"1 with pat gelsinger (and pete macwiliams) in portiand
70n397
Date. 13 Oct 1997 16:04 31-0700
Organization: Rambus, Inc
Lines: 10530
‘Sender: daemon@sun rembus com
Message-I0.
NNTP-Postng-Host 198.147.244.4
Mime-Version 1.0
Content-Type. multiparvmixed, boundary="
876808852:
rambus bamarketé3 0.32 19971013141354 006ed670@ 198 147 246.4
i had requested a 1:1 with pat g as a result of his request, thatsurprised us, to lower our
rdram royalties to <0.5% that he made at ourlast gelsinger/parker meeting about a month
ago. the insert below isfrom an email pat sent me following up after the meeting giving
hisreasons; >>>
pat gelsinger 9/17 Note, we expressed two concerns: - segment 0, eg. other lower
cost solutions will evolve/arise. Thesesolutions need not be equivalent performance,
in segment 0, absolute cost is thecritical factor. - The overall industry and individual
royalty that is being paid. Individual or groups of companies will analyze their $10's
million royalty and insiston developing alternatives. The combination of the two of
these is what raises our concern to an unacceptable level
i'd requested a 1:1 but pat had pete macwilliams sit in also(pat/pete/randy b and some
others by video conference had a rambus reviewmeeting that finished up just before i
arrived). the slides i used areattached (powerpoint/pe format),
dave told me to give away little or nothing but leave pat happy - i thinki did this, although
i set some expectations/deadiines for concreteaction.
update ——-- on rambus update pat's one question was what % of the market do you
thinkyou are going to win - i said that from an assessment of our valueproposition we felt
we would win 80%-ish of the market; the other 20%were not (yet) performance-limited so
we'd win that too as volume draveour costs below siower commodity memories,
i did a recap of direct rambus program progress. i told pat this was anoverview and we'd
dig into details at the regularly scheduled execmeeting next week
‘on power, pete agreed there has been good news BUT he says we need somemore due
to thermal issues otherwise we'll have more costs problems (dueto heatsinks?)
‘on dram partners they feel there is progress on motivation. i said i wasin process of
meeting with dram partner execs to tell them about warrantsfor TTM and 10MU shipped -
R 229266
RH 361761 (OUTSIDE COUNSEL ONLY
¢x0961-001 |30K each for first 3 to qual direct rdram and70K each for first 3 to ship 10MU. i explained
‘we were doing both sothat laggards like micron could feel they could get in on the
2ndmilestone if not the first. pat seemed to like this, i gave copies topat of LG's rdram
ad, samsung's ad that mentions rdram, and ebn's articleon top 3 koreans being
‘committed to rambus. also copies of the dvorakarticle (he'd heard about but not seen)
and hp test ad
i mentioned micron even has assigned a design team - pete said that as ofthe week
before IDF terry lee was calling raging on him; he is supposedto have a telecon with
terry lee and will give me feedback later thisweek as to whether there is any difference in
behavior by terry lee onthis next phone call, pat described micron as "a mutual thorn in
‘ourside” - he's not too worried about them alone - their read is that justcompag server is
listening at all to sidram and maybe micron’s pc groupis doing a chipset for sidram but
it's not concerning them.
die size is the new BIG concem at pa''s level - he’s concemed that ourdie size premium
could price us out of much of the marcet in 99,especially if the sdram guys keep
shrinking and the rdram guys don't. iexplained that TTM is priority #1 for rdram now and
massing engineersfor shrinks will happen for rdram too (as it is now for sdram)
‘oncevolume is seen to be growing, pete's concems are 3: 1. is there some fundamental
reason why rdram's will have a PERMANENT diesize penalty vs sdram (i.e. if equal
engineering was invested now toshrink, what would the delta be - what will it be when
this happens downthe road). 2. the dram guys are not planning rapid shrinks - if
directrdram/camino/colfax are successful the ramp of rdram will be the fastestswitchover
the dram guys have ever seen and are they prepared for it. 3. this is going to be a
month-by-month issue where we have to beshrinking die size with partners continuously
to stay close tosdram-100. i explained that dram execs ARE skeptical of intel’s claims on
howquickly the market will switch over because they've heard it before fromintel but the
switch always takes longer - so they assume it will thistime too. "ACTIONS" -
ALLEN/DAVE R: need analysis/story post-nec/samsung die size analysesthis week as to
why we don't have a permanent impairment - JOINT WITH INTEL: need strategy to
convince the dram guys the ramp isgoing to be real. pat's feeling is to do this in q1 after
they've seenintel deliver on the sdram-100 product ramp so intel has highcredibility. pete
says he hasr't seen 2nd generation shrinkplans/schedules from most dram companies
which concems him (ALLEN/DAVER??).
as for meeting at comdex, this was a suggestion IF dram execs happen tobe there
anyways - “ACTION™: GEOFF to check to see if dram partnerexecs for top 6-7 will be
in comdex 11/17-18 and if so set meetings withrambus/intel-pat g/gerry p(?)
pete said that for large x86 systems they are getting issues on chipkill. | said we have
already presented a technical solution to largesystems companies but thought it wasn't
an issue in the intel x86 space. pat says with merced they are getting commitments form
high end systemscompanies that never before used x86. “LAURA F"*- need to arrange
forpete to see billy’s chip kill presentation
pat was very happy with IDF and with our presentation at IDF. pete saidthe feedback
from attendees was that the memory track was #1 rated. patsays IDF will run
TWICE/year, next will be february - suggested we lineup our first RDF to tie in with
february IDF. ““SUBODH™
patipete says they agree sidram is going nowhere, but they don't agreethat dar is losing
ground. in fact they see that ddr has GAINED groundin the last 3 months - 2 big guys
went to asia recently and said they've decided on ddr, unidirectional strobe and told dram
companies to get behind it or not gettheir business (ibm/hp?) for production late 99-ish
(after direct rdrampe's) - they are getting concerns from pc oem's, even ones that are
behindrambus, that ddr-based pc's could be a threat to them and how does intelstack up:
intel is having to go out to oem's to explain ddr’s weaknesses, etc. - they think ddr-based
R 220267
RH 381762 (OUTSIDE COUNSEL ONLY
¢x0961-002 |C's could be in the market 1H98 with potentiallybetter performance than sdram though
far short of direct rdram: they areconcerned that 2nd/Srd tier vendors, like in taiwan, who
sell only oncost could be woo-ed by DDR - they think there will be several DDR chipsets
in 1H98 and samsung isstill pushing hard - i told pat/pete that for example compaq went
from considering DDR 3months ago to being firmly in direct rambus camp for all
systems; andthat things were going on with large systems and other customers where
wewere winning designs in q3: so ddr wasn't dead but we had taken businessaway from
ddr - we agreed that we both needed to keep ddr #1 on our threat list andwork to beat it
overall relationshipiprocess for business issues/pat's request
pat confirmed he is the executive owner
of the rambus relationship. hehas regular MRC (management review committee
meetings) with other execswho are interested in rambus (gerry, albert yu, avtar - need to
addserver/mobile reps soon he says). barrett is not directly involved butis kept up to date
periodically along with lots of other stuff
i reviewed the deal we agreed to with intel 11/96 and covered all of theconcessions that
intel got that no other partner has. not clear if thiswas news to pat or not - i think some of
it he had taken for grantedbefore.
pat asked what our market cap would be without the intel deal - i saidmaybe $200-300M
versus >$1B plus today. i told pat we are happy andgrateful - i just reminded him that we
gave intel a LOT in the deal too. so it's win-win
i said since the deal was signed in 11/96 we have had a great workingrelationship with.
huge progress and we are basically giving inteleverything they wanted - the best
solution, on schedule - yes there areissues, yes we have to work harder but we're in the
zone.
i said on business issues we have had a lot of requests from intel sideto change terms.
often uncoordinated and unilateral. pat/pete objectedto this and said they tried to do a
good job of funnelling through chung. i used pat's royalty request as an example and
contrasted to what chunghad told us the day before - huge difference. (i did this
diplomatically’ think without any putdowns). i said asking for changes is ok but theway
we do it with other companies and recommend doing it with intel is - raise, discuss and
reach agreement on business issue major changes atthe EXEC level: then get 1-2 page
email confirming the agreement andmid-level can work out the detail. i asked them NOT
to raise majorissues through chung as chung is not empowered and we lose
valuableinfo/details in communications since chung hasn't got the request directfrom
gerry/whoever. - need to balance gets with gives as much as possible; if there
issomething painful to rambus being requested, how can intel give somethingelse to
rambus that helps to offset?
pat said that maybe we felt intel was in win-lose mode but he didn'think that rambus
making money necessarily meant it was coming out ofintel's pocket, i said overall we feel
that intel management views thedeal as win-win, although out of the 100's of people at
intel there werea few isolated individuals that might view it as zero-sum (pat smiled
andsaid he knew their names - i think we were both thinking of larry paleybut we didn't
say his name)
pat elaborated on his concer raised at the pat/parker meeting in earlyseptember. he
said their market now is 2-3M servers/workstations; 20Mhigh-end desktops, 60M
mainstream desktops; and 15M mobile, and that thebig change is the mainstream
desktop market is bifurcating into 2 chunks- the upper chunk is very similar in
dynamics/needs to the high-enddesktops, just lagging in time: the lower chunk is
segment 0, 10MU todaybut possibly 30-40MU in the near future, where the needs are
verydifferent and cost is more critical than performance. the performanceintel already
has is adequate for segment 0 and more performance won'twin it - lower cost is what's
needed. so intel is doing new chipsets/newmotherboards/cost reduction efforts/even new
R 229268
RH 381763 (OUTSIDE COUNSEL ONLY
¢x0961-003 |