Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
spokeshave - elmer
Thanks for sticking in here. As for Elmer, his devils-advocate-role is indeed worthwile for balancing the discussion. This role can be misunderstood very easily as to be emotionally biased or at least skewed; which I find not the slightest indication he would be. For this misunderstanding he has taken hits bigtime on other boards. I would wish this board appreciates any reality check as long as it is done in appropriate form - which you cannot claim Elmer does not provide unless you take his statements as peronal attacks which they are not intended for (admittedly very easy to misunderstand, as I said).
As for AMDs approach to be considered independent from feasibility, in fact from an investment-standpoint it is higly interdependent as long as you intend to capitalize on it.
And as for the weather forecast, yes we should consider to pick an umbrella for tomorrow - even if we thing we would not need it - people selling sunscreen-lotion we have a lot already.
And yes - sometimes it is good to have somebody trying to save us from ourselves - they will most probably fail with it anyway. K.
elmer - hype
If you understand "confidence in AMDs abilities" as "believe the hype" the answer is yes. And I still do. As for "proven unmanufacturable", not yet - but not proven manufacturable either. Appreciate your advice however, welcome anytime.
Klaus
Elmer Replay
I remember... iirc IBM and Apple were already in an uphill struggle in these days.
If you consider Hammer as a replay, X86 will win again. :)
What convinced me from the very first moment of the idea is its evolutionary approach of transition. Everybody who tried a revolutionary approach in the last decade failed badly.
Btw, some DNA of the original PPC still lives in IBMs fourth (and soon fifth) generation of CPUs. Going into Macintosh as unicore-970 maybe - could stretch the life of this architecture for sometime longer.
K.
Elmer, Apple and the bigger picture
You sure know a lot more about PowerPC than me. However, what I know is the architecture was done quite a while ago. Today, neither MOT, IBM nor anybody elso would make any thought of designing and manufactoring a leading-edge CPU based on definitions of Apple, you bet - unless Apple is capable and willing to carry the cost for it. Apple can only buy whatever they basically already have (which is good enough - Apple can still make USP-products with what they get).
In the bigger view:
Developing leading-edge CPU-Chips for leading-edge process-technology (highly interdependent!) has become so expensive nowadays that only really huge volumes of the resulting product can return the investment.
Btw, Paul Ottelinis vision presented lately at Goldman Sachs (rebroadcast available via Intc.com) is at the end nothing else than the above from the viewpoint of Santa Clara - working out implications for intels endless prosperity.
I would agree on his vision first-hand and even consider this market as "closed shop" of some five manufacturers for the forseeable future - and likely tending to three within the next years. AMD btw is one of the shaky candidates: Even assuming their architectural concept is competitive and they prove it is manufacturable in desktop-volumes on 130nm-SOI they can probably not make it on their own if they cannot keep the pace in volumes because of lagging the 90nm-process for more than a year or so - unless 130nm-SoI shows excellent yields plus scales far enough. We all have experienced what happened when they were just couple of months behind the curve for the 130nm-node.
Just for the fun of it - there is an even bigger picture:
Imminent contrentration-processes (not only) in capital-intensive technology-sectors were brought up long ago by Karl Marx for the "late period" of capitalism. No, i do not share his conclusion what will happen after that period. Growth and concentrations are somewhat self limiting from my experience on this planet. (This coloured picture - of what mankind is driven by - does not fit in the frame of this posting)
So to say, Marx and Ottelini are both most convincing under the ceteris paribus restriction - which is hard to assume forever in an industry where the only thing that will never change is that there is change. Which is fun to follow - and can be really painful sitting on the horse trailing....
K.
neye - niches
In contrary to SUN, Apple is not designing and producing CPUs, which is what my posting was about.
Klaus
yourbankruptcy
Not that I would think you made any mistake. What confuses me on pricewatch is:
Looking at the CPU summary e.g. for 2800-model I find two lists, one marked as 333. However looking into the other list (without 333 specified) there are plenty of 333MHz FSB parts in. Some vendors specify the core, some not. Some vendors are in both list, some only in one section. Sometimes looking just at the summary you see a sharp drop just to find out going into the list that this is for minimum 20 CPU - the following day you see the price rising because this offer has been taken out.
So I do only use pricewatch for a rough longterm-trend on prices, especially to find out which model is travelling from performance to value. As long as channel-inventory is properly aligned, you can make something out of it how the process is going (which was totally impossible last year).
As for ASPs btw, the volumes going to OEMS definitely have pricetags which have little to do neither with any official pricelist nor the prices you see on pricewatch. So from pricewatch-data you cannot really calculate anything for ASP - maybe it helps to confirm any guesses or to adjust it to one side or the other. Same is for unitsales and distribution of speedgrades. Pricewatch only represents consumer and maybe partially SMB-market in U.S. (where AMD lost unitshares bigtime last year). Looks somewhat different in Europe (where similar sites allow to do the same as you did)
and totally different in Asia (where such site exist as well, but in local languages). K.
Combjelly - disagree on niches.
Until today, Sun has (only) lost the competitive edge, marketshares and some money recently - but not its reputation and the "critical mass" for their business (the latter except for Chipdesign, which is the problem). Scaling down to niches would make it far worse for SUN. There is an analogy to AMDs situation by the way, Bob Rivet adressed that recently as "You just cant cut yourself to profitability in this industry". That is true for SUN as well.
As for the pity, sure hope Scott has an excellent plan I am not aware of. If this involves AMD in any way, even better.
K.
CombJelly - definitely agree
And might add that I dont see a longterm way for Scott to be able to keep SUN alive without giving up its proprietary hardware anyway. (To prevent any misunderstanding here, their design is excellent - i only see no way to achieve volumes or prices for SUN to feed this excellence to keep up very much longer. So it'll most probably go the DEC-way. A pity.)
K.
yourbancruptcy - CPU availability
What data did you use for this analysis? Pricewatch?
(I can see the drops in 2200 and 2400, for the rest going into details i get confused).
In this context: Anybody has heard anything about Core-stepping to Tbred-C? I'm asking because this stepping i assume to adress yield issues - likely sacrifying top-end clockrates.
This could explain what you found: Waferstarts since November or so for Bartons (TbredB-core for 2500+ upwards) and TbredCs for Value-products below that.
K.
Combjelly - thanks
Touchée. Kasey Holman definitely speaks for SUN. Apparently they will use AMD for Uniprocessor-Servers first - hand.
The P3-based Server SUN sells today (as uniprocessor and dual server) is called LX-50 btw, that is what wmbw referring to,
guess he assumes that is common sense and needs no support.
Klaus
yourbancruptcy
Maybe you are right with some or all your maybes. :)
A lot will depend on the platform-side of the equation as well and many other things, so to make it clear: The cited statement about the CPU-compromise as to be the only parameter to choose any CPU is nothing less than rubbish, insofar i do agree in any maybe
K. (who would be anything but unhappy if SUN would replace what they are using now (that is what wbmw stated correct))
CombJelly - c'mon
The private statement from somebody in France cited in your link pretending to speak for SUN says explicitely that a Chip from AMD not yet released will be used in the second half of this year.
Neither SUN nor AMD commented on it.
A lil bit thin so far, dont you think?
Plus, what is said about Athlon-M as the CPU of choice does not really make much sense: If SUN indeed is only looking for a compromise between density, heat dissipation, power-consumption and performance, lets face it there would be most probably a better choice for a X86-32-CPU appropriate for server-use in the second half - at least unless prices come into the equation or strategic thoughts come into the equation.
So, what you are so sure of is not impossible at the end. But that is about all what you can really claim for today. K.
kgoodrich - thanks
for sharing the seasonal patterns of AMD common you discovered. As you already did the work and explained your approach as well i have no doubts your findings are accurate.
Klaus
sgolds - thanks
Sure. I seem to be thinking in sort of tunnel. Maybe time for a break... Klaus
sgolds - thanks
Sure. I seem to be thinking in sort of tunnel today. Maybe time for a break... Klaus
Combjelly: Fab25
Thanks for the link. The qualification of FAB25 as a foundry for FASL does only say that the products from Austin are to be considered as meeting FASL-specifications - so Fujitsu could sell it as well.
If AMD Flash sales go accountingwise via FASL this makes no difference as long as AMD does not pay more than it gets for any module. (After a second thought it is not impossible there is some subsidy granted for FASL in exchange for the allowance for AMD to sell in parts of Asia.)
For FASL corporation, its future organization is still under negotiation currently; which besides possible integration of FAB25 involves financing of JV3-Retooling, the Gresham-thing and other AMD-guaranties etc.
K.
kgoodrich:
Not offended, but just curious: Where do you have monthly data from for AMD?
(If not based on monthly data, how did you extrapolate for your revolving results presented?)
Klaus
Buggi - outlook
Hi there. Welcome aboard - hope you are ok.
Als for Flash and Fab25 - I just recently learned from a posting here that there is currently no Mirrorbit produced in Austin. From w:o you already know I never was a flash-hypester, now not only my production model for FAB30 is doubtful since couple of weeks but output in Bit-shipments from FAB25 were obviously miscalculated as well. In result, I am expecting nothing much better than the street for the next two quarters. That is - still a lil bit better. Plus there is still a lot of short interest in the stock - so any upside-surprise plus some flash-squeeze could very easily catapult the stock into the teens in just some weeks- maybe coming back to a p/e-valuation instead of a discounted bookvalue-pricing.
Would be not the very first time to happen.
OTOH, taking into consideration what AMD has put on the agenda of s/h meeting for conversion of stock option strike prices these guys are maybe not really interested in a short-term recovery as they have an opportunity for their own pockets to ride the wave out from bottom to top - but only after May 1.
K.
p.s: Reminds me I promised you to mail you the model-file.
And: I sure owe you that "bottle on the beach" from our ASP-bet.
CombJelly -
FAB25 is not part of FASL. So there is only little increase in costs for variable Costs for Flash from there.
As for FASL, I guess anything produced in JV1 (on 350nm process) is still selling well below full costs. Probably the parts in JV2 are achieving ASP to cover full costs with the help of Mirrorbit (230nm). JV3 is the only 180nm FASL-Facility, currently going partially to 130nm.
As for Palominos, quite sure sold out end of quarter, and TbredA probably as well. Not so sure for Durons though, because it is the only explanation I can imagine for inventories still to "align". These parts must have been selling through really slow last year in most parts of the world. K.
Nice reading for sunday's breakfast
http://www.washingtonpost.com/wp-dyn/articles/A59534-2003Mar7.html
Btw, i like this board - thanks to everybody contributing context and organization. Klaus
econo re: Outlook
For the current quarter and CPU-revenues, Bob Rivet said in his Morgan-Stanley presentation there is still some "channel"- inventory to be "corrected". However, taking a six-digit number into consideration for that it still leaves 6-6,5 Mio units to account for Q1-revenues. In Dollars a lot depends on how many Durons are still on the bills booked this quarter.
Flash should have a better contribution for the current quarter not because of big increases of revenues but for the increased output from Austin reducing bills from FASL - which is on the cost side. (So achievement of "cost reduction" is not really a miracle). In result slightly up seems achievable for me - which would mean an upside surprise.
For Q2, black ink is not completely impossible, but I can imagine that only in a "very best case" scenario under stable market conditions - or with the help of some warm rain falling on Sunnyvale from the flash-skies if Intels vision of an upcoming capacity-sqeeze for high-density modules will materialize.
As for people already starting to look out for the next thing - maybe, and AMD-common would be definitely the right company to look at - as soon as profitability is visible.
However some wording from Bob Rivet in the presentation mentioned above like "it is still a long way to go" plus what AMD intends stock-options-wise may well limit any overly optimistic expectation for AMD common within the current year. OTOH, as I underestimated the downside-potential of AMD common by far, most probably that is true for the upside-potential as well..
K.
CombJelly
There obviously is an expectation of some squeeze for High Density Flash in Santa Clara during the upcoming months - which could easily happen as nobody has significant inventories now; not that i believe it will hold for long but maybe long enough for Intel to take the shares back they concede now using their inventory no other manufacturer has and get better prices for their stuff for immediate delivery; plus allows contracting longer term contracts for better prices as well if signed during the hike. Interesting to watch how this gamble works out. Hope Intels vision is on the spot as it would help AMD for better prices as well. K.
CombJelly
Thanks for updating me on erase-cycles for some 230nm-parts.
This one I missed.. :)
Klaus
wbmw thanks for the wise advise,
about buying the stock however i am afraid it comes too late for me as i am fully loaded. Actually more than that.
As for Mirrorbit or conventional flash I do not care about what they make currently as long as FASL capacities in Japan are fully utilized and they are ramping up FAB25 successfully - as long as it sells for prices to return the business back to profitability. You know, one obvious reason for the flood of red ink last year was that they had all the fixed costs for the Austin-Fab but only marginal revenues for the Durons and a lil bit of Flash coming out of it. As soon as the facility covers its costs the numbers altogether will look different even without big earnings coming from flash.
As for Mirrorbit they already proved it works on 230 nm. Good enough for me as long as Flash is not bleeding anymore. If shrinking to 130nm in one step proves to be unfeasible, there is still a good chance to make it to 170nm and use 130nm for conventional flash instead. So the story is anything else but dead - just a longer one maybe? Taking into consideration they already have a continuing chapter ready (Saifuns 4bit per cell-approach) it is sure intended to stretch the hype until the next book (Polimer at the end of the decade) will be opened.
For what people might assume today thisrespectively, as AMD always shows the mirrorbit-story and FAB25 ramp in the same context the fact people mix these two things together is a mistake easy to make - and, carefully said, AMD is not really unhappy this happens. ;) At least me they caught here - until your enlighting posting that is.
K.
p.s: Which makes me think your ranking in Pete's quarterly contest is founded by tight grip on what is going on.
Chapeau!
combJelly: Flash Charts, structures
Right, AMD ist producing Flash in 170nm currently in Austin,
just not Mirrorbit, but conventional Flash. You can find the products specified as to be 170nm there (64Mbit-Modules)
For confirmation of this, see last Q&A here:
http://www.amd.com/us-en/FlashMemory/TechnicalResources/0,,37_1693_1774~20983,00.html
Q: What lithography is used in MirrorBit technology?
A: Initial MirrorBit Flash devices will utilize AMD’s proven quarter-micron technology.
Q: When will MirrorBit technology move to a smaller lithography?
A: AMD plans to begin producing MirrorBit devices using 0.13-micron lithography during the second half of 2003.
Klaus
p.s.: What would you understand from the (semiconductor language) term "begin producing during the second halt of 2003"?
p.s.s: The 1.000.000 cycles are only guaranteed for modules made in 350nm structures by the way...
wbmw: Thanks for your enlighting posting
In fact, realizing Mirrorbit is currently still two generations behind Strataflash in terms of structures, any cost-advantage of this technique for AMD will only materialize if they can successfully lapse one generation (170nm) and prove Mirrorbit is makable in a 130nm-process in volume next year. No idea how probable it is that shrinking from 230 to 130nm can be done successfully. It at least sounds sort of ambitious to me. However under the assumption it is feasable FAB25 could considerably increase output in Mbit-shipments next year again.
So the Mirrorbit-dream still lives on for a while. Just with the same questionmark as for Hammers: Can they make it true in time?
As for Intels pricing-strategy: For the quarter being, Intel concedes some 10 percent Flash-Revenue and marketshare - however if their expectation of demand exceeding capacities will prove to be right in the next couple of months, they will get the marketshares back with inventories nobody else has and still get the prices they looked for in Q1. A gamble Intel can afford to play - AMD in contrary depends on every Million of short-term Cash Flow in the financial situation they are; otherwise I guess they would have tried to balance prices and share-gains a lil bit - as they see some tightness in supply as well for the near future.
@facsnotfiction
IIRC P3 1,13 which was recalled was on 180nm.
The one released later on 130nm worked fine.
K.
Hi there
Just curious... three AMD-Boards?
how do you guys manage this??
Klaus