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.
Keith
While I believe the article is on the spot wrt to OrNand, the spin on aggressive pricing pressure from Intel is probably much overdone. I have seen estimates that AMD only lost about one percent of mss to Intel in the handset part of the business, but five percent in the embedded space to others. No isupply figures out yet for last quarter afaik, and no second source to confirm these estimates yet. But what Intel and AMD said in their calls about the markets fits into this picture.
K.
mmoy
Prepare for all sorts of such stickers: 64-bit-high precision joysticks, 64-bit high definition printers, etc. etc.
I'd guess even in my wildest fantasies i could to predict what the oddest one will be. Gonna be amusing, anyway. Somebody should make a contest of it. A nice opportunity for equipment makers to make a new carton and sell their old lamps for a premium with a 64-bit driver.
K.
upndown
Some will complain about drivers for legacy peripherals, but the dealer won't be on the hook to support those.
Yeah. Dealers will have "64-bit"-periphery ready to make another hundred dollar or so when people come back and complain.
K.
Buggi
First retail offerings of new boards are usually from pilot production to cream off enthusiast premiums. OEM-quantities not before eight weeks later. Visible in retail as well by significant price drops for these boards. As I said, appropriate CPU shipments could happen in this quarter already - optimistically.
K.
Joe
Yes. Demand in terms of migration from K7 to K8 was not shifting at the pace AMD migrated packaging facilities in the last couple of months imo, creating an imbalance in the mid single digit unit percentage only - in revenues in the high single or even low two-digits though. Halting migration of packaging facilities for a while will solve the problem during this quarter, optimistically, or in Q2, conservatively.
Keith' lamenti about lack of K8 socket wins in the last quarter were obviously justified. If i understand him right, he blames the platform-side for it, among other reasons. From this angle, we gotto watch availability of VIA890, nForce4 and ATI-boards as indicators for the probable inflexion point of K7/K8 units. From this perspective, considering the first month of the quarter is nearly over already, I am leaning towards Q2 currently. Might change if a flood of boards rolls ashore next month.
K.
cd
Indeed I do. Do I miss something?
K.
Mike,
thanks for posting again. Although it is frustrating to see AMD being about to become trapped again in the value cage it has been in for years. Finished goods entry level Semprons inventories have been dumped end of last quarter to show at least 730M. Apparently packaging facilities for K7 are fully utilized for OEM commitments currently. Demand in the middle part below expectations (pricecuts last week), and top-end is supply constrained. Not exactly what I wished to see.
But well, maybe I am seeing just ghosts.
K.
wbmw
Well. Reading through your post I see our assessments of what is and what is to come from either side do not differ materially at all, it is basically just couple details here and there that we look at from slightly different angles.
I do agree in this as well:
In terms of a technology path, I don't see anything new for AMD. There are no new standards on the horizon, save for a nebulous thing AMD is calling "security and virtualization", but at least the virtualization part is late and Intel will likely set the standard with VT. In terms of anything else, I don't see anything from AMD.
Now, the angle I always have been looking at it is that if the market accepts AMD64 it would change the competitive landscape even without AMD delivering something "new", just from the ecosystem synergies to be opened up over time. Server designs, Compilers, Operating systems and application optimizations etc. etc. I see all this coming at we speak. Not yet really obvious for many, just because nothing really "new" is happening - still below the line, so to speak. Second, there is many opportunities within the architecture to be realized without making anything "new": Glueless dualcore, SSE3, working the power down are just beginnings, dual memory-controllers and coherent cache to follow, further system-integration (e.g. bring RAM on-chip) is next - and there is many many more - Hector's "u ain't see nothing yet" is about the essence of all that. But again, looking at the list, this is all nothing which would be perceived "new" in the markets - except for the few savvy ones out there. What I expected in this respect is a perpetuous followup on Hector's "innovating for custumers benefits, not for technology's sake" in terms of balloons, bladders and gimmicks. Did not happen. For the reasons below.
In essence: I do acknowlege your point. Sadly, the likely outcome is that AMD will continue to win awards and loose money at the same time.
Ironically, I see AMDs management-team being somewhat aware of the dimension of some gap for quite a while now (e.g. custumer-advisory board, customer centric approach). "Somewhat", because awareness still appears to be unshaped, or halfbaked if you want. More precisely put, on some occasions i had a strong impression folks are presenting some consultant's foils without really having fully internalized its underlying content in the dimension it was meant (which is more the rule than the exception anyway if the content is in a dimensional context your clients are not used to think in). Insofar, it is not surprizing I don't see any indication yet of cognition what the missing link would be to bridge the gap.
K.
wbmw
The competitive situation between Intel and AMD, however, may change dramatically if AMD gets to 65nm a year later than Intel.
AMD always was a year behind Intel in process shrinks as long as I can think of. From my vantagepoint, this fact is somewhat more worrying for Intel than for AMD - here is why: For decades, shrinks meant diesize-economies plus progresses in performance within roughly the same power envelope. This competitive advantage is part of Intels business model. It probably will work again for the Pentium-M architecture in 65nm but is very doubtful for Netburst. Going multicore and reducing clockspeed is only efficient for multithreaded application - whose codebase is still scarce. What I am saying is 65nm could easily be hitting a wall if Intel is unable to push Dualcore down everybody's throat. (But I believe it will).
AMD has very reasonable diesizes at 90nm - and is entering its 300mm fab with a process they know already pretty good - which bodes well for initial yields - iaw they take are ramping in Fab36 with a two years learningcurve of volume-manufacturing to Fab36 and Chartered. Low risk profile and probably not much less output than from a juvenile 65nm process. But they can't go fullsteam into dualcores with it. And ramping up a fab takes time. I don't expect much volume from Fab36 and Chartered already in 2006. Both 300mm fabs together will only exceed output of Fab30 in 07, it my model is worth a dime.
As for Smithfield, its a twincore design. Two Prescott-dies. http://www.siliconinvestor.com/readmsg.aspx?msgid=20970302
Not sure AMD will gain mss in 05. I see the snowball rolling slower now than it rolled three months ago. Handicapped in the ecosystem. Dumped valueparts again and still lost mss last quarter. Taking flash into the equation i see them have a hard run to make any money at all this year. But I still see them on the superior technology path. AMD64 is continuing to penetrate the markets, execution is solid, capactity is on the way.
K.
wbmw
As for what you see, yes, we are looking at the same roadmap obvously.
As for conclusions:
1. Not for me. I think it is commonly known I did not expect an initial 65nm volume ramp in Fab36 and Chartered, but second generation 90nm. The recent disclosure that Cell will be fabbed on 90nm instead of 65nm as previously expected is only one more supporting tidbit for it.
2. Not sure if it is DDR-2 that requires a new socket. Anyway, I would expect AMD to make LGA as well. And I believe there will be Socket 939 packages (and probably 754 as well) for quite some time, probably for as long as it takes to clear out inventorized dies of Fab30.
3. Certainly. For several reasons. First, Intel is stretching the term Dual Core anyway. AMD would have less difficulties to make a lot more if they chose Intels way of understanding the term. Second, capacity. AMD only has one 200mm fab, limiting its abilities of churning out Dualcore-dies significantly.
4. I believe AMD is pushing technology as hard as Intel. Just not in the same direction. And I am not sure if VT will be big yet.
But in summary, I agree on the conclusion of diminishing leadership from looking at this roadmap. Actually even stronger: it clearly shows AMD is following the big dog again.
AMDs control of its own fate seems to be perceived currently as being not yet as complete as it appeared a while ago.
K.
AMD Roadmap
www.tweakers.net/ext/f/49970/full.gif
K.
deleted duplicate
deleted duplicate
jj
Item one is highly appreciated.
As for item 2, suggest to leave it as it is currently.
I would indeed expect the announced changes of structural organization will be followed by corresponding changes of reporting. Not sure it will apply for the current quarter already, maybe rather in Pauls first quarter at the helm. (Unless current reporting format would seem to be "inappropriate " for serving q1 results wrt to the company's (best ;)) interest, that is, which would "require a change of report format to reflect the changes in the company for better transparency for shareholders and financial analysts" - along this line.
K.
cd
I am assuming Smithfield will ship for reviews together with at least one popualar game in multithreaded code producing shining bechmarks on Smithfield. Add some apps like Photoshop, CorelDraw which are known to make use of a second core already, plus at least one video-application. These are the figures public attention will be directed to.
As I said, I would be delighted to post about a world of educated buyers, sophisticated opinion leaders and an independent press.
From another angle, i believe any attempt to explain enthusiast's buying decisions (and opinions, the latter being far more important in business-terms) with rationale would not be capable to come even close to the crucial point.
I'd suggest you take half an hour or so and talk to one of the enthusiast/opinion leaders inclined to AMD. Let him explain why he opts for AMD. Then scratch the surface of his arguments. In case you don't know one of these chaps in your neigbourhood you can do that online as well. Log yourself into an AMD enthusiast site as Melissa or so and play dumb, saying you want to know which computer you should buy - and why. Whatever approach you take for fieldwork, it's an interesting experience - making clear these chaps are far more loaded with conviction than with sophistication. There is a reason I suggest fieldwork - see below.
Now from an empirical angle: Couple years back, when P4 was launched, I thought exactly what you stated today: High end consumers, enthusiasts, sophisticated, if even I see the architecture is crap its only a question of weeks until the writing is on the wall. Now, we both know how this one turned out. The takeaway from this experience is I had to acknowledge people are not where I thought they were in this particular respect. So obviously you gotta pick them up somewhere else. Before you can do that you gotta find out where they are. While I have seen couple of approaches into this field from AMD (eg. Customer advisory board) and I see a trend towards the right direction, the way I see AMD moving here never gave me an appeal of excellence. Iaw, I suspect there is a lot of groundwork to do. Nothing that can be done in a short-term horizon. If i am anywhere close in this assessment it is btw no miracle AMDs marketing initiatives remain disappointingly unsuccessful.
Where I see excellence in this field is in Santa Clara - that is because I understand their products (the term to be understood strictly in marketing-terms) as projections of the outcome of their marketing-research (this is what I am really talking about). They've been on the spot with Megahurtz, Centrino. I expect platformization and Smithfield to be on the mark as well, because these might be founded on the very same solid ground of marketing-research.
Circling back to the origin of these musings, whether I like it or not, i am afraid i have to get used to the idea people out there are indeed where Intel aims its offerings at.
Using the generic concept of marketing paradigm of value exchange, you probably could extract some simplified perceived benefits of Smithfield (meant just as examples to point out the dimension):
Dualcore delivers superiority. You can tell (and brag) about having two cores instead of one like everybody else.
Dualcore delivers attention. Your peers will be interested and ask you about what you have.
....
K.
DURL
Tx. Actually, its one of these where I would be happy to be proven utterly wrong. :)
K.
Paul
While I can't answer for Keith, it is probably irrelevant. Who cares about details.
Two CPUs instead of one is what people out there will understand.
Buy one get two is what people are familiar with. It's easy - and it will work. People will feel proud annd strong with a two-processor machine. Perceived value is what sells, not value.
I mean you have seen Megahurtz work - this one is a lot more compelling.
It'll work. And it will make a huge dent into the consumer high end segment AMD owned for a while.
K
mmoy
Got it. Not sure HPQ will take the pain to offer a Win64-system for these ones. I wouldn't.
K.
Chris
Naah, u did not miss it. Delayed for business reasons, i've been told.
K.
mmoy
Sure. However, wouldn't Server 2003 SP-64 and Opteron 1xx be the combination of choice for a workstation-type of product?
Or do you think more of the wannabe professional users?
K.
mmoy
There is no retail act for MS to get together wrt Win64. These chaps do know exactly why they make it an OEM edition only.
Its the reasons you pointed out in your post - plus driver issues.
K.
mmoy
If you have the wrong hardware, sometimes it's very hard.
And in some cases even close to impossible. Say, if your notebook-vendor does not supply a supporting bios, it is impossible unless you take the chance and flash some bios from elsewhere which could make it work. With a fair chance to run into deep sh!t if it is going awry and your notebook won't even post.
K.
jayxx
Thanks very much for hosting the contest again - and all the improvements to the site you made in the last quarter.
Congratulations to all winners.
Klaus
mmoy
Thanks. Interesting. Because it fits my fundamental gutfeeling since yesterday (not looking at technicals at all).
FF version Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7.5) Gecko/20041107 Firefox/1.0
K.
mmoy
Just to reconfirm, are you making a technical call for five bucks something again for the common from the inverted flag?
K.
btw, got a strange issue with FF again: It does not store downloads in the directory i specify, and loses these when closing FF. Ever heard of this one? Got a suggestion?
Keith
When I read Kevin's remarks last automn I had a stong impression of authenticity.
I cant help what I read today gives me the impression of a puppet talking balloons.
However the message is crystal-clear.
K.
Gothik
Tx for the elaborations. However I am not sure it is really necessary to have the books on RAM: When I use chess-programs together with chessbase-books (which are indeed large) the latter are on disk. During analysis access to the books on disk does not seem to be the bottleneck.
K.
buggi
Guess u gotta ask wporsche to to that.
K.
gothic
I did it on an Athlon64 notebook (1,8GH, 1MB L2, 512MB 333MHz DDRRam). Got 205900 nodes/sec.
64MB is the most hash memory selection.
Just curious, what am I missing? While you blame MS for not being 64-bit capable and look for 64GB memory machines, 64MB of hash mem is the most your prog can use now?
Thanks for elaborations in advance.
K.
drjohn and aleph
What was the most amount of ram you selected for performing the test?
K.
Keith
I don't see manufacturing of memory as a viable (let alone ideal) N plus one solution going forward for obsolete CPU-fabs.
And I agree in your assessment of Spansion and implied conclusion. As I said earlier, I would apprecitate AMD to exit its Flash business - if there is an opportunity to do so. Risk/reward considerations are fine if you have options to choose from. I am not sure AMD really has many options currently wrt an exit of flash - i would be happy to be proven wrong on Tuesday already.
Insofar, look at my previous post as just to paint the worst case with some lighter color to make visible it is not as scary as many believe.
K.
cd
U r right. Thanks for setting it straight.
K.
Keith
Spansion could indeed loose half a billion this year if it goes bad. Makes it 300M for AMD after short interest. I am confident CPG can make up for 300M this year, although it has to carry the burden of a 300mm fab in the works and some costs in Singapore as well. That is why I believe AMD can keep its numbers together this time.
K.
Keith
Well, AMD's earnings exposure was probably closer to 20 than to 25 percent last quarter already (after minority interest). If Flash will go even worse for the next half year or so, exposure will likely drop into the teens percentagewise. It still hurts badly, because it damages the sprout of confidence in AMDs solidity grown in 2004.
K.
Combjelly
Good you pointed out why nine month numbers are published.
Its getting tough in the numbers game. Disinformation at work.
More to come.
K.
Comjelly
I see. However, your posts helped me to understand better what is going on in the notebook grafics field. Many thanks.
K.
Combjelly
Thanks. Just to reconfirm: Are GeForce 420/440 Go single chip solutions in the envelope of 6 Watts or so?
Tx
K.
Combjelly
Interesting. I did not know about the single chip discrete grafics controllers from nVidia. And yes, anything up to 5 Watts or so can be operated with passive cooling afaik.
Thanks for enlighening me on this. It also would deliver an explanation for nVidias manufacturing model for K8 chipsets - although i have not seen it in combination with one of the chips you mentioned.
K.
Keith
Never saw one of these tadpole notebooks out there. It's not even listed on Tadpole's current notebook offerings page as well.
A dog died a sudden death I guess.
K.