Seeking Alpha

H. Bruce Campbell

 
View as an RSS Feed
View H. Bruce Campbell's Comments BY TICKER:
Latest  |  Highest rated
  • Intel And Micron And Depreciated Fabs [View article]
    Russ articulated his popular views about this vigorously and in great detail in numerous prior articles. They're terrific reading, still timely, and readily available by clicking on his articles link of course.
    Aug 6 04:11 PM | 2 Likes Like |Link to Comment
  • A Fairy Tale For The Amusement Of Vringo Vs Google Et Al Followers. [View instapost]
    An entertaining comment, thanks! And may it prove to be true!
    Aug 6 08:56 AM | Likes Like |Link to Comment
  • USPTO Checkmates Google Over Vringo Patents [View article]
    If you might find a satirical story vaguely related to the Vringo vs. Google et al saga appealing, I composed a fairy tale for your amusement. It's in my blog here: http://bit.ly/11IsJek. It has precious little if any investor value - it's just an amusement for stressed out Vringo longs.

    All your responses make perfect sense (even though some conflict slightly with others), and I'm very grateful for the conversation. I am a bit of a reactionary by nature, and sometimes perceive more monsters under my bed than are actually there. However, though more rarely, I sometimes perceive fewer than are actually there...

    A substantial part of my alarm springs from the fact that the Obama administration is involved in this general issue at all. It seems to me that it's nothing like important enough to warrant his time or attention - there are very serious matters of international relationship dysfunction begging for more enlightened constructive engagement, very serious long term global environmental threats begging for charismatic leadership, a genuinely bankrupt economic system on the brink (from a long term perspective) of collapse and with no means of repair in sight, and a long list of other very, very serious matters which directly and deeply affect people's lives. The alleged 'Paten troll' issue simply doesn't rank - it's nothing like important enough to warrant his time and energy and distract his administration - not even close.

    So I agree with the basis of Sheldond's remark "I am not overly concerned with this currently as the president has many economical concerns that are far more important IMO." But it begs the question: Then why the devil is he involved in this general issue at all? It simply doesn't rank. And tampering with these issues isn't without repercussions - Korean business and government interests, though displaying some public restraint, are privately furious about his intervention in the Apple vs Samsung matter, and with very good reason. I'm an Apple product user and fan, but this was a matter for the ordinary bureaucracy and the courts. The fact that Obama was willing to create significant international friction with a good ally over this relative triviality sends my personal investing radar into overdrive - as a national leader, intervening in that squabble was wholly irrational even if he felt that the US International Trade Commission's action was deeply flawed. Obama's involvement suggests, in my view, either remarkably poor judgment or an unsavory hidden agenda.

    I also agree with Sheldon that "... but then again presidents try [to] stretch their power often." They do indeed. They live in a world which is ultimately disconnected from reality and with narcotic effect powers which ultimately corrupt. And political money is of course a prime behavior driver.

    I don't question that the President can't simply trump the courts or Congress. But presidential power has grown dramatically in recent decades, and even when the office was weaker Roosevelt found a way to confiscate privately held gold. That may be a bad example - the nation was under more dire stress then. But it demonstrates that presidents occasionally do shocking things. So while I can only speculate wildly, I can't shake the feeling that Obama has a plan to materially help his very generous buddy Google in this matter. He's been on a very visible 'Patent trolls are a destructive evil!' bandwagon ever since, metaphorically, Google began waving money in his face. (Actually I have no information about the timing of those events, but do believe they're intimately connected.)

    Extreme fast track 'Patent Troll' law pushed through Congress in trade for something the opposing party wants? Retroactive provisions with Vringo in mind? Bully pulpit rhetoric which crushes Vringo's image and bargaining position (we've already heard some of that)? Executive order to massively tax or uniquely regulate 'Patent Troll' firms? I'm not convinced that something couldn't be contrived...

    And every day Google engages in more delay and obfuscation tactics, I find speculation that they have a secret end run game plan in play more compelling. I suspect Google very deeply intends to crush Vringo by any means, and hasn't been just sitting idly by (ordinary and plainly visible trial machinations notwithstanding) waiting for Judge Jackson to rule against them. I'm suspicious of background plots which could deeply affect my investments...

    It's all pure monster under my bed class speculation. And I'm often just plain wrong - very often. But my nerves are rattled - my radar keeps beeping, as if there's more going on than power brokers care to reveal.

    And all the monsters under my bed are wholly and utterly indifferent about my precious investments - every penny can go down the swirl for all they care... I'm nervous...

    I use Blekko.com for search. Long vrng & vrngw
    Aug 5 06:36 PM | Likes Like |Link to Comment
  • USPTO Checkmates Google Over Vringo Patents [View article]
    I hope you're right, honest. But even though the image is rather foggy, my radar's giving me the willies...

    How about an executive order? I don't have a clue what limitations it has in modern times, so I don't know whether it'd be feasible. But if it were, I suspect there'd be precious little backlash. The vast majority of citizens have no clue what Vringo is, nor care about patent squabbles. And they won't connect any political contribution dots unless the press encourages them to do so, which strikes me as unlikely in this case. Congress may not much care either, and in any case is probably too dysfunctional to counter with much more than sporadic lip service.

    But I'm often just plain wrong, and I do hope you're right in this case - I'm not a fan of injustice.

    Sorry everyone, "Sometime justice wins." should have been "Sometimes justice wins."

    I use Blekko.com for search. Long vrng & vrngw.
    Aug 4 08:25 PM | Likes Like |Link to Comment
  • USPTO Checkmates Google Over Vringo Patents [View article]
    The million share Vringo purchase on 26 July was a very bright flare alerting everyone that a significant Vringo vs. Google et al event was imminent - it seems very likely to me that the buyer became privy to some tightly coy information, and used it to best advantage. I'm sure many of you suspected something substantial was afoot when you saw that purchase too.

    And a second bright flare may have been set off yesterday as well: http://bit.ly/1b5v45Y

    Much earlier this year, Obama said, for all practical purposes, 'I don't like Vringo'. Now of course he said "patent trolls" rather than Vringo directly, but we all know very well what was in his heart, and as he emerged from his meeting with Google officers who've filled his political party's coffers with exceptional generosity, he telegraphed it quite clearly to anyone who watches Vringo or Google closely. You could almost smell the stench of power brokers in action from Mars...

    In my estimation Obama is, at best, ambivalent toward Apple - they don't contribute much to his party, and power brokers resent rich entities who don't play ball with them, or do so at only a token level. But he needed political cover for an event to come, so he bit his lip and came to the aid of Apple. But that action wasn't about Apple at all...

    It was about his very generous buddy Google.

    He had to create political cover for later, when Google runs out of options, and it's time to provide a favor in return for Google's exceptional generosity.

    Shortly after Vringo's awarded a highly favorable ruling by Judge Jackson, Obama will act - he'll announce some action to nullify it. Frankly, I don't know what his administrative options are, but in my heart I feel almost certain that his administration will find some mechanism to turn a just Vringo victory into an unjust defeat, and do whatever they must to make it stick.

    His action in the Apple / Samsung matter was just a crass means to create political cover - to enable him to later claim that helping Apple wasn't about Apple, and helping Google wasn't about Google, and that he just hates to be involved in these matters at all, but he must act because a broken patent system is harming American industry so grievously that he simply has no choice but to intervene directly until the system's fixed.

    The Apple related action was just a transparently crass political cover ploy. It's nothing more than obfuscation to cover overt power brokering, and it stinks so offensively that it makes an outhouse smell like a rose by comparison.

    But what about us? Do power brokers care a whit about Vringo longs? Please...

    I've been very patiently and solidly long with Vringo for a long time, and with a lot of personal skin, and remain so, because I seek lucrative but honest returns on my investments, and it seemed obvious that Vringo had a chance to generate them for the simple reason that justice is very clearly on their side. But sadly, I can't ignore the flare which lit up the skies yesterday.

    I've made no final decision. But shortly after Judge Jackson rules - perhaps very shortly after - I might very well decide to jump ship. Because, let's face it, this matter isn't just a question of justice. It's also a question of power.

    And throughout the entire history of civilization, justice has struggled mightily against power. Sometime justice wins. But yesterday Obama sent script carving fireworks into the sky for all sharp eyed people to see. The glittering sparklers said "Not this time."

    It's bloody sad...

    That's just my sense of matters of course, and obviously speculative. I could be wrong. I am of course often just plain wrong. But I'll bet those of you who've walked this planet for a long time and have seen a lot of the disgusting side of mankind in action know that raw shenanigans are rather common, and suspect something quite nasty may very well be looming for Vringo and its investors.

    I hope I'm wrong. But the scent I'm detecting in the air didn't come from any rose...
    Aug 4 06:55 PM | Likes Like |Link to Comment
  • Micron: The Story Isn't Tech ... [View article]
    I don't know what the proportional magnitude of that sector is for Micron, but yes, they do address it:

    http://bit.ly/13J1hK6
    Aug 2 07:38 PM | 1 Like Like |Link to Comment
  • Micron: The Story Isn't Tech ... [View article]
    Exactly right in my view.

    In addition, I'm inclined to believe that in this particular case a credibility torch has passed from most financial firm analysts to this ad hoc group of investors. Most professional analysts simply don't possess the technical credentials of key members in this group, and thus can't accurately assess the proportional value of technical developments which can be key to a technology firm's future fortunes. With deference to a rare few highly respected analysts with genuine technical credentials, for the rest of them full disclosure should include a technical background disclaimer, perhaps akin to "I connect dots as best I can, but I have no idea which end of a soldering iron is which, much less how to don a bunny suit or operate a diffusion furnace. I'm not a nerd - I'm a Wall Street money man. So generally my 'insight' has to wait until a firm shows me the money."

    This group possesses a great deal of acumen over a broad range, including deep technical, financial, corporate logistics, international exposure, and, due to age related long experience, human element insight. It may simply be the superior judge of Micron's future fortunes (and Intel and other technology firms as well) - it's simply in a superior position to perceive significant developments at much earlier stages, when far more lucrative investments can be made.

    And thus far it's a uniquely uncontaminated group of respectful, insightful, enthusiastic, honest, and mature investors with a simple and utterly transparent agenda: Honest maximization of our personal investment returns.

    Professional analysts on the other hand - maybe often not so much. Because, purely speculating, and with this personal disclaimer: "I'm a nerd. I have no clue how to don an Armani business suit." That said:

    Frankly, I have no idea how broad institutional ownership of Micron is. If too few firms own too great a proportion, the temptation to manipulate the security, especially near key option expiration periods, might be overwhelming. If two or three institutions own most of the institutional holding, a wink or nod triggered "Let's do it together!" effort to shake out weak main street hands and undervalue expiring options may be pretty easy. And perhaps very safe too - I doubt the SEC has the resources necessary to discourage such activities. Again just speculating. But this planet isn't 100% populated by pure hearts.

    But many main street investors will never be privy to these realities. That's too bad, really - they'd be wise to take the batteries out of their remotes and spend more quality time reading honest conversations about their investments. That would benefit us too - if main street investors couldn't be rattled because they possessed a thorough understanding of the root story, the security couldn't be effectively manipulated, and we'd reap more fruit from of our hard won research and experience because we'd be able to hold tighter to margin loan limits and our options would expire at more honest common share prices.

    It's not as if it's expensive - this group's thoughts are utterly free to anyone who cares to invest a bit of reading time. Main street investors just need to take an interest...

    Complacency costs an inattentive main street investor their hard earned fortunes. And, in America, with ever increasing aggression in my judgment. And the ancillary effects force us to pay a price too. Alas...

    Oshaberi sumimasen...
    Aug 2 06:59 PM | 2 Likes Like |Link to Comment
  • Micron: The Story Isn't Tech ... [View article]
    Having secured Elpida, and likely already possessing impressive momentum for graceful and efficient corporate integration, hopefully Micron's executives will continue to explore for the next big opportunity, whether technical or business in nature. One mustn't sit on success of course...

    One already well under way is of course TSV and HMC, which in my novice estimation are very roughly as big a deal technically as the Elpida acquisition is strategically. So this statement in Steve Shen's article piqued my attention:

    "Micron's current plan is to use Elpida's expertise in DRAM development and to establish a critical DRAM research department within the Hiroshima plant."

    I suspect Micron and Intel have already extensively discussed and agreed to a general plan for planting TSV and HMC technology in the Hiroshima facility, and that it's already begun, or will soon. So perhaps by 2015 regular production of magically swift and energy efficient HMCs and will flow from Hiroshima. Or HMGCs (Hybrid Memory & GPU Cubes). Or SCs (System Cubes).

    (With TSV based chips for the other functional blocks coming from Intel, then wed into cubes in Hiroshima, or the reverse, with TSV based DRAM chips shipped from Hiroshima to Intel to be wed with other functional blocks into cubes in America.)

    Then perhaps a lot of those cubes will flow to Cupertino.

    Because: Is it just me, or do others share the feeling that tablet popularity will wane somewhat, whereas laptops, ever more compact, elegant, and efficient, sporting SSD only storage and HMC and Haswell lubricated lightning speed, superb power, and ultra stingy energy consumption, will rise in popularity. Because they'll have it all - they'll address all needs at very high performance levels, yet be nearly the same size and mass as tablets. When you want to accomplish serious work, you need a laptop. And when a laptop is only scarcely larger than a tablet, it'll dominate.

    I assume most of you prefer the other OS and thus hardware option, but as I compose this on my MacBook Pro, I suspect Apple's taking another crack at a big hit new laptop product, and that it's been in development for some time, separate from their normal laptop model evolution. Microsoft seems to think all cyber systems will become touch screen centric. I see no evidence Apple views the future that way - the Mac OS and iOS remain separate creatures. My sense is that Apple's apparent view of that aspect of the future is more realistic, and that laptops will remain mainstream.

    Whether Apple continues to lead in these fields or not, I suspect Intel and Micron view TSV as the foundation of the future, and it's natural to speculate that Hiroshima will play a key role in it. Offhand it seems like there's no better place for it, at least insofar as DRAM is concerned, and maybe cube assembly as well. Particularly if the Yen continues to decline, as seems very likely.

    To stretch the speculation further, one could imagine that Apple, in coy partnership with Intel and Micron, is currently developing a TSV based ultra high performance laptop whose size is only scarcely larger than its display, sporting at the very least HMC, or maybe entirely new functional combination cubes (such as an HMGC), or an SC. The SSD NAND will come from Micron Singapore. The TSV based DRAM or maybe complete cubes will come from Micron Hiroshima.

    If so, others will follow Apple with their own laptop competition. They could equip them with SSD NAND from Micron competitors. But for a time at least, there'd be only one source for TSV based DRAM. (And, to the joy of most of us, only one source for processor related cubes.) Perhaps for a considerable time...

    All just speculation of course. But I'm fond of my Micron common. And I'm very fond of my January 2014 Micron call holdings. And perhaps even more fond of my January 2015 Micron call holdings. And when even more distant calls appear, I'll update my shopping list.

    Because in my estimation Elpida, great as it is, isn't the only big Micron story. The TSV sequel might prove to be at least as exciting.

    And Hiroshima is a very nice place. There is a certain deep beauty to this...
    Aug 2 05:37 AM | 3 Likes Like |Link to Comment
  • Elpida Is Now Micron - Or Back To (From) The Future - Part 1 [View article]
    Ever since reading Russ's wonderfully insightful and thorough article I've felt this strange compulsion to rub my hind legs together...
    Jul 28 10:29 PM | Likes Like |Link to Comment
  • Intel: OK, Where Are The Apple 'A' Chips Really Going? [View article]
    "And me too of course!" That is, I'd love to hear from you about it too!

    TSV Enabled opportunities seem very exciting to me in general terms. But I don't have any direct monolithic design, fabrication, capex, nor other work experience with such firms. So I could very well miss critical technical problems (though I'm very skeptical that the oft repeated thermal management issue is a genuinely significant hurdle), and I might badly misjudge financial and corporate logistics issues.

    You bring direct experience, analytical skill, and your popular personality to the table, filling in a great many important blanks which I can only vaguely guess about. Almost everyone here has skin in this game, and in my estimation many would like to further explore how TSV might leverage their skin into more skin.

    And I'll bet you've already got a couple of ideas about how to make it clear in the article that it's not just about technology, but skin too...

    May Seeking Alpha judge wisely...
    Jul 17 04:51 PM | Likes Like |Link to Comment
  • Intel: OK, Where Are The Apple 'A' Chips Really Going? [View article]
    Hi Russ,

    No formality necessary for me - Bruce please. And "That was very well done." from Russ Fischer is a very highly prized trophy on my wall, thank you!

    A wall built in substantial measure by funds your experience, insight, exploratory heart, tireless energy, and community team spirit made possible. Again, thank you!

    The profits are very helpful of course. But ultimately the trophy holds the greatest meaning for me.

    Given the unique position Intel and Micron forged together for leveraging this important new technology, and its inherent and very substantial business implications, perhaps Seeking Alpha would approve such an article - maybe they'd perceive that it's intrinsically investment relevant. They certainly should. I'm sure the community here would love to hear from you about it. And me too of course!
    Jul 17 11:26 AM | 3 Likes Like |Link to Comment
  • Intel: OK, Where Are The Apple 'A' Chips Really Going? [View article]
    Just speculating wildly of course:

    Apple strikes me as either possessing a terrific secret rapport and loose partnership with Intel as Apple executes a terrific secret long term plan for a transition to Intel technology, or is simply adrift - at a loss for how to steer their corporate ship into the technical future. (Insofar as mobile products are concerned.)

    Intel possesses a superb but partially coy working relationship with Micron which will enable it to leverage TSV technology to its limits. If TSV and chip stacking prove to be relatively cheap and easy to fabricate, then Intel will be (or is) in a position to reallocate functional SoC blocks plus functions previously relegated to other die, like system RAM, into a very tight knit communities of cubes, in which every cube can communicate with every other cube at very near internal die speed and power levels, or even into a single SC (System Cube).

    At this time firms try to cram as much as possible into an SoC die. That's tough at 32 and 22 nm, and perhaps bloody tough at 14 nm. TSV might remove most of the pressure to put everything into a single SoC die. For example, the GPU might be reallocated to the HMC, making it an HMGC (Hybrid Memory & GPU Cube). The GPU would then be extremely intimate with system RAM, yielding speed and power conservation performance gains. Shared cache might be shared across cubes as well, or, if RAM access speed were sufficient, reduced to very modest size, leaving little more than processor cores on the SoC die. The processor could then be fabricated on a much smaller die, perhaps dramatically improving yields at challenging process nodes such as 14 nm.

    Near die level communication speed and efficiency within and between cubes bestows a great deal of functional block allocation freedom which simply didn't exist before. And that freedom dramatically transforms fabrication logistics, opening entirely new options. Intel might be in a position to allocate functional blocks to separate die with almost no performance loss, thus allowing them far more options to optimize yields at challenging fabrication nodes.

    If TSV can be fabricated efficiently and consumes only modest die real estate, to me the implications are profound. But a partnership with a robust and technically advanced memory firm is critical. Intel possesses that with Micron of course. To me that's huge. So far as I'm aware, no other firm has anything even remotely like such a partnership except Samsung, which has both fabrication technologies in house of course.

    If Apple's not deeply involved in this, they're desperately adrift in my estimation. But Apple would be crazy to allow any hint of such a long term plan to leak to the outside world. And given their history of rather severe angst with Samsung and Google, my guess is that everyone at Apple is thoroughly wedded to a corporate culture of absolute secrecy in such matters - every soul at Apple understands the critical sensitivity of long term core technology plans as the firm tries to protect its innovations from the likes of Samsung and Google. The pressure is especially great because Samsung is in a position to wed processors with HMC if they're technically astute and forward looking enough. So no American firm wants any hint of custom developments in this area (outside the HMC Consortium) to leak to Samsung.

    And Intel knows this too, and thoroughly respects it, as does Apple respect Intel's secrets.

    For many mobile products a single SoC cube incorporating all significant functions - a mini 'Smoking Hairy Golf Ball' of sorts - a single cube of processor cores, GPU, RAM, radios, and perhaps more, might prove to be the highest performance and most die and product fabrication efficient means to implement primary silicon functionality in the product.

    TSV may become (or already is) truly efficient. I suspect that's inevitable because of the dramatic functional block reallocation options it opens, and thus the dramatic die fabrication options it opens, plus the performance gains it enables (especially in RAM access speed and efficiency) for CS's (Cube Systems).

    If Mother Nature isn't blocking this path, in my estimation Intel and Micron are pursuing it extremely aggressively. And with as much secrecy as possible. If Mother Nature is blocking this path with TSV size or fabrication hurdles which are simply too daunting, then precious little of my speculation has any merit. Otherwise it seems to me that we're looking at a cube intensive future. Which Intel and Micron seek to dominate.

    And Apple would be nuts and truly adrift to be on the outside of that evolution. And TMSC and their peers have no Micron like partnership to enable such a future. Samsung is the wild card and sole threat. And needless to say, Apple sure as blazes won't wed with them for any such future.

    That leaves Intel and Micron. To me, the writing's on the wall. Or maybe etched deeply into the wall...

    All just my personal wild speculation of course. Intensively long Micron common and calls.
    Jul 17 08:36 AM | 5 Likes Like |Link to Comment
  • Micron: The BIG One For Wednesday [View article]
    The *C (Anything Cube) and CF (Cube Farm), in my rambling boring opinion:

    I've never believed any human being is willing, ultimately, to settle for 'enough'. Eventually we always want more. Maybe if we were immortal creatures it'd be different. But we're not, so we seek to fill our short lives with new excitement every day. 'Enough' might be perfectly practical. But it's not exciting, so in the end, rich or poor, we seek more.

    That sentiment drives every nuance of life, including material life. We always want more. And when we get it, we become dependent upon it by configuring the rest of our lives tightly around it. We never stop pushing the envelope. And no product is immune (including phones).

    Cyber hardware gets more powerful and feature rich. Exciting! More elaborate new software and networking is created to take full advantage, giving us more functionality. Exciting! But that wipes out the hardware performance margin just achieved. So there's pressure for another round of hardware advancement, and thus it's created. Exciting! Once achieved, yet more elaborate new software and networking's created to take full advantage, yada yada yada, an infinitum. All competitors on this stage are in a constant advancement battle of survival. (Some by honorable means, some not quite so much...)

    Okay, forgive me for stating the obvious - we're all perfectly well aware of that of course - we completely understood that for processor chip firms (and others), constant aggressive design and fabrication advancement is a mandatory matter of survival.

    At any given node, such as 14 nM for example, it seems to me that a critical design factor lies in optimally proportioning functionality among the processor and its ancillary functions, such as RAM and GPU for example. If more functionality is left to ancillary chips, the CPU is smaller and simpler, so more fit on a wafer, and yield is almost certainly higher as well, and thus they'll cost less. And perhaps some on chip performance advantages can be achieved due to generally closer spacing of the CPU's remaining functional blocks, and other factors. But all this comes at a serious loss of performance in those functions which were pushed off chip, because that functionality is now way, way out yonder on the horizon, and thus hideously time distant, relatively speaking.

    So the on chip / off chip proportioning optimization dance has favored lots of on chip functionality lately, even though large chip sizes and thus low chip count per wafer, and yield penalties on top of that, exact serious cost pain. Thus the age of the SOC.

    And I get the sense that a lot of the industry, including end product OEMs who purchase processors, are essentially convinced that that's the inevitable direction of evolution of the technology, even including the very many who've focused on better matching their less demanding product niche, like phones, with more suitably scaled processors. They seek to match CPU cost and performance with their product as wisely as possible, but in their technical hearts they still feel that ever higher SOC integration levels are the natural and inevitable path of evolution for the industry.

    But what happens if they can suddenly purchase a significantly higher performance new SOC black box which costs less, but when they peer inside they see that it very rudely flies in the face of their sense of wise levels of integration? Well, they suck it up, abandon their preconceived notions, and they buy it for their next generation products. Because first and foremost they seek to survive.

    The question is, does TSV and HMC lead to that? When ancillary chips are snuggled right next door in time rather than way out yonder on the horizon somewhere, traditional thinking about how much functionality should be integrated on the processor chip for optimum overall system performance (including component price performance) changes.

    Simply stated, I suspect Intel's very keenly aware of this. And Micron too. I'm not convinced ARM, Qualcomm, and many others quite get it yet. Or they've just failed to vigorously prepare for it. Wildly speculating, I suspect Intel's very quietly preparing modified versions of Haswell and other processors which are fully optimized to incorporate very tight HMC. Though there's not the slightest outside evidence of it, maybe the GPU will be stacked within the HMC too, where it'd be exquisitely intimate with RAM, and thus could achieve superb performance at dirt cheap cost. And leave a far simpler and smaller CPU which will pop out of wafers at eye popping density, and thus also be dirt cheap.

    I'm not privy to a great many of the design and fabrication nuances involved in this arena, and my ignorance might very well be glaring brilliantly at this moment for all to clearly see. But from my outside perspective, TSV and *C (Anything Cube) change everything about how to optimally allocate silicon real estate for best system performance, fab yields, and overall cost. It's just wild speculation, but I suspect Intel and Micron are coyly preparing for the day when they'll spring the technology onto the marketplace rather abruptly.

    There'd be a very cool presentation, but in understated style, because the dramatic performance and cost numbers would glow brilliantly on their own. Near the end Intel and Micron execs would say, "And that, ladies and gentlemen, is our first nascent 'Smoking Hairy Golf Ball'. It's a substantial step forward. But it's just an initial glimpse of a much larger future."

    In my estimation the days of "Well, you just have to keep cramming more and more and more onto the SOC." may be over, at least for a while, replaced by "Well, you just have to keep cramming more and more and more into the CF (Cube Farm).

    But that doesn't seem to be popularly recognized yet, even among many refined inside corporate technologists. Or, of course, I'm just all wet. As I very often am...

    Where's Apple in all this? I just don't know. My very wild guess is that they're in the thick of it, but will reveal no hints. In my wildest spy novel moods I imagine their visible relationship with TSMC is partly just a backup supplier arrangement, and partly a cover story to help keep a far more dramatic development better obscured. The entire Apple culture might be convinced of the critical importance of secrecy now. Everyone at the firm knows what Google and Samsung did to their iPhone innovation - even the mail clerks may genuinely embrace, from the heart, that loose lips really do sink ships.

    And therein lies an example of our eternal craving for excitement. I have zero evidence of such an Apple posture. It's just exciting to think it might be true, like a good industrial intrigue novel...

    Things change. If I've got enough of this about right, it seems to me that some pretty cool advancements are around the corner. To the delight of excitement craving consumers. And greed excitement craving Micron and Intel investors, like me.

    But my engineering experience wasn't focused in this arena, so I'm just speculating on very shaky ground. I guess we'll see...

    Oshaberi sumimasen...
    Jul 12 02:29 PM | 2 Likes Like |Link to Comment
  • Intel Could Be Hiding Something Huge [View article]
    I disagree with a couple of points in this specific discussion branch. My sense is that the memory wall is a significant performance bottleneck for even relatively pedestrian systems such as tablets and laptops for example.

    However, the overall line of thinking in the discussion makes sense to me. Specifically I thematically agree with Fidgewinkle that "They probably want to take some of the ever increasing cache off of the CPU die to better manage yield.", although I don't necessarily view it as just a yield centric issue.

    Regarding HMC and processor cache design, presumably there are lots of superb analysis of cost / benefit tradeoffs for on chip cache size for each significant type of processor (such as those for phones, tablets, laptops, desk tops, and servers) when the overall product utilizes traditional RAM. It seems to me that HMC slants all those guidelines toward smaller cache sizes, since HMC RAM is much less time distant from the SOC than traditional RAM - less RAM interaction latency suggests a reduced cache size for optimum cost / benefit tradeoffs.

    In the case of modest performance products such as phones, where the HMC is likely to be exceptionally close to the SOC anyway, I'd think the need for on chip cache would be minimal, allowing dedication of much more actual SOC silicon real estate for other functions, or to simply reduce chip size and thus cost. Perhaps rather dramatically so. (But I don't know how size and speed critical the SOC's GPU's memory needs are.)

    Just guessing extremely roughly, maybe there'd be about 2 cm separating the SOC and the HMC in a phone. And while normally I'd assume a practical signal propagation speed of about 17 cm / nS, in the case of a close proximity HMC maybe just half that, about 8 cm / nS, due to unusually high predominance of irregular transmission impedances, since a high proportion of the short distance involved might be mechanically complex and irregular.

    But still, that suggests that HMC is only about 500 pS distant round trip. But add delays for the HMC system itself to manage data flow. But we're still in the very rough neighborhood of just one clock cycle for a 1 GHz processor. And maybe only half that, because the HMC / SOC separation distance might be only 1 cm or even less, and maybe the signal paths can be made relatively clean and thus propagation speed closer to 17 cm / nS.

    So with such quick access to HMC RAM, I wonder whether the cost / benefit of on chip cache falls dramatically to a very modest cache size. Thus freeing a lot of silicon real estate for other functions, or to reduce chip size, and thus fabrication cost.

    I don't know how much real estate Haswell allocates to shared cache, but I'll wild guess 15%. Maybe competitive processors allocate about the same amount.

    If you could eliminate or reallocate most of that real estate yet suffer no overall system performance degradation (or perhaps gain overall performance, including power conservation), and if ARM, Qualcom, et. al. couldn't do that, I'd think you'd simply have a superior product, and capture a lot of extra market share. Good is never good enough after all - everyone wants the best. And ARM, Qualcom, et. al. might find themselves offering only second best, possibly in all three categories - effective speed, power consumption, and fabrication cost. And possibly by significant margins.

    Intel's not a member of the HMC consortium. Whether Micron serves as their proxy might just be a question of semantics. I simply view Intel as disinterested in being captive to the consortium's specs. From their perspective it makes far more competitive sense to, in tight partnership with Micron, and sans any HMC v1.0 specification constraints, develop custom HMC / SOC products for all categories of products, then sell those to OEMs at great competitive advantage compared to ARM, Qualcom, et. al. options.

    I'm a bit suspicious that competitors, perhaps already well behind Intel in incorporating HMC logistics into their processor designs, will as a practical matter be forced to honor the HMC consortium's standard because they won't be able to fund custom HMC design. They don't have anything even remotely like an Intel / Micron partnership, and aren't likely to have anything similar for the foreseeable future.

    If all that's very roughly correct, it implies some very happy days ahead for Intel investors.

    However, I'm just speculating as a general circuit design EE - I don't really know enough about the esoteric details of processor design to feel terribly confident in my appraisal. But we do all know what the speed of light is - there are some immutable facts here. And we also know that Micron and Intel have a very close and effective relationship, and that Intel isn't a member of the HMC consortium.

    And we know that ARM, Qualcom et. al. don't have a Micron like buddy. To me, with HMC nearing implementation (in my estimation), that's huge.

    I still have lots of questions. Technically, I don't know how critical the SOC's GPU's timing to memory is, and thus whether HMC can reasonably meet its needs. That could limit how much cache could reasonably be eliminated.

    And I don't know what the devil Apple, also very conspicuously absent from the HMC consortium, is thinking or doing. But Apple does owe Intel a favor, and is understandably very secretive. So it's only natural to wonder whether Apple and Intel are close but extremely coy development partners on an important project. The TSMC / Apple relationship notwithstanding...
    Jul 11 07:35 AM | 3 Likes Like |Link to Comment
  • Micron: The BIG One For Wednesday [View article]
    Oops, my sincere apologies. Again... 78%, not 90%. Babbling incoherently:

    As I recovered from abdominal surgery in Miyazaki, a roommate and I christened ourselves 'Oshikko kyodai', peeing brothers. Urethra catheters leave a bit of tissue damage which makes urination quite an adventure in pain, and we bonded over the common experience. It sounds improbable, but my entire surgical experience was a uniquely cherished time in my life which I remember with immense fondness. All there were, and many still are, like intimate and beloved family.

    This is very, very nice too! So much skill and good will among good friends. And no pain anywhere, other than the inevitable occasional angst of worry of course. And, so far, money rolling in. We'll all celebrate together, or some black swan will rain on our parade together, as kabushiki kyodai (stock brothers).

    I relinquished part of my already quite modest Intel holding Friday to free cash to place an order for a block of 17 August 2013 $15 Micron calls, but only about 25% of the order executed at my limit price then. Today did the trick - the remainder of the order executed. I missed the bottom by several percent, but I'm very happy to have acquired them. Besides, lucky Intel sale timing. (I still suspect Intel will do very well soon though.)

    My hope is that there'll be a bit of a viral news effect when the Elpida acquisition is formally announced as complete. Maybe analysts will be motivated by the announcement to crunch fresh numbers for Micron, be impressed, and effuse some excitement. (Maybe some TV blowhard will even do cartwheels over Micron, who knows...) Maybe the news will snowball enough to cause Micron common to surge a bit. It's just a raw gamble of course, but it looks favorably asymmetrical to my star struck eyes, especially with Fred's news about Inotera - wow, 24% sales growth MoM. That's got to impress analysts, and juice their wonder about the implications for the other fabs. I'll bet I'm not alone with this bet - I seem to remember Russ saying he's playing it, and I suspect others too - maybe Fred, Seeker_2, and Terry S5492 and I are all 17 August 2013 Micron call kyodai.

    Yes indeed, fingers and toes crossed. I hope we can all roll in rewards later, and that the IRS or equivalent inexplicably forgets we exist. But even if not, we know we played a good game, and the sweetest things in life will still be the sweetest things in life.

    Sorry, "black swan's" should have been "black swans" of course. The lesser of my sins captured the greater of my focus, a big priorities mistake. In my excitement, I completely forgot about another modestly significant holding I'm waiting with extreme patience to play out, and my IRA account. You'd think I'd check such a dramatic key number before publishing it, instead of worrying about "black swan's" class typos.

    So to set the record straight, my Micron holdings are, as of NY closing today, about 78% of my grand total securities holdings, including my IRA account (which sadly has no Micron - it's stuck in that extreme patience play, alas).

    I'm plenty embarrassed to have made such a drama queen class dumb mistake, especially to this group. I make so many mistakes... It's a wonder I still have all my body parts after over six decades of stumbling over my own feet...

    I see that Hawk Glen cashed out 64,750 shares from (I presume) calls last Friday. That doesn't strike me as significant, though if I were him, I'd have waited at least until next month if allowed. But his name is still Hawk. Bruce is okay - I've grown fond of it. But Hawk is pretty darn cool, whether stock transactions are mistimed or not.

    Again, sorry - I'll try to be a bit more careful about key numbers next time. And sorry to babble...
    Jul 8 04:27 PM | Likes Like |Link to Comment
COMMENTS STATS
171 Comments
169 Likes