You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
It looks like there might be serious trouble on the horizon in Intel either failing or seriously stumbling in their 3dXpoint memory. This has been a costly effort for both Intel and Micron and if it partially or totally fails will have serious ramifications for both. The silence has been deafening in this area with no real clues. The various possibilities are serious design flaws that cannot be easily corrected or manufacturing problems on a similar scale. It is obviously something very serious and being dealt with at the highest levels since so much is as stake for both companies. It already must be a problem, for both product and information have been pushed out into the future with no clues given. This could be temporary and still a huge win for both Intel and Micron or turn into a very expensive disaster. This would be a bigger disaster for Intel than Micron for all the problems they have been having, especially with TSMC literally blowing past them in not only market cap, but technology(and that's only technology we know about or can see). A wall of silence which is the current strategy it seems, is the worst solution for it leaves a cloud of uncertainty, which is the worst possible option of all. Hopefully we will hear shortly from either party and this would be the best way of dealing with any outcome, win, loose or draw. This is just but one of the many serious problems Intel faces.
Addition: When you add in Microsoft using ARM, Nvidia and AMD attacking processors, and other companies attacking their market like piranha, Intel's problems look even more serious for this indicates the successful Andy Grove culture is either gone or severely diminished.
Intel will more on from Washington County and Maricopa County. Intel will be fine. Intel at 10 nm will still have the best architecture and the best technology.
Thanks for sharing your thoughts. To be clear, I don't know where things stand so I am just speculating. That being said, my guess is that if there are serious issues you will see the first public repositioning from Micron. Sanjay M., as the new CEO, won't hesitate to get ahead of this and reposition Micron, thereby reseting expectations regarding this technology. This may come as soon as this week on their Q3 concall. Micron is expected to beat expectations and guide aggressively for Q4 so slipping in some bad news on XP might be the best possible way to handle it.
The northwest isn't developing the engineers and talent it should and they may not be getting the support they should in the private and govt sector. I can't recall a single coherent product that came from Oregon's campus at Intel alone. In this era of ecommerce and search engines it's a disaster for the state.
Portland, thanks for the insight. I believe with all tech advancing so quickly and accelerating, internal training will become a major competitive advantage for those companies that master it. With technology advancing so fast, external education will become less of a factor than companies that have their own human resources development program. I feel the standard college track of developing talent might become obsolete in the not to distant future.
I always wondered how engineering and comp sci programs in the PNW were so weak considering all the tech companies not just in Portland but also Seattle.
I always wondered how engineering and comp sci programs in the PNW were so weak considering all the tech companies not just in Portland but also Seattle.
It was interesting to see all of the students at #54DAC this year. Very few however were American. Other countries around the world are pushing science much more aggressively. Right now computer science is very hot but that is for application programming (social media). Google is paying CS grads $150k while EEs are closer to $100k here in Silicon Valley. I realize it is a supply and demand issue but without EEs we wouldn't have the need for so many CS grads.
Arthur, would you mind sharing why you think "it is obviously something very serious [with Optane]". You seem to take it as a fact, but I have heard nothing about it so far and I do not know what I am missing (your other post about this topic did not shed any light either).
Yes, the higher-end SKUs (P4800X) are almost impossible to find in the channel, but if they are so good then the hyperscalers could be taking in as many as Intel can make them. Then, I also hear it is reasonably easy to buy as part of complete systems/servers from the major brands, so it is not as if it had been pulled from the market.
The consumer SKUs I can easily find - but I still would not say they are widely available because many large retailers have it for pre-order only.
mbello, the reason I think there may be a problem with 3dXpoint is that when I do a search on deliveries, which Intel said should be in full production now is that I can't find any evidence of large scale deliveries or companies touting products containing the technology. At this point if production was in even near fully scaled we would see it on the market. My feeling is there may be either a speed bump or a serious problem in bringing it to full, economical production. As one respondent said we should know 6/29 when Micron, Intel's partner in this venture reports their quarterly results. I hope it is just production glitches holding this up and Intel/Micron work out the details. I follow this closely for I have substantial holdings in Micron, so I don't take this lightly. Intel/Micron could easily clear the air with a single announcement and have not done so, a very poor move at best, to keep customers and the market in the dark. As an investor with a substantial interest in the success of 3dXpoint, I hope things work out for the best, but don't like how current information on progress or lack there of is being handled. TSM has always been straight up and I also have a substantial holdings in. Intel should learn from TSM. I wish Intel had another Andy Grove.
I'm thinking that it must be very hard to compete against a mature technology like NAND on price.
There's an interesting comment in EETimes article about XPoint memory: 3DXP's Memory Role Unclear | EE Times
Flash-memory analyst Jim Handy is quoted as saying:
"that the manufacturing cost of the 3DXP drives that started shipping earlier this year could be an order of magnitude more than their high prices."
Optane (with its promise of 1000x faster than NAND and ready for persistent DRAM applications) was announced in mid-2015.
A year later we were told that select enterprise customers could test/benchmark their apps against a cloud-storage system based on Optane.
What Intel finally shipped after eighteen months is very expensive, high performing yes --- but high-performing along particular metrics (like 99th percentile latency), not uniformly blowing away NAND the way it was suggested.
It doesn't help that the use cases Intel has suggested for selling Optane drives are idiotic ("instead of using a 128GB flash drive to accelerate your PC, why don't you pay more for this 32GB Optane drive whose additional performance will make no use to your perceived speed?")
And THE supposed value case for 3D-Xpoint, persistent DRAM, remains MIA.
All of these are not the pattern of a technology doing well. You don't fire up the hype machine for technology you don't plan to introduce (in substantially neutered form) until two years later. You don't sell lousy pointless products if you can create more compelling alternatives. (To give a calibration point --- you don't see Apple announcing ARKit and promising "you'll get to use this cool tech in 2019, maybe 2020". Likewise at IO, Google announces what will be in the next version of Android; there isn't a wave of hype around what won't ship for years --- or if there is, like Google Glass, it's subsequently accepted that the project went very wrong, not denied...)
The claim is not that 3D-Xpoint is dead or can never work or whatever, it is "something very serious [with Optane]". This is essentially a claim about timing --- Intel is not rolling out this technology on a timing schedule that makes sense. No-one is (yet...) making more serious claims like "Optane is dead" --- we don't have info to decide one way or another.
IF Intel can fix whatever is preventing its use in larger SSDs (yield?) or in DRAM (power?) it may have the future it was promised. But right now these problems seem much more substantial than Intel imagined in mid-2015.
Certainly the claim that something has gone seriously wrong (based on the timing schedule) strikes me as simply common-sense, indisputable.
Totally agree with the above. Hype almost seems like a compensatory mechanism in tech, where the more hype there is about a product the further in advance of launch the worse the product ends up being. It's like "Oh shit we've invested hundreds of millions into this but it's got serious flaws, better create a bunch of hype so we can at least get a bunch of pre sales before everyone realizes it's crap and hopefully keep our jobs long enough to find new ones"
On the other hand, the best products seem to be those that are kept secret until almost right before launch, then catch everyone off guard with how amazing they are - ie iPhone. Which makes sense because if you have a truly great product the last thing you want is to tip off your competition early.
count, sometimes touting a product very early in the development cycle is a deliberate tactic to scare off the development or financing of a competing product. Tactics can sometimes be important or even more important than the product itself. Look at the inferior products that have won the battle through tactics instead of attributes.
I get what you are all saying, and you may be correct. I just pointed that there is a hypothesis that could explain everything you are saying (Optane shortage, delayed launch of new products) and that is if Optane is doing really really well.
Optane is actually a very disruptive tech, it may not be 1000x or even 100x faster/better than NAND, but 10x performance and much lower latency can already do wonders in some cases. Therefore, if nothing is wrong with durability/reliability than it would not be too much of a surprise if the demand was much larger than the supply.
For NAND which is much more mature, Intel still has a supply problem, why couldn't it be the same case for Optane?
Intel a global corporation. This is Oregon fucking up again. I lived in Washington County it's a lutheran country club atmosphere and those people say the things nazis say. It lacks the character and integry to succeed in tech. Everyone at intel knew the project would fail and if you didn't than you haven't been paying attention for the last four years.
count, sometimes touting a product very early in the development cycle is a deliberate tactic to scare off the development or financing of a competing product. Tactics can sometimes be important or even more important than the product itself. Look at the inferior products that have won the battle through tactics instead of attributes.
Intel does have a pattern of spinning big stories to the market, and than disappointing. For example about the IOT, mobile , altera, nand.
Does it get punished in the stock market for that ? or the reverse happen - creating hype raises the stock by X percent, but disappointing(especially when done while declaring some other thing, like self driving cars), reduces the stock by much less than that X percent, or at all ?
I'm probably missing a lot more. Intel's a global corporation those problems are Oregon specific. When Washington County could get the H1B1s things were fine and now they can't they been exposed. I wouldn't want to live in that place and I don't blame people for not wanting to either.
