Intel’s palpable desperation on display with Rocket Lake
semiaccurate.com
semiaccurate.com
I'm willing to bet my intuition is wrong, especially given my extremely deep bias against MBAs and 'this quarter' thinking. Any great sources on the full story?
https://www.extremetech.com/extreme/227720-how-intel-lost-10...
https://semiwiki.com/semiconductor-manufacturers/intel/28919...
The TL;DR is that Intel has always been a vertically integrated shop (meaning that they usually fab and design their own chips), and that is starting to bite them because pure-play foundries are improving their tech at a faster rate.
Intel has been unable to keep up with process advancements in their foundries, and that has led to pure-play foundries like TSMC taking massive market share. As chips get smaller and smaller, Intel has failed to keep up. They can only do 10nm for their mobile stuff and 14nm for their desktop stuff, whereas fabs like TSMC have been in 7nm territory for a while now and are moving into 5nm territory.
Are you saying that vertically integrated companies are inherently disadvantaged because pure-play companies have a bigger list of orders and can spend more on evolving technology?
There is advantage to be vertically integrated, namely a better global optimization across domains, like between design and manufacturing. Would be interesting to know why it's not enough - if it's always not enough. If not always, then this particular case needs more reasons.
With intel their foundries only have internal customers, who can only go to their internal foundries. There’s no competitive pressure on the foundries, and if they are ready early that’s wasted capacity. So so capacity and technology planning is based on a common road map to meet the needs of their slowest (er, I mean only) customer.
If Intel builds Intel (I believe their contract fabing is a rounding error?), then they're directly tied to Intel chip sales.
This sets up a potential death spiral. Intel misses a process node deadline, Intel's products are uncompetitive, Intel sales decrease, less demand for Intel fab, less money for Intel fab improvement.
Intel can temporarily paper over this by shifting money from other areas of the company, but it's not a good path to be on.
Conversely, as you might expect, if Intel sales are increasing then the opposite, virtuous cycle holds.
So essentially, Intel's fortune is tied to the Intel_sales : (global_sales / number_of_leading_edge_non-Intel_fabs) ratio.
And with regards to that, two huge things happened in the marketplace recently: (1) mobile chip sales explosion, (2) GlobalFoundries exiting leading process race.
If Intel hadn't been screwed by a process engineering miss, longer term trends would still have hit them hard.
The latter depends on your market share. The former only depends on the total market.
In the long run Intel screwed themselves over by not leasing out fab time to other companies. They put themselves in a niche in an industry that is naturally dominated by the largest player. And it's extra embarrassing that they did so because they knew very well how important it was to be the biggest--they were for a few decades!
Maybe Intel could have held on longer if they had a successful mobile chip to stuff into billions of smartphones, but their mobile efforts were short lived and seemed to be treated with disdain by the management. The first product kind of sucked and instead of sticking with it and improving they just threw in the towel, both on mobile processor and the baseband chip. An embarrassing misstep for a company as big as Intel.
sounds like Google who also treated their cloud/fabric computing as a competitive advantage to be kept to themselves and as result they missed the cloud business.
>their mobile efforts were short lived and seemed to be treated with disdain by the management.
classic. Low margin high volume future usually can't survive in the shadow of the high margin cash cow of yesterday that is still being milked.
So AMD's strategy also leads to lower fabrication costs: because they can make a far cheaper 14nm chip to handle the slower portions of I/O (talking to RAM, or PCIe), while the expensive 7nm parts of TSMC are used only for the cores / L1 cache / L2 cache / L3 cache.
Zen 2 has PCIe 4, so you could very well be correct.
What's keeping Intel from adopting this same practice? is it IP?
Intel can do it, they just haven't decided to do so yet. They have the tech for sure.
Its simply a matter of priorities. Its not so much that Intel "isn't" investing into it, its arguable that Intel just hasn't invested "enough" into it.
AMD went all in: they literally bet their entire company on advanced-packaging, with AMD GPUs using an active interposer with HBM2, and now Zen-chips taking a chiplet strategy. And to be fair: AMD had to do something drastic to turn the tide.
We're just at a point where AMD is finally reaping the benefits of a decision they made years ago.
--------
If I were to take a guess: Intel was too confident that they could solve 10nm / 7nm (or more specifically: EUV and/or quad patterning), which would have negated the need for advanced packaging.
AMD on the other hand, is fabless. They based their designs off of what TSMC was already delivering to Apple. Since TSMC leapfrogged Intel in technology, AMD can now benefit from TSMC (indirectly benefiting from from Apple's investments).
Intel's failure bubbled up from the fab level: Without 10nm chips, Intel was unable to keep up with TSMC's performance, and now AMD is advancing.
----
AMD's strategy just works really well for AMD. AMD is forced to keep buying chips from GloFo (which are limited to 14nm or 12nm designs). All of AMD's decisions just lined up marvelously: they fixed a lot of issues with their company by just properly making the right decisions in a lot of little, detailed ways. A happy marriage of tech + business decisions. I dunno if they can keep doing that, it almost feels lucky to me. But they're benefiting for sure.
AMD took something that seemed like a downside (the forced purchase of 12nm or 14nm chips, even when 7nm was available), and turned it into a benefit.
Did Apple actually buy a significant stake in TSMC or are you just referring to the fact that Apple is one of their large customers along with Qualcomm, Nvidia, Huawei (until recently) etc.?
I'm talking more like the later: all of these companies (Apple, Qualcomm, NVidia, etc. etc. and of course, AMD) are effectively pooling their money together to fund TSMC.
I don't mean to single Apple out as if they're the "only" ones funding TSMC's research. (And I can see how my wording earlier mistakenly can be interpreted in that manner. I was careless with my wording). Its more of a team effort (although Apple does seem to spend significant amounts of money trying to get first-dibs on the process).
The only time Intel really did that was with the end of the P4 and frankly even then they waited as long as they could before doing it. The rest of the time it's all carefully planned stepped increase and safe design change.
Both have their advantages, but for your question it means Intel will have to take a leap they clearly don't like taking.
There's some overlap between the 2, but 100% of Intel's current struggles can be chalked up to dysfunction on their process side.
Given the lead time of chip arch (2-3 years?), the chip side of the house is arriving at manufacturing start day, and the process and specs they'd been optimizing for just aren't available.
Until Intel's process catches up, other parts of the company have limited options. (TSMC!)
The process side of this industry is in a really scary spot right now. TSMC is killing it. Nvidia is using Samsung to fab the RTX 3xxx chips, and there's some rumblings that low yields are a reason why those are in such short supply (not to mention, whenever Samsung releases a new phone only some regions get the Exynos chips, because historically they couldn't produce enough of them; Samsung used to have a pretty cool relationship with Qualcomm whereby Snapdragons were fabbed with Samsung, but more recently, they've moved to TSMC).
On the high performance side, everyone is moving to TSMC. If Samsung and Intel's fabs continue to exhibit issues at these smaller process node sizes, the monoculture is only going to get worse. We need Intel to get their act together, not just because it pushes better design innovation from AMD and Nvidia (not that they need it right now), but because their fabs are a critical, independent part of the software supply chain. They're the last western company with any form of high yield fabrication on high performance chips. At this point, we shouldn't just be worried about Intel's bottom line; we need to start being legitimately worried about national security (both in tactile cyber-warfare terms, and also more nebulous economic terms with regard to western manufacturing).
TSMC is majority-based in a country whose land and government is claimed by another, nearby, much larger country. A country willing to leverage international economic options to further that claim.
TSMC's global importance has direct implications on the world's appetite for intervening or preemptively selling arms against a hypothetical Taiwan Strait invasion.
I’m sure it’s not lost on folks in the region the importance of TSMC when Xi is telling troops to “prepare for war”
Hopefully their plan to deal with the 30million extra men isn’t war ( https://www.chinadaily.com.cn/china/2017-02/13/content_28183... )
That's not true.
The reason for that has always been Quallcomm's patent shakedown.
That's not true.
The reason for that has always been Quallcomm's patent shakedown.
I have a close friend, who is also a close friend of Jim's family. Apparently the arrogance of Intel management hindered his ability to actually put together the type of team that he did at AMD. I'm actually concerned about Intel now.
I thought Zen 3 would be, maybe 5-8% better than Intel's current offerings, and newer chips would be on parity with them, but this is just embarrassing.
Basically the concept was instead of engineers having an assigned office or desk, the office would have a large field of unassigned desks and generic equipment. Every day you would come in, reserve a desk, take any belongings from a locker, and set up shop. At the end of the day, you would tear down, lock up, and leave.
It is beautifully efficient from a top-down perspective, but it turns out employees like the consistency of a known work location each day. Its also nice to be able to put up a picture of family, from what I gather. I suspect the cost of personnel replacement dwarfs the saving of n% reduction in desk need.
Dude, that guy's job probably relied on him being on the phone all day - he's just doing it. It's not his fault that you can hear him, it's your office manager/designer.
Some of us really do have to spend a huge chunk of our day talking to other people on the phone, zoom, etc. There's n other way to get our jobs done, and if they don't give us private office space to do it in, then we have to do it in the open space they do give us.
Open offices usually at least have quieter spots if you get in early enough, in my experience.
I would be miserable in such an environment. I need my keyboard and screens in a certain layout. Books need to be in a certain place. If that was taken away from me I don’t think I could live with that.
It's not that simple. I worked with hotdesking at Intel. Those workstations, although still small, were bigger than the (assigned) ones I saw at Google and Facebook when I interviewed there, and Intel's one offered more privacy. Visiting those companies for an interview was depressing - finding out your crappy cube at Intel was still better than the ones at FAANG. Surreal.
It is, but not within the timeframe that someone will get their bonus/promotion for the cost savings and move on. Same scam as outsourcing.
So dedicated office space can be expensive.
Sometimes I think the way big corps do offices is a legacy of mass production factory stuff. If you have an assembly line making toasters, yeah you need everyone in the same big building on a fixed schedule. But any business where people can work from home it likely means you could in thoery let individual groups decide where they want an office. Give a group of 12 engineers a stipend of $4000/mo to rent office space, see what happens.
I worked somewhere once that tried this open-plan hot-desking nonsense. The HR dept kept their nice private office with fixed desks however. They said it was because they had to deal with confidentiality. Whereas us dumb engineers developing the IP of the firm presumably didn’t...
Throwaway account as I work there, but I'd like to make some minor corrections to this: I don't think most of Intel ever switched to hotdesking. It was a trial program. In my building it was on one or two floors. I can't speak for all the locations, but most of the other buildings at our site did not have this at all. Unfortunately I was on that floor and hotdesked for 2 years.
I'm pretty sure none of the process/fab folks were hotdesked (I used to work in that area and stayed in touch with those folks). Nor did any of the circuit designers/architecture folks get hotdesked. After a few years, the trial was over and everyone reverted back to having their own permanent desk.
It's incredibly unlikely that this is a reason for Intel's decline.
As an example, my current cube at Intel has walls all around (albeit short, but some buildings have walls taller than most people), 2 file storage cabinets with a fair number of drawers, a decent sized desk (enough for 2 people, not that we'd do that). Correspondingly, the cube is enough for 2 people. Also, I have a whiteboard.
...so you would never be able to have a computer on for more than the length of your workday? That sounds insane --- I could never get any real work done that way, because so much time would be spent on "restoring state" to the way it was at the end of the previous day. I wouldn't call that efficient at all.
Laptops can be suspended rather than shut down, so maybe they're counting on that as the solution.
Sun solved that 20+ years ago with their Sun Ray technology https://en.wikipedia.org/wiki/Sun_Ray
These days you could do it with persistent VDI.
François Piednoël, performance architect at Intel for 20 years, recently gave a presentation that covers a lot of the reasons behind Intel's decline in his presentation "How to Fix Intel" https://www.youtube.com/watch?v=fiKjzeLco6c
Anyway still some interesting insight about what he saw happened there.
The situation can be explained in a very boring way in any case. You had the tick-tock cycle, then Intel 10nm was broken, but they thought not anything they could not fix with one more year of process debugging, maybe 2 worst case. And they had so much advance that they could even have tolerated 3. Except 10nm basically never worked. And the new microarchitectures were designed for it... They switched to thinking what to do with 14+++++++++ far too late, and even then I admit the result is suboptimal.
SemiAccurate is... let's say very very opiniated (as usual), to be polite, and I don't think Rocket Lake will be that much a disaster, but also yes Zen 3 is solid, although not that much magical and a bit overpriced for now (but I'm sure AMD will be able to adjust if really needed)
The final thing is: Intel had really an insane advance, and medium/high core count is not that much required right now in volume for consumers. I'm not sure about the situation in datacenters, could be more of a problem for Intel. But Intel has a level of market addressing that is insane and far above AMD. So it is not that much a big deal for Intel, at least if they manage to come back on tracks in the coming years.
So did Intel became that much uncompetitive? I don't really think so. Enthusiasts are just so a bit too much now that AMD really is competitive. Choosing Zen 2 was still a (often excellent) compromise. Choosing Zen 3 over Intel would be a no-brainer if cheaper (and when you can!), and depending of what you do it can often be the good choice even at the current prices, or at least for some specific models.
I don't personally believe that Apple would be going to in-house silicon instead of Intel for its flagship laptops if there were a viable way to avoid doing so. Intel is so hurt right now that I surmise it'd be willing to negotiate a fat discount for a flag-carrying customer, and the loss of Windows functionality is kind of disappointing at the upper end of the market (where some tools are clearly better supported by windows or at least x86).
It's not just Intel's failure. They must think their chips are competitive against AMD's as well (or they're all in on iOS apps on Macs).
Plus probably still cheaper than any x86 alternative.
Apple marketing always reminded me of how for decades, Rolls-Royce advertisements never would explicitly say things like the engine horsepower and displacement-- just "ample."
Now it will be that much easier to dodge performance questions. "Our machines are not built to run (mainstream software or game), so of course the performance is sketchy in the emulation penalty box. Just run the seven pieces of native MacOS software and it really flies."
Personally, I'm very skeptical on Apple beating AMD.
Yes moving to a new architecture is challenging, but Apple has done it before any this time it can be for keeps. Never again will they be beholden to another company’s priorities, or stuck with me-too processors their competitors have equal access to. A better Intel road map might have resulted in putting the transition off, but I think it was inevitable eventually ever since Apple bought PA Semi.
The problem with Intel missing on 10 nm is not so much that 10 nm is critical, but that it is critical to their roadmap. Large CPU design is heavily pipelined (like large CPUs), so you miss on the process node, you've still got a team building the refined next release for the year after, and a third team working on the more refined release for the following year.
Then you have decision making; it's hard to get a sense for if you need to go back and make a good new design on 14 nm, if 10 nm is going to be ready enough soon (but it's been several years now of not ready enough), splitting design resources.
I wonder if the migration from "tick-tock" to every third iteration is a case of believing your own PR. "Everything is fine" is what they should have been telling us, while internally it was flashing red lights and klaxons.
Or maybe this started even earlier, with the generations of hardware that gave us Spectre. The target became unreachable, they used smoke and mirrors, and when that blew up they just sort of gave up. Maybe the intervention should have come back then but didn't (cite the "MBAs took over" comment elsewhere in the thread)
Doesn't help they switched from large cubes with five-foot high cubicle walls which did a good job on minimizing noise and visual distractions to having cubicle walls with a height of about 3-4 feet high and much smaller. Plus they installed sit-stand desks and you have people stand and make phone calls that can be heard 30 feet away. Doesn't help for concentrating on problems.
Even this would be a substantial discount for most qualified people.
$120/hr * 40 hrs/wk * 50 wks/yr = $240K before self-employment taxes and paying for insurance. Ends up being comparable to typical compensation in cities that aren't SF, Seattle, or other top-tier markets.
Now if someone was doing more typical ad-hoc freelance work on a project basis, $120/hr might be too low for typical work. However, $90-120/hr as a contractor with a long-term contract and stable pay over 12-18 months is not uncommon.
Unless someone has a lot of new, small contracts knocking at their door nonstop, it can make more financial sense to take a $120/hr long-term contract over sporadic $200+/hr work that changes month to month.
$240K/yr, minus self-employment taxes, insurance, and other overhead, not to mention the lack of other non-monetary benefits and comp, is not a lot for someone doing serious machine learning.
I'd imagine that the type of specialized talent that Intel needs to catch up to AMD have plenty of options open to them. A comp strategy based on median industry comp might work to maintain their position, but probably won't churn out new industry-leading tech needed to jump ahead.
Some professions typically require both a related degree and one or more certifications or licenses. Accounting is one of those (CPA, CMA, CFA, CISA, CIA, EA) and this ends up being reflected in the hourly rate.
That harder work seems to only get done when you can't leave (or when leaving doesn't get results).
"We've done a survey and our pay and benefits are on par for the industry" = we are proud to be a profoundly mediocre place to work in every possible way.
Can't wait for this pandemic to end and the market for my industry/region to pick up again.
The industry trade organization did make a snazzy movie to try to attract young graduates into the industry.
You could use the parking lots as a proxy for employee prosperity. The chip guys drove better cars than the students, but the government and other corporate people had better cars on average. The chip bigshots made bank though. The chip people would also have crunch periods where they worked crazy hours.
The workers making bank were the tradesmen building the facilities that housed tools.
Those aren't mutually exclusive.
You can have more than one priority, even though you can't have more than one TOP priority.
At the same time AMD has been fortunate that TSMC has been able to continue with its node shrinks - supported by demand from and cash generated by high end mobile devices.
None of which is to say that Intel has been well run (cough McAfee cough) or that your criticisms aren't valid to some extent.
Are we saying this all started to unravel when most of the money moved to mobile and Intel still, after 20 years of warning shots, didn't have a compelling strategy there?
Back in the day before the electric grid really caught on, hydro- and coal-powered electric generators could be found in businesses large and small that had needs that went well beyond the limited voltage and amperage requirements of the meager electrified buildings at the time. It made sense for a business to have its own power plant engineering department to just feed and maintain the beasties if the productivity from thirsty electrically-powered machines far outweighed the enormous cost of an in-house power plant, consumables and staffing. As the grid evolved though, this need went away, and only really heavy industry these days exceed the delivery capabilities of the grid will run their own on-site power plant (more often they just try to "peer" co-locate next to a "real" power plant, kind of like HFT's with exchanges). Power plant manufacturing is now a heavily-specialized industry with margins much lower than Intel has become accustomed to.
Could the "grid" of the Net, cloud and mobile devices evolve to the point where the existing PC and server markets retreats in relative scale into "heavy industrial" applications like cloud infrastructure, and shift out Intel's cash flow from under them? Under such a scenario, the PC and server markets don't shrink as much as grow dramatically slower on a relative basis to B2C-grade gear and infrastructure that explodes in growth comparatively. Intel will still be plenty profitable if well-run, and still be mighty, but won't grab all the headlines, and evolve into a GE-like "mainstream" business.
It could, but the extractable rents of centralized data (eg. the Law of Conservation of Attractive Profits) will continue to push against this for a while yet.
Arguably, the hardware and connectivity we have now is already more than capable of dispensing with data centralization for most use-cases, relegating the cloud to niches like commodity backup-and-restore services, and premium on-demand compute capacity analogous to the use-cases for supercomputers (of which AI/ML is a growing subset).
We're already hitting use-cases and scenarios where centralized synchronization is becoming a hindrance that imposes additional costs, but the loss of control of a locus for extracting attractive profit margins implied by alternative software architectures generally limit investment.
Absolutely. Many analysts are predicting that the future is mostly public cloud (heavily ARM) accessed from mobile devices (100% ARM) so they're imploring Intel to get back into mobile before it's too late.
It's more than process node. Zen 3 now has superior IPC as well; the now unqualified performance lead of AMD devices happens at lower frequency. The process node problem is just the most visible and embarrassing problem Intel has. It is not the only problem. Intel's 10th gen is also still well behind the curve wrt side channel attacks. That gap will persist until Rocket Lake appears.
The rot at Intel is old and runs deep.
The one area that Intel has managed to maintain superiority is quality. Intel is still the thing to buy if you want minimum glitches.
Intel claims Rocket Lake will deliver 10% IPC improvement over Cannon Lake on the same 14nm node. Clearly 14nm hasn't been tapped out despite the five years and 3-4 (depending on who you ask) previous microarchitectures they've used it for. Yes, smaller devices enable better designs. Ironically if Rocket Lake actually delivers 10% then it's corroboration that Intel has been slacking on microarchitecture as well.
> for the architecture team who have to design for two processes at the same time rather than focus on just 10nm?
Intel has been designing devices for at least two nodes as SOP for 13 years now; they called it "tick-tock" which is the cycle of developing a new microarchitecture on one node and then porting to a smaller node.
The difference this time is that they're going backwards... In any case the argument that there is some great difficulty with moving cores between nodes is tough to support given the history.
Which is probably why AMD has better IPC now.
> they called it "tick-tock" which is the cycle of developing a new microarchitecture on one node and then porting to a smaller node.
The whole tick-tock cycle has been interrupted by the 10nm failure which has almost certainly thrown off their architecture development cycle. It's not about moving cores to a smaller node its about knowing whether you're designing for 10nm or 14 nm.
This is kind of a big one, and why I prefer Intel. I can't see it continuing, though. Intel will feel pressure to:
* spin products to seem even somewhat competitive
* reduce their margins on clock speed (historically, they overclocked well because of conservative margins), and run a little too fast/too hot
* push things out the door before they're ready
I don't care very much about a 50% performance difference, but I do care a lot about stability. If Intel can maintain a quality lead, I'm likely to keep buying Intel.
Few executive teams have that long-term discipline.
There are lots deliverables, and thus lots of things that can have superior quality I guess. But my recent experience has been the reverse of yours. Intel's early WiFi modules on a CPU board were so bad I was forced to buy a go down the suck it and see path, but a whole pile of them from various manufacturers on EBay, build a room of 60 machines so I could test failure point of the bloody things. Intel at about 10 clients or so, I didn't find the fail point of Atheros and it was 1/4 the price. Then there was the i915 which hardly worked for a year after laptops were being sold with it, followed by a seeming unending series of bugs from the Manage Engine, follow by Apple cancelling heir Macbook refresh because of Intel bugs. Then I had the misfortune of using Intels T265. To name but one bug, it you would not re-initialise it's USB bus (say after a CPU reset). Intels response is "won't fix, suggest you splice in a relay USB power line".
You must be looking at a difference sort of quality to me. Right now, Intel is terrible. In the mean time I've had the pleasure of using a few 10's of AMD GX-420CA MPU's, and they are just working after 5 years, zero failures.
This is high risk business. Unseen problems that come after pathfinding when big choices have been made may kill the success.
Seems like there were years when intel did not have much competition from AMD. Each year the "big news" from intel was their their Celeron/Pentium/iX/whatever was now up to 100MHz faster than last year's (during limited situations where temperatures allow blah blah blah). Or they add some new strange trademarked exotic-sounding technology brand name each year like "The new Intel Celeron/Pentium/iX with..." "Intel MaxT"/"Intel TruEvo/"Intel Cartara"/"Intel Artuga Plus"/"Intel Versuvia"/"Intel Centurior Max"/"Intel T-Xe" that no one really understands but is basically some sort of mundane enterprise feature no one actually cares about that does something for remote wifi management
Quick - rush to the store to get a new laptop!!! I totally need an extra 100MHz single-core maximum boost on a 4-core 2Ghz CPU (i.e. 5% max...) with Intel TruStep MagrittePro (TM) technology that does something to the colour gamut of my monitor in certain content creation scenarios.</sarcasm>
It strikes me that Intel have been caught napping and resting on their laurels. AMD appear to have come out with a great competitive product and Intel don't seem to have anything to compete with it with because they've been milking the market for the past 5+ years with tiny incremental clock increases and nothing actually "new".
They've allowed AMD to eat their lunch.
Maybe they've got a "real" new product they've been keeping in reserve that they'll bring to market now and surprise everyone with. Maybe they'll bring out something amazing next year. Who knows. Maybe. Maybe not. Seems like they've blown their lead for now regardless.
Basically he just needs to keep it afloat long enough for Intel to be able to find its version of Satya Nadella and Azure to unlock the next leg of growth.
They could lose a lot of market share and then come up with something better in the end, like in the Athlon days.
Not obvious that it will happen, but not impossible.
As an outsider to MS, its unclear that any one particularly saw Nadella as the obvious successor to Ballmer until they announced it. He somewhat came out of wood work IIRC. Same with Lisa Su.
[1] https://www.macrotrends.net/stocks/charts/INTC/intel/income-...
and that’s the pivot the software players made in tech, they monetised every click of the mouse (and every tap) while intel is stuck with the burden of the platform costs
Everything has sensors now, from manufacturing processes, to cell phones, to your optical mouse.
Imaging is much bigger than film.
Kodak R&D had a lead there too, but blew it bringing it to market.
I'm not even sure it was the first wave of digital cameras that did them in-- people bought boatloads of digital Kodaks in the 2000s. I recall they were very big on dock ecosystems-- drop the camera on a base to charge it, and I think they had some printer docks. That should have helped to differentiate them from cheap Olympus/Nikon/Canon/etc. point-and-shoot cameras.
Yeah, there wasn't the residual income but there was a fair bit of an upgrade cycle for a few years to make up for it, which could have bought them time to find a solid place in the market.
I think the problem was the second wave: the point-and-shoot consumer camera disappeared (outside of novelties like Instax). After 2015 or so, you're either looking at interchangeable lens systems or other prosumer/pro-level kit, or you're using a phone. Kodak never made much of an inroad in the pro-level digital market, and I'm not sure they had a business adaptable to cameraphone sales-- I don't think they made their own sensors or optics, so a licensing or subcontracting business would have nothing to offer.
I think I agree with the point you're making but I still feel obligated to point out that cheaper and better printer (like the brother laser pointed in the other thread) isn't stopping HP from making huge bank with their ink scam for decades.
Kodak was always a big player in the camera business, but they did it to increase the film business - they wanted people taking and printing as many pictures as possible, because that's where the real money was. It's hard today to imagine the scale of that business. And that's the problem, if Kodak had sold every single sensor that went into every single digital camera, it wouldn't have been enough to save them.
It's interesting to contrast the reaction of Kodak to that of the other big film company Fujifilm when the market for film cratered. https://petapixel.com/2018/10/19/why-kodak-died-and-fujifilm...
The current Intel CEO has been there for 4 years as a CFO prior to taking over as CEO. We know he can spell Intel... we know he isn't even a little bit technical (his degrees are business administration and MBA and just about all of his previous positions of note are as CFO). So basically he's going to run the company like a beancounter... which always works out well at companies that need heavy R&D to stay competitive.
WACC: weighted average cost of capital
On the other hand, attempting to quantify R&D seems like a fool’s errand. Maybe quantifying between different paths of R&D to see which could pay off more is beneficial but cutting the R&D budget outright because it’s not profitable? Seems like you’re cutting away from your company’s future at that point.
That's the problem with "MBA thinking". You can't put a dollar value on R&D. You just can't. Anything you do will be a gross approximation, and because the number is so fuzzy, you'll be encouraged (even subconsciously) to fudge it around to make other things look better, which will undoubtedly lead you to undervalue R&D to the point where it negatively impacts your business, but you have no idea why, because you're solely focused on dollar figures and not what really matters. Y'know, like developing the ability to solve actual customer problems in an exceptional manner. This sort of thing is very difficult to quantify, and any approach that starts from a finance perspective is always going to be sub-optimal at best, but often just flat-out wrong.
I see so many people optimizing the things that are quantifiable at the expense of the things that aren't, when the loss of the qualitative things is what's killing their business. MBA types seem to never get this, and the form of your reply is basically a textbook example of that.
It’s the same as how some firms get more return out of spending the same amount on factories as others. They are better factories.
It's on the income statement five years from now, not the one you have when you're making the decision today. The R&D paid for five years ago will commonly have been under different market conditions.
> Then I can show that some firms see more of a return on that R&D spending than others, and that impacts how valuable they are.
The question is, how do you cause your company to be the one getting more of a return?
Very simple: Cultural Decay
"January 31, 2019, Swan transitioned from his role as CFO and interim CEO"
They got what they paid for, I guess...
A few examples:
* Lisa Su (CEO of Advanced Micro Devices (AMD)): BS, MS, and PhD in Electrical Engineering from MIT, and is a fellow of IEEE
* Jensen Huang (CEO and founder of NVIDIA): BS in electrical engineering from Oregon State University, master's degree in electrical engineering from Stanford University
* C.C. Wei (CEO of TSMC): Ph.D. in Electrical Engineering from Yale University
* Simon Segars (CEO ARM): Bachelor of Engineering degree in electronic engineering at U of Sussex, Master of Science degree from the School of Computer Science at the University of Manchester
* Sundar Pichai (Alphabet/Google CEO): Has an MBA, but also has an M.S. from Stanford University in materials science and engineering
* Eric Schmidt (former Google CEO): BS in Engineering, M.S. degree for designing and implementing a network, and PhD degree in EECS, with a dissertation about the problems of managing distributed software development and tools for solving these problems
* Jeff Bezos (Amazon CEO): Bachelor of Science in Engineering (BSE) in electrical engineering and computer science from Princeton
* Mark Zuckerburg (Facebook CEO): In Harvard studied Psychology and Computer Science (did not earn a degree, but did study it for a few years and implemented the first version of Facebook).
* Tim Cook (Apple CEO): MBA from Duke University and a Bachelor of Science degree in Industrial Engineering from Auburn University.
* Satya Narayana Nadella (Microsoft CEO): Bachelor's in electrical engineering from the Manipal Institute of Technology in Karnataka; M.S. in computer science at the University of Wisconsin–Milwaukee; MBA from the University of Chicago Booth School of Business
* Reed Hastings (Netflix CEO): Bachelor of Arts degree in Mathematics (Bowdoin College), MS Computer Science (Stanford University)
I'm sure there are more examples, but I think that amply demonstrates my point.
The only other example I found of a no-tech CEO leading a tech company was Safra Catz (Oracle CEO), who has a bachelor of arts and a J.D. (Law School). My search wasn't exhaustive, but it was illuminating.
Now let's compare this to Bob Swan (Intel CEO), who received a bachelor's degree in business administration from the University at Buffalo School of Management in 1983 and his MBA from Binghamton University in 1985. No tech at all. Maybe Mr. Swan can do well anyway, but his lack of technical education is extremely unusual when compared to most other tech companies.
I think you've made the point extremely clearly.
They're extremly good (better than most) at keeping the machine well oiled and maintained to ensure it performs as best as possible, but when the machine won't be able to do the job anymore they will have a hard time foreseeing it or finding the new solution.
Just like what is happening at Intel.
So has there been organizational dysfunction for decades? I think not: there are things Intel has always been doing very well and post-Netburst improvements after Core 2 have also been significant.
I believe that a big element is basically luck: you invest in a certain progression path and that investment will yield results so you keep going. Another path might be sufficiently better to write off the inferior path investment but you don't know that. Perhaps the unsatisfying path taken is the least bad of all. Even Netburst improved over time, a bit, and so did whatever forgettable AMD had been building between the Athlon glory days and those of Ryzen. As long as you see progress, it's very hard to just give it up for a fresh start (that may or may not be better). We can just be lucky that a luck/lock-in imbalance has never persisted long enough to make the lucky one the only survivor, because then they would never leave the left dead-end they'll inevitable run into some day.
Phenom II! I had six cores, 7 years ago! They were dog slow, but there were six of them (and that's all I could afford).
And even when it is obvious for particular workloads, the relative value and demand of various future workloads is definitely non-obvious.
But then a company in The Netherlands solved the extreme ultraviolet generation and optics problem and Intel failed to push their existing process to 10nm allowing TSMC and Samsung to overtake it as a technology leader. And Intel now has to deal with a situation that it had no experience for the last 40 years when the company was a technology leader.
INTC deserves that 9 P/E ratio
There are lots of moving parts with an enormous organisation like Intel.
But if you ask me there is one thing to point the finger to, it would be ex-Intel Chairman Andy Bryant ( He retired earlier this year ). Raised to CFO in 90s. Promoted to CAO in 00s. And later head the process and manufacturing operations while forced out Patrick Gelsinger. That was I think in 2009. The Otellini and Bryant era.
Paul Otellini retired in 2012, and BK ( Brian Krzanich ) was picked ( by Andy Bryant ) as Otellini's successor.
There seems to be lots of power play within Intel that we will never know.
This is an insightful analysis of what’s going wrong at Intel and the lead-up and fall out after Keller left:
Part 1: https://web.archive.org/web/20200730185751/https://twitter.c...
Part 2: https://web.archive.org/web/20200729091548/https://twitter.c...
Part 2.5: https://web.archive.org/web/20200802065407/https://twitter.c...
Part 3: https://web.archive.org/web/20200807105015/https://twitter.c...
Part 4: https://web.archive.org/web/20200812073050/https://twitter.c...
That means the best case scenario for Intel would be (barely) scraping back their "single threaded gaming performance" crown while completely giving up against the multi-threaded performance of AMD's higher core count Zen 3 chips. The only way Rocket Lake would make any sense in the market would be if these are priced less than $400 (probably a lot less), and so Intel's margins will be much lower on what is likely to be a much larger die with more transistors.
I don't think it's possible to call this anything other than a pure desperation move.
Amd has had Gen4 out in desktop CPUs for over a year now.
Pure speculation, but maybe someone's bonus depended on delivering and this technically satisfies the requirement. It's not designs problem fab can't deliver the right node.
My understanding from the article is that the answer to that question is that Intel is unable to produce any quantity of chips at 10nm that makes (enough) money. The author surmises that rather than go a year with AMD essentially unchallenged in the market, Intel back ported a 10nm chip to a process that they have better margins on 14nm, so that they could tout "improvements" on an architectural basis while skipping over the question of improvements on a system basis.
For me, what is most telling is the 500 series chip set which seems uncharacteristic.
From a purely speculative point of view (that is code for pulling a wild ass guess out of my butt here) I could see this as having been the design they had done for Apple's next generation Macbooks before they lost to Apples' A14x chip. The factors that lead me to that guess are that it really looks like a 'point' product (specific changes, not a general set of family changes) to me that doesn't fit into thee PC ecosystem as well as other chip introductions (like IceLake) did with respect to their overall roadmap.
Do you mean "is a horrible sign for Intel"?
More and more, Apple's timing on their switch to in-house ARM designs seems perfectly timed.
Now they either need to halt some Mac lines or develop all kinds of CPUs, from mobile (which they probably can do, because they should be similar enough to iPhones) to server-grade (which they have no experience at all). And there's no way Mac Pros would sell enough to offset development costs. I just don't understand how they are going to manage that situation.
How is that supposed to affect anybody else who is still buying the regular stuff, or be better for Apple customers than the same thing but with an architecture transition and the inability to natively virtualize the x64 editions of Windows and Linux?
> and a lot of manufacturing power would have been wasted on Macs making it more expensive for other people.
Apple and AMD are both TSMC customers. They come out of the same fabs.
For one, these CPUs would make the fabs busy, that means AMD would produce less CPUs for the "masses" and probably that would make it more expensive for everyone. I think the inability to virtualize would kind of fit the walled garden Apple is going for.
> Apple and AMD are both TSMC customers. They come out of the same fabs.
It's not like you can produce different CPUs at the same time. It requires setup and probably particular fab is able to produce one type of CPU in a batch. The TSMC resources are finite as well.
The main point of switching to their own silicon is to unify their entire product range onto the same architecture.
Investing in a switch to AMD would be doing literally the opposite of what they're trying to achieve. Plus the shift of the big cloud players moving onto their own ARM chips means they're going to abandon x86 and move into a space that's already gaining traction in server grade.
'Now' they don't need to do anything but build the products. You act as if they only started working on the mac chips after they announced it. They'll have been planning this since the first A processors started performing really, really well.
With the ARM transition, Apple has perfect compatibility with the iPhone which is arguable more important for Apple than Intel compatibility.
Apple is already investing tons of money in CPU design, they are just doing it via proxy with Intel building the chips for them. Bringing it in-house means they can add the features they want to the CPU on the time frames that meet their needs.
Going with AMD just shifts who they outsource their chip design to, it doesn't give them the sort of control over their architecture they want.
We’re going to move towards AWS, Microsoft, Apple chips. It’s really hard for a company like Intel or AMD to serve hyper scale customers.
Although this is possible, it would be a huge hit to Intel's image.
I think one can state that Zen 3 is higher performing than any Intel's architecture/model across the board. What happens when they'll release Rocket Lake? They'll introduce a new(er), slower architecture, to tackle... the budget segment?
I think that, from an engineering point of view, Intel is going to be in deep for a few years, with the hope that they'll manage to pull a new architecture in 3/4 years.
On the other hand, I've specifically written "engineering", because Intel still has copious amounts of "green persuasion", which shouldn't be underestimated.
Maybe somebody can provide some context.
It might not be fair, but I do like to know the mindset the article is written from.
Just a reminder. Intel bribed, intimidated and blackmailed hardware manufacturers to alienate AMD while AMD was offering better product. Intel was fined record 1 Billion Euro. Intel hasnt paid a single cent of that fine to this day.
https://www.reuters.com/article/us-eu-intel-antitrust/intel-...
The counterbalance is that he's got one of the absolute best networks of moles in Silicon Valley and always seems to know things, often big things, before anyone else. Sometimes even before it's on anyone else's radar.
So if you're familiar enough with him to back out his style to see the underlying (likely) facts, which isn't too hard, you can learn a lot about the big chip industry. And that's not just interesting to people like us, but very very valuable to a lot of people.
(Disclaimer: I haven't read SemiAccurate much since he paywalled it ~10 years ago, a move that I don't like at all but completely understand from the perspective of his business.)
He's ranty and very aggressive with his writing. His best stuff is kept behind the paywall. It wasn't really worth the money to me (and I only subscribed once as a curiosity). But those "moles" he talks about are clearly the real deal.
Specifically I know a lot of industry insiders have a dislike of some of Intel's practices. When S|A talks about them, S|A knows to take the "ranty" tone, as a means of signalling up front to readers, "Hey, this isn't going to be a bland slide presentation..."
As for his grudge, it probably stem from Intel's underhanded tactics. AMD had the Athlon XP that buried Intel's Pentium 4 in every measure.
Intel made some very backhanded and immoral moves- * Bribed Dell to not use AMD * Threatened Taiwanese mobo manufacturers to not build Athlon mobo by withholding North bridge chips * Blackmailed the said manufacturers to not even show up for product conventions held by AMD * Forced RAMBUS down everyone's throat until Athlon forced them to release the i815 chipset (finally supports DDR1!)
As a result AMD never managed to grow their revenue to a level that can sustain their R&D. When the Intel's Core arch arrived 4 years later it was over for AMD.
The Inquirer had a lot of contacts that suffered greatly, and it must have left quite an impression on Charlie. I remember that one of his reports was that during a trade show, a Taiwanese mobo rep had to meet Charlie in secret in a hotel room to talk about AMD, in fear of Intel catching wind and retaliate.
BTW Charlie also has a grudge against Nvidia for similar reasons, but Nvidia has been competent enough for the last decade that Charlie couldn't find anything to complain about.
And while I agree with your proposed first step of the solution, and the urgency to do it given the delay between that and any impact on results, their tendency to look only for short term gains and benefits will probably impede them.
They've learned to love easy money, and they didn't need to really pay their people well because there was no serious competition, not in fab, not in chips. Now they're fallen behind on both, as easily predicted, but for a short while they extracted massive amount of money from their margin.
Silicon backdoors are the new nukes.
https://www.eurekalert.org/pub_releases/2020-11/ou-1ei110320...
Essentially, if you trap a few photons in little resonator cavities that rely on mostly all of the electrons making up the cavity, the next set of photons can pass through the cavity without interaction.
This means the light interacts with engineered electrons in a specific way to create an optical switch. This is a critical component of photonics on a chip, and was just demonstrated with Si (which is a requirement for something in the mid term).
All-optical switching (i.e. photonic computing) is like fusion power generation. It's been "right around the corner" for 20 years, yet we never seem to round the bend.
It is, however, great for physicists writing grant applications.
> look at this recent development: https://www.eurekalert.org/pub_releases/2020-11/ou-1ei110320...
This was on HN here: https://news.ycombinator.com/item?id=24997573
My response here: https://news.ycombinator.com/item?id=25012670
Summary: this is converting information from the photonic domain to the thermal domain. Heat diffuses very slowly; you won't be able to put a lot of these devices near each other and run them at usable speeds without them causing each other thermal interference.
[But seriously, the green light of a standard green laser pointer is generated from the non-linear interaction of pairs of 1064nm photons.]
Um, no. It is caused by stimulated emission in the lasing medium, when electrons drop from a higher to lower energy level. Pairs of photons do not interact with each other.
https://en.m.wikipedia.org/wiki/Laser_pumping
In rare situations you might have an ion or an entire atom changing energy levels. But never just a photon.
https://en.m.wikipedia.org/wiki/Stimulated_emission
If you don't believe that photons are bosons I don't think we're going to get anywhere with this discussion...
In a typical green laser pointer a 1064nm light is produced via ordinary laser mechanisms, then 1064nm light is upconverted in a non-linear medium to 523nm.
https://en.wikipedia.org/wiki/Second-harmonic_generation
Yes, in a medium. You also don't see a lot of conventional computers constructed without semiconductors. :)
TSMC is now the only reason Taiwan can count on the US coming to its defense when China invades. You can't put a price tag on that.
Single Core:
AMD 5900X - 3643 points (+15% faster)
Intel 10900K - 3178 points
Multicore
AMD Ryzen Threadripper PRO 3995WX - 88673 points (+130% faster)
Intel Xeon Gold 6248R @ 3.00GHz - 38521 points
By the way, 5900X is not only faster than 10900K, but also consumes less power.
https://www.anandtech.com/show/15762/the-acer-swift-3-sf314-...
Or if going up to 15" is acceptable:
https://www.anandtech.com/show/16153/the-acer-nitro-5-review...
This isn't possible. The cores are already as fast as they can be, so moving from 10 big to 1 big would only reduce performance.
What does Windows support heterogeneous cores at all? I don’t think it does. And until that happens, would you be able to sell a chip to the public?
A lot of people I see on hardware forums seem to have the deception that we'd suddenly have 10 GHz cores had big.LITTLE x86 processors appear.
Reality is we'd have few of exact same cores we have today and a dozen more atom level cores put there solely for the purpose of marketing spec race.
And why is it only AMD that has a GPU baked into a CPU? Vega graphics are simply amazing and beat Intel's integrated graphics hands down. It's so nice to have only a single chip without buying an extra graphics card.
Unfortunately, AMD isn't very present. I see Intel sponsoring every eSports event out there and handing out i9s as prizes. I'm not tunes into the gamers but I still have the impression they believe Intel reigns supreme...
With RISC-V about to enter the market and AMD beating Intel on benchmarks, Intel better pull a rabbit out of their behind to stay competitive. But that's just a layman's view on the subject.
It can only execute on its core business which the MBA infestation has been riding for a decade with their IP lead and now it finally has run out.
Despite that, it is still probably too late anyways. My industry has moved on to GPUs and FPGAs for real-time compute, but I remember when everyone was waiting for 512 after utilizing AVX-256 for some time. Although the thought of a 64-core threadripper with avx-512 at 5GHz on all cores is appealing.
Also, AVX512 can be on many workloads much faster than gpus.
https://cpu.userbenchmark.com/Compare/Intel-Core-i9-10900K-v...
https://cpu.userbenchmark.com/Compare/Intel-Core-i5-10600K-v...
https://cpu.userbenchmark.com/Compare/Intel-Core-i5-10600K-v...
But that huge market drove investment in fab technology outside of Intel, and allowed them to close the gap and provide competitive, now superior, fabs to AMD.