Intel’s Turnaround and the Future of Chipmaking
economist.com
economist.com
That sounds ludicrous. If true, it has to count up there as among the most expensive way to get at "OPP" (in the Naughty By Nature sense). I don't see how the BoD and all sorts of DD processes could have signed off on it. Then again, I simply don't know how politics and power works at those levels, so I'd certainly welcome enlightenment.
My two cents: Intel is full of people who are career oriented and not product oriented. Their main goal is to get a promotion, and often find means to do so without contributing anything meaningful to Intel's products. It's also full of senior leaders who believe strongly in credentialism[1], complexity[2], and style over substance (i.e. how the message is delivered vs the content).[3]
From a SW standpoint, I have not yet been in a team where all teammembers can handle branches.[4] This is quite acceptable.
In one team I was in, I was leading the efforts for a product that required features A and B by the customer. I was a junior member of the team with no domain knowledge, but I was somewhat of an expert for that customer's domain. Everyone was on board with the technical work. In every meeting we had for about a year, there would always be some person in the team who'd suggest things that would nullify feature B. I would have to remind them that we agreed to do features A and B." The response would always be "Oh, we're also doing feature B?"
The person who would say this varied from meeting to meeting. But I was very frustrated that they couldn't remember this basic fact, and often ended up writing code that had to be undone. And then deal with their frustration as if I had never mentioned feature B to them. I can understand if this happens once or twice, but I have to remind them in every meeting.
But this was normal behavior. I was the odd person who thought this was unacceptable.
Oh, and coming to meetings unprepared is the norm. No one will read your emails briefing them about the meeting ahead of time.
[1] "Let's hire the PhD with no experience and not the internal MS employee who's already doing the job they are hiring the PhD for"
[2] "I don't care if your code sped up our workflow by 5x. It's just what, 200 lines of code? Anyone can do that."
[3] Presentations break most rules of effective communications/presentations. A senior person once told me "You explained things too well, and your slides are fairly sparse. Fill it up with jargon and lots of plots, and don't explain it as well as you did. If senior management understands your work too easily, they will believe the work you did was trivial. If they have trouble understanding it, they'll be in awe."
[4] One former manager: "Every person will get his/her own private branch. Do all your experimental work there. There will be no more branches." A senior member in another team said "Why complicate things by adding new branches for our various experiments? Let's just keep it in the main branch and enable the different algorithms via command line flags."
For perspective, this is typical LargeCorp behavior, especially at a company like Intel which makes 77 billion/year in revenues (700k/employee) at a 25% net income margin! The unwritten rule is that managers aren't incentivized to police this behavior. It's liberating to recognize this and modify one's approach.
> Oh, and coming to meetings unprepared is the norm. No one will read your emails briefing them about the meeting ahead of time. >A senior person once told me "You explained things too well[...] If they have trouble understanding it, they'll be in awe."
Wise words. In cultures that do this, you have to adapt and work more on narrative/story-telling. In many ways, this is how things actually work no matter how efficient you think you're making the team. Instead, write the same emails, but only to gather your thoughts. Then, lead the discussion. In large teams, it feels like this approach reminds me to avoid doing someone else's work.
Unless it is the manager who has to constantly do the reminding. Then there is swift policing :-)
Yes, this is actually normal "human" behavior. But this level of extreme was ridiculous, even within Intel. I quickly left the team once the project was over. Life is too short.
> Wise words. In cultures that do this, you have to adapt and work more on narrative/story-telling.
Narrative/story telling is good, but is orthogonal to the issue here. The usual flow is to use narrative/story telling to explain the why (motivation, etc). However, some senior management will expect you to also talk about the details. And this is where the advice came in: "Put the details, and make sure they don't understand them." My sin was that I was presenting the details in a manner where they could understand it (without losing the nuances and details - I was mere presenting the same material "well").
A more severe example will enlighten: I once solved a challenging problem with a really simple solution. My manager had multiple sessions with me to coach me on how to present that simple solution in a much more complex way. He emphasized that senior management should not realize that the solution was simple - no matter how impactful it was.
Yes - this is also a general "human" problem, and is common in lots of places. However, when you're striving to be the best company in X, it is wise not to settle for "average".
Basically, a company that's in an extremely strong position from prior successes may lose internal feedback loops that kept the people and culture on track.
Doesn’t mean something bad is happening, yet.
Some improvements like fit & finish, tolerances, endurance, so on are not visible on paper. Others like battery life and weight are visible on paper. When these combine with the xOS (where x={mac, i, iPad, tv}) ecosystem, whole thing becomes visible or palpable, one may say.
It's not an ubiquitous thing to have in technology space, and only possible with tight integration Apple provides, and strives to keep.
I like Apple at its best, but I wouldn’t say they’re the best in every (or even nearly every) category they’ve entered.
Yes they have iPhones (top range, even if they have competition from top range Android alternatives), iPads (peerless), Apple Watches (dominant), iPods dominated the MP3 market back when that was a thing anyone cared about…
But they also make really weird mice (Magic Mouse 2: charger on the bottom; many models: one button; Mighty Mouse: scroll nipple; iMac: unergonomic puck), the Apple TV and HomePod are nothing special (the latter is a pity, given I don’t trust their Voice Assistant competitors), and the device power cables have been infamously fragile for a long time.
The first ARM, that was a tour de force. The 6809 was too, and of course the 4004. But most other CPUs to me are run of the mill and the M1 is in that sense to me nothing special though of course it will give Apple a bit of an edge, but when all is said and done it's just another ARM based SoC.
At best it is an optimization, the most interesting part of the M1 is the dedicated neural net, and I'm not sure if anybody has already done something with that that the main CPU could not have done.
Intel is now chasing some design decisions within M1 with Intel Thread Director with different types of cores: https://www.anandtech.com/show/16881/a-deep-dive-into-intels...
How not M1 is not an Innovation? I'm not claiming M1 is the first CPU to do it, I don't know, but they may have executed it flawlessly. Great performance, great battery life and great silence. If you don't call it innovation, at least call it breakthrough.
I have seen this being repeated again and again by DaringFireball, Gurman and many other places. It started with ARM bigLITTLE, not M1.
Not really. You can't buy an M1. It's a part of a bigger product. Is the M1 impressive alone? Or is the MacBook and everything else including the M1 impressive?
And it's not the first chip Apple has done. It's just the first one for a small selection of computers. The M1 a tour de force? No.
If you remove "tech" from this sentence I think it's still true, and it points you towards the solution: bias the economy towards small companies (for example, via progressive corporation tax). There's a bit of nuance here: I think a lot of the problem is with public companies. So we could probably do a lot by reforming the public ownership system.
That's why we have entire sectors of the economy focused on wealth preservation without volatility.
These people want a stable and steady climb upwards.
I think it's because there is an alarming lack of arrogance and self-confidence in the world. It's visible in every possible datapoint . People say those things are bad, but in reality they keep the system honest.
The best way to redistribute wealth is wealthy individuals overplaying their hand. It doesn't ever happen, there is something deeply wrong in Bill Gates being at the top of the Forbes rich list for 30 years.
His brain should have adapted to the new normal and compelled him to risk his whole fortune in the hope of making even more money, maybe becoming the first trillionarie. That's how wealth is redistributed; rich people becoming greedy and overplaying their hand.
It doesn't happen, rich people are happy being at the top and they are not ambitious enough to reach for the impossible milestones.
The only nut out there who does it is Musk but his wealth is not real yet, it's paper wealth.
We'll see how Bezos decides to go about it.
Why do you think this should happen? That doesn't sound very rational to me. If you have more than you could possibly ever need then why would you choose to risk that? This is one reason why we should never reward people with these ridiculous levels of wealth in the first place.
> The best way to redistribute wealth is wealthy individuals overplaying their hand.
This sounds like a very wasteful and unreliable way to redistribute wealth to me. Unreliable because as you say it often doesn't happen. Wasteful because if it does happen it involves wasting enormous amounts of resources on a failed project.
IMO the best way to redistribute wealth is to not allow people to become as ridiculously wealthy as they are in the first place.
Because you don't simply do it for the stuff. You do it for the brain juices which are released when you win.
Bill Gates biggest wins happened in the late 80s and 90s.
His brain should compel him to do better than that to release those elusive juices once again.
The only way to do so is having even bigger wins than the ones in the 80s and 90s
> Failed project
The entrepreneur's failed project is the worker dream job where you get paid for doing nothing.
Kinda like the influencers hired by Bloomberg in his 2020 campaign. Nobody was checking if they were really doing the job and they just pocketed the money.
Well sure, but those motivations exist regardless of any financial rewards. We are talking here about the additional effect of financial incentives. IMO the possibility of making say $1m or $10m is pretty motivating if you don't have much. But once you reach say $1B, financial incentives are unlikely to be as relevant to you. You're made for life, why should you do anything you don't want to do.
> the worker dream job where you get paid for doing nothing.
I think most workers would prefer to be paid to something useful/meaningful, unless they can literally go and do something completely different with no restrictions. But is rarely how billionaire's failed projects go. They usually involve a lot of people working very hard to no end.
The amount of the financial reward measures how much you are winning. The juices produced when you win 20$ and those produced when your leveraged long in biotech stocks make you 750k are not the same.
> They usually involve a lot of people working very hard to no end.
This is just a pessimistic outlook. Most stuff fails but the few thing that stick produce a big improvement.
And it's not just billionaires who face this problem. Government has an even worse success rate, but the home runs are paradigm shifts. The failures burn a lot, think about the CERN failure, that's 20B right there.
The ITER fusion seems to be the same thing and that's 50B
The ISS never recovered the initial investment made, that's 100B.
It's not just billionaires projects wasting money
> The amount of the financial reward measures how much you are winning. The juices produced when you win 20$ and those produced when your leveraged long in biotech stocks make you 750k are not the same.
I suspect the world might a better place if we gave fewer people who are motivated in this financial manner, and more people who are motivated by the impact of their work (of whom there are plenty, and they're just as good as the financially motivated ones).
Nobody does the work for the sake of doing the work.
People work for a combination of financial reward and social status, and they'd use that as a metric to measure how much are they winning compared with the other 8 billion people on the planet
There is only a very very tiny percentage of people who are happy with just citations and the tepid approval of other academics
So if you roll back financial reward all the way down , then you have to pump social status all the way up.
It's going to be even worse. The cult of Steve Jobs and Elon Musk and Donald Trump would be the normal and would trickle all the way down.
You say this, but you point at evidence saying the opposite.
They seem rather dull and desensitised to anything. Very robotic.
Very similar to the British royals and European old money.
They live in fear of losing their money and relevance instead of having the confidence to double down on their bets and hit yet another home run.
If you look at billionaires then it's pretty easy to pick the ones you'd want to be: Mark Cuban, Donald Trump, Rihanna, Cristiano Ronaldo....etc
Those type of billionaries who are big spenders and also continiously betting on themselves are the type of billionaires also good for society because they make the economy go around and their projects are cash cows for other entrepreneurs looking for easy oversells.
Can't imagine living in fear like Gates, the Waltons and Warren Buffett, they aren't human.
Who are we and why do "we" do that?
"we" currently reward people with this amount of wealth generally because of a majority belief in one or more of the following (none of which are true in my opinion):
(1) They deserve this money, because either:
(1a) it represents a proportional reward for work put in
(1b) any money gained under a market system are deserved/earnt, and taxing those earnings is stealing
(2) That rewarding people in this manner (without limits) is the only way to motivate people to work hard, and that this ultimately produces better outcomes than distributing our economic output more evenlyWatching some of Intel's recent events seem to be driven by sales and marketing in a very tone-deaf manner. It's almost always been the case, in my experience, that sales leaders don't see competition. They continually talk about how what they represent is the best and you can't have a lucid conversation with most of them because they don't fundamentally understand what may make the competition better, or even fundamentals of their own product. It's black and white and many seem to just go all in on FUD to a ridiculous level, when forced.
When executives are far more focused on what wall street thinks than what their own customers want you know that the cancer runs deep. I continuously hear things like "our customers want a subscription model for our software because it's more flexible and is often cheaper!". Neither is, generally, true when it's being broadcast like that. It's a dark pattern (most notably in enterprise software / hardware) for a company to pull in long term revenue on a product that is, potentially, not evolving fast enough to warrant customers wanting to pay for an upgrade or new version. And a lot of SaaS just to "SaaSify" something, even though the customer could very well run it themselves, where they'd like, and be far better off.
It's just frustrating to have been in the industry as over 20 years now and to see even less inspiring, less charismatic and ignorant people continually failing up and driving companies down. I know C-levels that have lied, cheated and been fired for the former - yet they continue to land better positions, somehow? Some days it's beyond maddening to watch. But more often than not these folks are in some way, shape or form tied to the overarching sales machine.
Both bring in revenue, but neither creates product.
If you have leadership that allocates resources to boost revenue, without an understanding of why customers are paying, more and more resources and power accrue to sales and consulting departments. Eventually, this kills the company.
Above anything else, the lack of a distinct and political sales and/or consulting org is what makes startups startups. Most of the people fulfilling those roles are dual-hatted elsewhere, or at least sit close enough to people who are that they have a more holistic view.
It's all large companies. They think they are too big and entrenched to fail.
I've worked for 2 large non-tech companies where managers and leaders loved to say "We are X. We have the resources to do anything. If we apply ourselves, we are unstoppable." One of those companies tried to enter a new market and failed miserably because they kept trying to tell customers what they wanted instead of listening to their customers. The other company is a classic example of what people talk about when they make fun of "too big to fail" companies.
Management will say this and in the next breath say that we either don't have headcount to alleviate an understaffed team or we don't have budget to retain some amazing engineer that they want to see how long they can underpay before they leave. I've lost track of the number of times I've told management about how import it is to retain certain individuals and how they absolutely should make sure those people are happy with everything within management's control, and then they ignored those warnings and then the talent left. These losses happen slowly and then suddenly when enough of the remaining talent looks around and realize that most of the good engineers have abandoned ship.
Admitting you have a problem, first step, etc.
disclaimer: ex-IBM.
What that actually means in a situation like this, where technology and production pipelines are years long and a misstep can set you back a decade or more (like it did for AMD, and like it's doing now for Intel), I don't know. I do know it sure seems like markets with at least three major players are more healthy than those with two, as we keep seeing again and again (iOS and Android being another recent example, if exhibiting a different problem).
The problem comes from the sociological fact that groups differ from individuals in serious cognitive, sociological, moral and motivational ways.
And as a general rule, groups of people DO NOT SCALE WITH SIZE.
This is the fundamental reality that has been shown to limit the size of companies: the ability to reliably transmit critical operating information from one side to the other as needed fails with increasing size.
And computers do NOT improve things because it's actually a specific type of information required: knowledge applied with skill in a timely fashion. That is what fails with size.
Related to this is the inevitably "Chinese Whispers" losses that arise as the organization grows vertically and as decision making is centralized: the knowledge about the markets served inevitably goes 180 degrees out of phase where every direction and decision becomes exactly the wrong or worst possible. The system starts to tear itself apart destroying its competitive abilities and opening up large swaths of market for smaller competitors to take and occupy.
The real world has more information in it than what any planning or system can describe accurately enough to use for predictions. So without accurate feedback from the outside reality, the corporation will fail.
Big tech components can only be "fixed" by breaking themselves apart so they can regain the proper feedback loop accuracy, or by doing what HP USED TO DO: pushing down decision-making to the lowest possible level in a Federated form. This isn't perfect either but it's better than a Top-Down command-and-control system when it comes to performance and competitive fit.
Control-freak executives always screw both of these up because they can't handle letting go.
Posturers tend to eclipse doers over time because doers are too busy doing to spend all their time posturing and social climbing.
It's worth noting that Intel 10nm can't be directly compared with TSMC 4nm. AFAIK Intel 10nm approximately equals TSMC 7nm. TSMC 4nm sounds like an iterative refinement on 5nm, so intel is only really one to two generations behind, not "several".
So counting "generations" is a little abstract anyways. You should probably look at the PPA numbers and the kind of scaling boosters they've managed to ship, if you want to compare fairly.
I don't think so, TSMC 7nm is ahead in many ways. For example, Intel 10nm power efficiency is much worse.
https://blog.cloudflare.com/the-epyc-journey-continues-to-mi...
Intel 10nm is maybe more comparable with Samsung 8nm.
We need a better metric, a marketing term doesn’t work.
Is this based on the lithography process, or because of the uarch?
The 7nm = 10nm was from https://en.wikichip.org/wiki/File:7nm_densities.svg, which looked at the density. Based comparison tables of lithography processes, it doesn't even look like power is something that they compare (eg. https://en.wikichip.org/wiki/7_nm_lithography_process)
They're not. Not even close. Especially, in combinational logic they're all over the place.
When verilog is being turned into gates there are many arguments the engineer can give the tool and they will give very different results. It's kinda like picking which optimizations get turned on during compile, like picking speed or binary size, although not exactly like that.
The industry tried using bit/area, because an important advantage of small transistors is more memory/area, but that has basically all the same complications as well so it doesn't really work like one wants.
So just counting transistors can be totally unfair when two designs are made for totally different applications.
In the end picking a node is a complicated, multi-dimensional, decision. Often, and more often than the public thinks, the smallest one is the wrong decision. Note: Google's TPU is at least a generation behind the bleeding edge, if not two, and not because Google can't afford it if they want to.
So, basically expect the marking names to continue for process node names for the foreseeable future.
For the more EE minded, see this deck: http://www.cse.psu.edu/~kxc104/class/cmpen411/14f/lec/C411L1...
If you don't want to go through the whole thing note slide 17 especially. Those are all the same boolean function but they have very different areas, speeds, and power. Which one the tool picks will be (mostly) controlled by the designer.
DAG-Aware AIG Rewriting: A Fresh Look at Combinational Logic Synthesis http://people.eecs.berkeley.edu/~alanmi/publications/2006/da...
Which is the inspiration for this tool: https://github.com/berkeley-abc/abc
Which how this flow does synthesis: https://openroad.readthedocs.io/en/latest/
Makes me wonder why, is it because they see a big future yet in on-prem and can't really embrace cloud computing until that future looks less certain?
I think that for on-prem they faced a treat from containers, which they are embracing by betting the company on Kubernetes.
Their bread and butter is in managing pets and cloud providers expressly don’t want to encourage people to bring their pets into the cloud because people get mad when a few get killed.
OpenStack with the KVM backend is better than VMware at multi-tenancy so there really isn’t much value for VMware to add even over open source state of the art IaaS.
- NSX (via Nicira)
- VMware in the cloud through all hyperscalers
- Tanzu (via Heptio)
Also made security as a core successful pillar for their offering.
They did not have the DNA to build a hyperscaler. They tried vcloud, but it was a failure.
We were getting sub ms latency between applications on the same DC site, across all layers up to application. With VM were we were well over 2ms and at the virtualised network level. With marginal improvements on that front observed after 2 years of back and forth support projects.
So take that concept further of what happens if/when China makes that move. What happens if TSMC's fabs are sabotaged/destroyed during this scenario? This would make today's chip shortage seem like just a blip. Would that be the thing that makes Intel relevant again? If this current shortage actually motivates US based chip manufacturing equal to TSMC, the loss of TSMC would obviously not be as severe. How much would China benefit from being the sole customer of TSMC at that point? Time to crowd source a new scifi/techfi plot
Every time Intel has a "revolution" they wind up pointed the same way, with less reason for hope. They've lived off blocking competition and stomping innovation for so long they have no idea how to do anything else. Kill the rotted zombie corpse and let people with ideas do shit.
Ryzen 3 is a low-end product line (currently 4-core APUs).
When companies like Cloudflare are openly announcing that none of their upcoming servers will use Intel, you have to admit something's wrong.
https://www.theregister.com/2021/09/01/cloudflare_picks_amd_...
> "We evaluated Intel's latest generation of 'Ice Lake' Xeon processors," Howells wrote. "Although Intel's chips were able to compete with AMD in terms of raw performance, the power consumption was several hundred watts higher per server – that's enormous."
Lake has turned out to be Intel's decline into Bulldozer chip style design issues. They might perform fine but they're too hot, too outdated and too expensive.
AMX might be interesting, too, although I currently stick to `Float32`+ precision, so I don't have use at the moment for <=16-bit FLOPS.
Overall, saphire rapids looks very interesting to me. Of course, being built on an upgrade of the same node as Ice Lake means heat/energy use aren't getting better, so perf/watt gains will probably mostly have to happen on the perf side of things.
Tiger Lake clocks much higher than Ice Lake, but I'm not sure how their energy use compares at the same clock speeds. Saphire Rapids is another node upgrade on top of Tiger Lake.
Because of that, it is much less efficient than Tiger Lake, which is made on the 3rd variant of 10-nm Intel process. Tiger Lake is still less efficient than the CPUs made with the TSMC 7-nm process, like Zen 3.
Sapphire Rapids will be made with the 4th variant of the 10-nm Intel process, the same as Alder Lake, now renamed as "Intel 7".
So Sapphire Rapids is 2 generations above the current Ice Lake Server and it should be much more efficient. We still have to see how Alder Lake will behave, but there is a chance that "Intel 7" might be more efficient than the now old 7-nm TSMC process.
However, when Sapphire Rapids will become available next year, AMD will probably soon launch Zen 4, using the 5-nm TSMC process, so it might become again better than Intel.
I agree. Although I think CloudFlare's is a bit more unique than most data centre usage. CloudFlare say it's all about efficiency here: https://blog.cloudflare.com/designing-edge-servers-with-arm-...
Also note that Intel's revenue is only about 1/3 data centre (DCG in https://www.intc.com/segments - I didn't find 2021 numbers).
Island is quite attractive and have a few data centers (Google for sure and IIRC some crypto companies).
But power is important for sure.
If you buy top line Intel processors they are. But reliability, higher speed and lower power consumption used to make up for it.
It’s also important to distinguish CAPEX vs OPEX.
Including Intel, apparently. I still think they were heading in roughly the right direction a decade ago with https://en.wikipedia.org/wiki/Larrabee_(microarchitecture), but then they they seemed to just... kind of give up on it.
These days if I hear rumors that Intel might be about to do something bold, I don't hold my breath. I'm not saying they definitely won't find a way to recover from this current stagnation, but I've been disappointed enough times now that I'll believe it only if and when I see it on shelves and in real world benchmarks.
Admittedly Intel having its own fabs looks good while everyone else is struggling for capacity, but I assume that will be fixed eventually. When it is, I expect Intel's decline to hasten.
Intel are already experimenting with using other people's fabs. I think this is their best path.
AWS-designed silicon has Intel running scared. Ampere is working with the other cloud providers on comparable initiatives.
Intel being obviated in the cloud is a real possibility.
They have grand plans. They've tried these exact same grand plans a decade ago and failed badly.
The culture hasn't changed. In fact the minority that warned of exactly what has happened to Intel - the Cassandras who got it right - have all already been driven out of the company. It's now just one single group-think of fail.
The proof is in the results which we will NOT KNOW for at least 10, maybe 20 years. The signs so far are not encouraging.
Microsoft haven't found a radically better way to sell MS Office, or Windows. It's just leveraged what Mr. Balmer built, an extensive sales network, and monster salesmen who make car salesmen look nice, to upsell big, fat clients into buying grossly overpriced web hosting (Azura.)
The best direct comparison would be if Intel starts force selling their corporate buyers an extremely overpriced Windows clone.
We had a visit by one MS salesman trying to forcefeed us 365, and cloud AD, instead of out self-hosted AD. We almost had him escorted out.
amd still has a good presence in affordable, good performing latpops. but that's it. everywhere else, they're basically only courting the high end. they got better & immediately stopped trying to serve the mainstream markets.
for example, go try to buy a business class small PC. there was a brief second where one could buy a 1L sized Lenovo with a 4850ge chip, and it was awesome, way better than intel. but AMD very very very quickly withdrew from price-competitive markets.
Like literally every single chip that is in shortage is something Intel does not and will not make.. well they do have Altera but still..
Once the semiconductor situation settles down and it all goes back to normal, you'll probably see more availability of AMD on the lower tiers - the situation is already getting better with most of the 5600 (and higher) series of cpus back in stock now.
As for me, ironically, once I could finally get it, I decided that I might as well await the next spin. Chipmageddon appears to have taught me patience.
On top of that, there's a container ship traffic jam for inbound ships going to the US.
There are others. What happens if TSMC has an unchallenged kingmaking node? Do they let AMD/Intel or AMD/NVidia keep their fat margins? Or do they say "the node makes the king -- start bidding"?
Chip shortages and supply chain issues, bruh. Let's not pretend it's not also affecting other suppliers. It's not an AMD issue.