Intel Betting the Farm
semianalysis.com
semianalysis.com
Intel has two legs. Intel microarchitecture business competes with fabless AMD, Apple, Nvidia, ARM, Qualcomm. And Samsung. Intel's chip manufacturing competes only with TSMC and Samsung in the high-end.
Intel has always been the first and foremost chip manufacturing process company. Intel was founded by chemists. The competitiveness of x86 didn't come from superior chip architecture. It came from superior manufacturing and volume. There is the more expected value in risking the whole company and betting everything on high-end fabs, process technologies R&D than just competing with fabless. High risk, high reward.
He's also targeting getting Apple back https://www.youtube.com/watch?v=QsrCOkIth_U
I can still believe things would be different at Intel, because of his long history there, and he seems pretty passionate about it (although maybe he seemed that way at VMWare too?)
His track record at VMware was also my major concern when deciding whether to invest in INTC stock recently. In the end I decided for myself that VMware was probably unsalvageable. They had a dominant position in the datacenter (and still have), but were unable to compete effectively in the cloud as there is too little overlap between the VMware culture, commercial model, and product with what's needed to run a hyperscale cloud. Gelsinger managed to become CTO at Intel in 2001, which is I think is not something you can do if you don't have any leadership skills. Given that the Intel mission is pretty clear, and given that I don't think Intel needs to total cultural overhaul (just the management), my view is that he has a decent chance.
In my personal opinion, this was generally to the detriment of VMware's products (downward pressure to focus on features that made little or no sense, except to give Dell marketing talking points; bad impact on morale by doing things like decreasing employee benefits quality so DellEMC and VMware could jointly reap discounts), played a role in frivolous acquisitions, and generally (and understandably) caused OEMs other than Dell to disengage and keep VMware at arms-length.
Very unlikely to happen. Maybe Intel could end up manufacturing Apple's M* chips, but it seems very unlikely at this point that Apple would ever go back to the Intel architecture - Apple would be ceding control if they went back and they're not going to do that.
Things would have to be desperate to do that.
The good news is Spectre style vulnerabilities are extremely hard to use and indeed have never been observed in the wild. Also, good security practices (only run trusted code on your own dedicated machine) provide full protection against theoretical Spectre attacks.
The only substantive difference is that Intel provides a stronger memory ordering guarantee. This is theoretically an advantage for Arm, but there are disadvantages too, so I'd call that a wash.
Apple hasn't released it's high power chips yet at 5nm. They've only released chips in the very low power laptop range (honestly lower than most any laptop uses today) and the power laptop range (not even high power laptop range). Comparisons will be interesting once we see the Mac Pro heat range M* chip.
It doesn't exist.
That's why people keep inventing hypothetical AMD 5nm parts to compete with the M1 as if it somehow scores points for AMD to win with an imaginary chip against a real one.
The irony of this is that the same AMD fanboys were (legitimately) dismissive of similar arguments from Intel fanboys for years prior to this. "If Intel was on 7nm Zen wouldn't be able to compete!" Maybe, but it isn't, and the only reasonable comparisons are with silicon that actually exists.
I have to come to the conclusion that this statement just a rallying cry. Woo the uninformed that Intel will get Apple back, and try and rally the informed using this crazy claim as a North Star: Intel will become so good in the future, Apple wished they could come crawling back, if not for their pride.
Case in point: 2x ProRes accelerators in the M1 Max.
These are critical to a significant percentage of Apple's core base and yet there is zero chance that we would see them in the Intel offerings.
The only way they could maybe have a chance is for them to have something better than the equivalent M1 5 years from now.
It's not going to happen. Especially with x86
First, Apple has a history of dipping its fingers into several cookie jars simultaneously. They source from different suppliers to introduce supply chain redundancy and establish pricing power. This way, they ensure that a transition from a main supplier company to its competitor is always viable, because they build relationships and allow the competitor to generate enough revenue to stay active in that particular product segment. So even if Apple wants to make its own chips, it has a reason to do so at least in part with Intel as the IDM 2.0 push towards foundry becomes a reality.
Second, there are still some significant drawbacks to Apple's new machines. I want more external monitors, for one.
This is neither here nor there, but personally I think Apple should have spent $500 billion on its own chip fabrication infrastructure rather than doing share buybacks over the past ten years.
“ Overall, the new MacBook Pro with the M1 Pro chip supports up to two external displays with up to 6K resolution at 60Hz, while the new MacBook Pro with the M1 Max chip supports up to three external displays with up to 6K resolution and one external display with up to 4K resolution at 60Hz”
More than this?
Experiencing 120+ Hz spoils you. That said, ask me to tell you which screen is 120hz and which one is 165hz in a blind comparison, coinflip. Versus 60hz, just scroll a web page and you'll know instantly.
If the limit is 6K at 60Hz, you'll be able to run 4K at >60Hz. My M1 MacBook Pro runs a 1440p monitor at 165Hz just fine over Thunderbolt 3.
Weird edge case, I know, but: using Thunderbolt 4, what is the maximum theoretical refresh rate I can get from a 480p display? 1000hz?
I share the sentiment that it's unlikely that Apple will go back to x86 any time soon. But if/when Intel has a leading node, it think it is actually /likely/ Apple would manufacture their M processors on it. Intel is saying their nodes will be easy to port from TSMC, having two vendors gives Apple leverage, they reduce geopolitical and other supply chain risks, and and the two companies are literally in each other's back yard.
Eh. Right now Apple is matching a (desktop) 5950X at laptop level TDPs - the iGPU is way lower than a desktop 3080 but so is a mobile 3080 at 115W. If you doubt that number, a Pugetbench result just showed up. TBH the specifics of the"which processor is 5% better than the other here" slapfight doesn't actually matter here, compared to the fact that Apple is doing it in what, 60W? And the AMD+NVIDIA system is using a 300-350W GPU and a 142W PPT CPU. Yeah, that's high-clocked kit, but with results like that there's zero reason to doubt Apple's claim of it beating a (mobile) 5800H (probably sustained 62W PPT) with a 115W mobile 3080.
M1 Max: https://www.pugetsystems.com/benchmarks/view.php?id=60176
Desktop 5950X/3080: https://www.pugetsystems.com/benchmarks/view.php?id=60613
pugetbench premiere pro definition: https://www.pugetsystems.com/labs/articles/PugetBench-for-Pr...?
https://i.imgur.com/dpBHGxN.png
Apple has never fibbed or bent the truth on their marketing benchmarks - and they have sandbagged before, the numbers have sometimes been higher than presented. They have nothing to be ashamed of here, this is an extravagant chip, it's twice the transistors of a 3090 to make your desktop 3060 ti / mobile 3080 run at 45W, and a near-fastest-in-market product in 15W for the CPU, enough to saturate the GPU for most things you'll want to do. In order to do that, they put the equivalent of an 8-channel DDR4 bus on it (it's 16-channel on DDR5 and DDR5 is significantly faster) and then used tiny LPDDR5 packages they could stack. Oh, and you can allocate any of the system RAM as VRAM, so you can do large-memory VRAM tasks like machine learning (albeit slowly of course - it's desktop 3060 Ti performance). For a portable workstation it actually pretty well owns.
Also the regular ol M1 had as much cache as a 9900K. For its 4 performance cores + 4 efficiency cores. The whole A15 design is just an extravagant display of transistors, Apple spared no expense, it's just a design exercise in "what if we built it ourselves and ran up the scoreboard and called that our 'external equivalent' expense". Apple will spend $300 to build the meme 5nm (nearly) icelake-SP-sized mobile workstation chip because a 5800H or 12Gwhatever would cost them $200 anyway and this thing dunks on those processors. Vertical integration matters at their scale, just like Google and Amazon and others.
With Zen4 on DDR5 and TSMC 5nm (actually N5P - it's a better node than what Apple is using for A15/M1) next year we'll see how true that actually is about it "all being node advantage".
I bet AMD can catch up to laptop-power-budget A16 (the next-gen Apple core that Zen4 will compete against - also on N5P) in performance with 142W PPT processors ("105W TDP") like their 5950X equivalent, I'm kinda doubtful they can close a factor-of-3 gap in perf/watt at laptop-level PPTs.
Yeah, Apple is clocking a lot lower and that's where some of the efficiency comes from... but look at the IPC difference here. They're matching processors that are running roughly twice as fast, at peak efficiency clocks. It's an insanely wide architecture, and it's hard to scale x86 like that - x86 instruction decoding has bad transistor-order-complexity for higher widths, due to variable instruction lengths and other problems. You can make all the execution units you want in the core, but it's hard to keep them fed if you can't decode enough instructions. x86 has already done a lot of the obvious tricks like instruction caching/etc.
I doubt he seriously is, I think that’s just rallying the troops by setting an aggressive, aspirational target, acknowledging the humiliation and trying to use it.
Which is what he should be doing so good for him.
It remains to be seen if he’s another Lisa Su. The odds are against it but I’d be excited if he could pull it off.
Independent Fabs would be ultimately able to compete with TSMC on equal terms - in fact on better terms for some clients as based in US. The fabs would be a growth business and x86 a cash cow.
Going back in history the move which Grove is remembered for - ditching memories - was partly about improving focus. This strategy is doing the opposite.
A far cry from betting the farm, really.
You have to ask Intel shareholders who have been used to a very reliable cash generating business that buys back a lot of shares, and pays nice dividends to use up 9 billion of their money in an year in the hopes that things turn around some time in the future.
I am not sure what exactly constitutes betting the farm, but they are getting into some very risky territory.
The thing with market cap is, most of it is decided by the last trade price, which can be very volatile. With deep order book on index and household name like Apple, that volatility doesn't show up until a black swan event like it did in March 2020.
https://www.apple.com/newsroom/pdfs/FY21_Q2_Consolidated_Fin...
marketable securities are volatile and don't hold their value exactly when there is a liquidity crisis. Even including these securities, totals to about $60B quite shy of either $200B or $1T
The equivalence standard you used is the accounting standard, which means you can get the exact amount in cash overnight if you so desire. It's too strict when considering what financial resources you have in making long term capital investments.
$9B is still small against Apple’s $200B, but fairly large against Intel’s $34B. So yeah I’d be scared.
I think Intel has some rough times ahead, but there are extra-market forces at work.
On the other hand, several of those nodes have been running in production elsewhere for a while now, so it's not quite as crazy as it sounds. But, highly ambitious.
Amazon has their own chips for their servers, Microsoft announced they're working on their own chips, Google just released their first consumer chip this week, and the M1 is crushing the competition.
I just don't see Intel succeeding with all of these things stacked against them. I hope they prove me wrong as more competition in the space is always good, but as it stands I'm not optimistic for their future.
AWS Graviton 2 ARM instances offer better performance/price than their intel instances. Does that qualify as viable? Or do you mean AWS has to make their own because there isn't one "off the shelf"?
This was what I meant. In this case, AWS is selling general-purpose compute, so there's minimal iDevice full stack synergy, and they're just looking for performance, power use, and total cost.
Intel has a long history of selling complete platforms, and they had a major role in disrupting computing in the microcomputer era; unlikely as it may seem right now, they have the power to turn the cloud story upside down and support repairability, client compute and decentralized systems through a radical new platform, with their new process nodes used as a hook. They may need to go lobby their case with governments to get some rule-making in place that drives a wedge in the existing monopolies, but they have the pockets to make a run at that, too, and it's a slam-dunk PR win to be seen protecting end users. Their endgame would be to once again sit at the top of the market, holding most of the customers, but it would be a very different market, and reflective of the kinds of broad transformations that are now happening to industry and supply chains across the global economy.
Intel is unfortunately another IBM/HP at this point. They will likely continue to exist in some form but their glory days are behind them.
These companies care a ton about how long it takes for capital expenditures to bear fruit. Semi fab is one of the longest-term investments in the technology industry. Even a company as large and established as Intel still thinks like a company, not a government.
Why is there a disincentive to spend on long-term investments? Look at INTC stock today and you'll understand.
On the one side, it's heartening to see the investment, but on the other... surprising to not see any aggressive market-making / market-winning moves. $40B buys a lot of PhDs.
Ultimately it sells chips and services for running software. Nvidia originally made chips for video games, now it is part of self-driving cars, data centers, and even genomics. They are one of the biggest AI powerhouses now and not by accident. Google and other hyperscalar sw+hw seems to be competing more w Nvidia than Intel at this point.
AMD way underinvested relative to Nvidia and we saw how that played out - I'm surprised Intel is doing the same. This stuff isn't secret.
Competing chip makers are showing how to grow on worse process technology by investing in software codesign and vertical codesign -- smarter designs, heavy compilers, heavy ai, heavy solutions, and going hard on the bigger markets within that. Likewise, the Google hw moonshots are more interesting in practice and scale, afaict. Intel knows this, and I'm sure there are clever internal proposals for other orthogonal approaches. But like AMD, even if that stuff is happening, these statements suggest it will be without muscle.
This is like a worse repeat of how they approached the rise of mobile devices. I want them to do well, it's frustrating to see them under leverage so much $ and talent year after year when everyone else is doing so well.
Building slightly better chips with light software support is 80s/90s thinking. They missed the wave in mobile, and I don't see the new CEO doing anything about using their position or $ to start competing differently. Jensen is inspiring in that he doesn't fight the last decade's battle: their architectures etc. keep getting better, but just as important, the market approach keeps evolving too.
2021 revenue projection is 73.5B, while 2022 is projected to be 74B.
Checkout the official earnings report. https://d1io3yog0oux5.cloudfront.net/_150ab36dee7b2e1e66231d...
I guess that's "poor" for the board members. But, since the new CEO is an engineer, I have high hopes for Intel. I've always liked Intel, and I think they finally realize that they need to invest more and push more to compete against AMD.
Gonna be interesting to see what comes out over the next several years.
First, it's a seller's market, there's no one who can replace Intel's chip output. They could have stayed on 14nm and still sell all their stock. Second, the US government is clearly not liking where the chip market is, and is going to spend a lot Intel's way. So income is guaranteed for that time.
The real threat is in the longer term. If the company loses technical relevance entirely, and loses marketshare enough the market can replace them. Then the US gov pulls away and Intel becomes Via or GlobalFoundries.
The current strategy is exactly the right one to prevent that: spend their current capital to be more competitive and regain marketshare. Meanwhile open up the foundry - both to be more receptive to external developments, and get easy money in the car sector.
IMHO, there's no reason Intel can't pull this off. Unless management is entirely oblivious, and this one doesn't appear to be. Intel's scale is large enough to innovate, and they've had fat years to build up a cash reserve even before Uncle Sam gives an infusion.
The previous position is unattainable, but Intel can remain a major player in the years ahead.
I know the Stockholm syndrome is real but try to not fall for this guy's bullshit.
Part of the issue is that they become a literal blue chip company which means paying a dividend and not focusing on growth. It creates a different company culture that can be hard to get out of.
It's hilarious how releasing the same product for a decade translates to "struggling to improve" all while making more and more money each quarter.
I'll leave this here for you and the people who are downvoting me just in case you're not simply fanbois.
https://jolt.law.harvard.edu/digest/intel-and-the-x86-archit...
whats an edge if not a leading edge? i guess the strange expression is accidental? that or im just ignorant of the english language
Bezos should buy the top say 6 levels of Intel management, and give them a new company whose sole focus is to keep any further advances in space travel from happening. That seems to be a high priority goal for him, and if the Intel folks can make the business pivot that's their jam just in a different market segment.
The only way to extract any of the value still extant inside Intel, that I can imagine, is a deep breakup where the resulting entities are small, focused, single purpose shops... In other words, I don't see a way. Some Liquidation / Firesale move would be better and keep the foreign companies off the first auction.
no
>The only way to extract any of the value still extant inside Intel
Intel is still a massive company with fabs in a world with chip shortages. If your car chips that aren't available such that they're shutting down car factories, are then made on 14nm+++ no one should be concerned.
Intel has plenty of runway to come back to competitiveness.