They used to be a crown-jewel of US tech. But it seems like every time I read the news, they are announcing a delay or shutting down some product.
Defense is starting to get a blank check with fairly bipartisan support for the first time in at least 30-40 years and it's centered on semiconductor supply chains. There has never been a better time to secure the fucking funding, have ASML send twice as many people as they already have, and power through it. The market is whatever you want and the margins are whatever you want: in a functioning system? You fucking do it.
And while I will believe that Intel has suffered serious attrition in key posts, there's no way that the meta-knowledge of how to debug "we don't have the fabs running right, who do we hire, what so we need to give them to get it done" has evaporated in 5-10 years from the singular source of this institutional muscle memory in the history of the world.
The failing here is more like a failing in courage, or stamina, grit, something. It's a failure of the will to do the right thing for both the shareholders and the country.
They've been doing the exact right thing for the shareholders: squeezing the living shit out of an asset (x86/64) for decades while cutting anything interesting or competitive to the bone to give shareholders more money. Money spent on something that could really have been competitive is money not sent to the retirement fund that keeps John and Jane Q. Public swinging in more ways than one at their golf course retirement community in Florida.
The problem is, you can only do that for so long. There is a minimum spend to remain a competitive company with regard to being able to market products to consumers. Executives don't have a fiduciary duty to create the best possible product for consumers to look at and potentially buy in the marketplace, but they do have a fiduciary duty to shareholders to meet an earnings projection. If these two activities can coexist peacefully, great. If not, the first activity stops while the company gets gutted.
The most expensive, highest-margin, technically advanced and risky business in the world is for investors who want that in their portfolio. If they wanted to milk a dying industry on the way down they would go buy Disney stock.
It is very clearly in the interests of long-term investors in Intel to maintain a commanding position in fabrication: it's been the secret sauce of the company since the very beginning, it's never been more in demand.
This idea that companies are obligated to do what will deliver some little bump in the stock price in 90-180 days is everything from not how the rules work to just a lazy meme for people who don't want to earn their princely salaries.
Don't make excuses for weakness at the top.
Different market participants will be trading on different signals, sentiments, or theses, and this will influence everything from the order types they use to the hold time of the instruments in question.
But one of many things they all have in common is that they know that what other people think about the future affects the price right now: an intuitive proof of this is that if some major announcement is made about e.g. trade policy, and the market deems it credible, you will see instruments transact up or down in price immediately.
In any effort to go deeper, one must be wary that this goes from market microstructure to Ito calculus to voodoo real fast, a closed form solution would be an infinite money machine! But a reasonable jumping off point might be the notion of a Keynsian Beauty contest: https://en.wikipedia.org/wiki/Keynesian_beauty_contest
The TLDR is that hold times across like 13 orders of magnitude (rumor has it a cutting edge FPGA fielded by someone like Optiver or HRT or Virtu can pull a whole ladder in 20-50ns glass-to-glass, so whatever Buffet does divided by that) the market still in some sense reflects expectations about the future: it's "priced in" in trading parlance.
"Too late" has some ways you can use it meaningfully in finance, but it's not in the sense I take you to mean above.
Inflation-adjusted INTC [1] is the same as it was in 1997, including dividends! Shareholders have no real return from INTC for almost 3 decades.
I would quibble with the exact right thing phrases but otherwise agree. Intel indeed followed a formula which is intended to and often does produce massive return for some time frame. The formula is indeed "gutting the company" - squeeze every part of an enterprise and return the results as profits. Whether destroying the companies long term prospects is worth these short term profits is a complex calculation.
A managers' duty is to promote long term value and stability, actually, but return enough short term profits and you trump that long investment income.
And in the days when fraud or "fraud-adjacent" behavior carried serious costs? When violating the social contract around pensions and severance and stuff had real reputational costs that followed the principles around? People used them when necessary. You sold off the assets sometimes.
But beginning with the LBO "innovation" in the 80s and running a line through Milken and shit all the way to the Vanguard/State Street/Blackrock "quasi-sovereign" level of PE asset capture?
People started arbing it, not by seeing value where others had missed it! By betting that Gordon Gecko had enough fans to make the arb work. "Gutting" a company slowely and painfully is in a bucket I'll call "fraud adjacent": usually not outright illegal (lotta "gray" work, gray edge), clearly not what society wants or intended, and you know it's a scam when you look in the mirror every night. i used to get wasted with these guys at Catch when I lived in NYC: they'll tell you everytging I am and more on five gin and tonics
There's no place for the word "duty" in any version of that argument unless you also use the word "derelict".
Don't excuse weakness at the top.
By betting that Gordon Gecko had enough fans to make the arb work. "Gutting" a company slowely and painfully is in a bucket I'll call "fraud adjacent": usually not outright illegal (lotta "gray" work, gray edge), clearly not what society wants or intended
Sure, but if society doesn't intend this, society has to f---ing do something. Clearly it's not.
IE: if the only legal obligation of management is promoting maximum legally possible valuation over time, "squeezing" still makes sense if there's enough money to be had. The only way to change this is forcing the issue in some fashion or other.
Clearly not as they've ended up suspending dividend payments altogether.
At a certain point, you hollow yourself out, and you can't recover. Top-level talent doesn't want to work at a place that doesn't have a real shot. So things just... peter out.
Usually opposing parties have had the common sense not to immediately hit the undo button once they take office. E.g., Biden leaving most of Trump's previous nutty tariffs in place. But "common sense" isn't on the agenda these days. We are, to all intents and purposes, under attack from within.
In 1998 Meriwether and the rest of LTCM nearly crashed the economy, needed the Fed to get involved, and they were personally ruined, guy never opened a ten thousand dollar bottle of wine again and probably never had anything again. Shortly thereafter, Jeff Skilling took out offices in 9 cities and pension plans all over the country with shady accounting. 24 years in prison (reduced later to 14). Ebbers/Worldcom 2002: died in prison.
By 2008? Zero prosecutions. Bonuses the next year.
Around the same time Clinton got caught lying about chasing (consenting and of age) skirt in the Office: nearly ended his presidency, definitely ended his policy agenda, real consequences and he caught a shooting star to avoid far worse. The public was not going to accept it, Congress was not going to let it slide on either side of the aisle. Today? Something like that barely makes the press. You have to be accused of sex trafficking to even get an investigation started and everyone will probably walk.
The idea that this became uniquely bad in January, or even 2016 is demonstrably untrue. At some time in the last 30 years we started accepting leadership who are dishonest, nakedly self-interested, lie without consequences, enrich themselves via extraction rather than value creation, collude with no oversight, and sell out the public.
This is a completely bipartisan consensus on these norms. Speaking for myself, I think Trump represents a new low, but not by much, he's just the next increment in what history will probably call the Altman Era if his ascent to arbitrary power on zero substance continues on it's current trajectory.
Reasons for that are easy to come up, imo chief among them being web2.0 (social media) and the ever increasing degree with which people exaggerate everything just to get a reaction.
Under that context, what's a little skirt chasing compared to what people usually say about the politicians? And how are you gonna remember he did something a few months ago, when so many more extreme things have happened since?
Really, I feel like social media will be considered the most destructive force to society in 20-50 yrs
The thing about those sources is that for the most part, it wasn't really economically viable to alienate half the population by leaning hard right or left. Any reduction in audience would likely translate to a commensurate reduction in advertising revenue.
Today, there are many, many sources of 'news' available in various forms around the internet, and of course people are free to choose what to pay attention to. This means it's entirely feasible for each source to cater to a particular viewpoint, even at the expense of definitely alienating half or more of the theoretical potential audience.
I theorize that the reason for this is that people have voted with their feet, balanced sources aren't as profitable and that's why there are fewer of them. It makes sense, a more balanced take on events is by definition not nearly as sensational, and almost always requires more mental effort on the part of the listener.
That by itself would probably be enough to explain the current situation, but on top of that, we also have the fact that many people receive the above mentioned information via algorithms designed to feed them more of what they already like (i.e. agree with) and nothing else, which of course only amplifies the effect further.
I have no idea how we get out of this situation (or if in fact we will), but in my mind it's not surprising at all.
I think the internet just supercharged a change that was already well underway.
I usually can think of at least a few plausible/possible solutions to most problems. But I am not at all sure what the Democrat's right response should have been.
However, a severe lack of legal tolerance for businesses that use technology to super-scale poisonous conflicts of interest, like surveillance backed ads and media feeds algorithmically manipulated for addiction/attention behavior would have been part of it.
Zuck should have been put away for life a few accidental genocides ago. (IMHO)
Sometimes I find myself thinking about that experiment with the perfect rat paradise. The overpopulation got so bad, the normal social functions of the rats started to break down and the rats started acting like sociopaths. Sometimes, I think that's what we're doing to ourselves by exposing the average human to millions of voices through the internet.
Of course, ironically, I'm ignoring my own advice and still engage with the Internet. Though I mostly keep to HN and some IRC.
It was just as wrong as predictions about human overpopulation like Malthusianism
Like there are any number of extremely specific issues which are not "screaming and posturing" unless you're dead set on not talking about them.
In other words, I completely agree.
The child sex ring that was uncovered in 2008 resulted in ludicrously light consequences and then after a repeat offense in 2019 was systematically ignored until now.
The same President Clinton, with the same Congress, repealed the Glass-Steagall Act in 1999 and thereby effectively legalized much of the next 26 years of corporate malfeasance, starting with much of what enabled the 2008 Global Financial Crisis. And when they did this, it was popular and politically nonthreatening.
We got what we voted for.
I don't think Al Franken would agree with this
They heavily pushed the idea that the opposition could not have legitimacy. Gingrich did it through the exercise of power and Limbaugh did it on the airwaves. It wasn’t just that the opposition was wrong or bad for the country, standard democracy stuff, but that the opposition had no right to hold power at all. Once you start thinking that legitimacy is based on which side you’re on rather than who you are or what you do, you won’t care about bad leadership as long as it’s yours.
Trump's entire rhetoric relies on this tactic. Anyone who disagrees with him or tries to shut him down should be impeached, jailed, whatever, because they shouldn't be allowed to exercise their power against him, no matter how legally they wield it.
It just makes me so angry to hear Vance say things like "judges aren't allowed to control the executive's legitimate power". Yes, they are! That's literally one of their jobs, specifically enumerated in the constitution! But that's the tactic: train people to believe that the judicial branch is not legitimate when it comes to executive branch decisions.
Really? Because:
> During Donald Trump's 2025 speech to a joint session of Congress, the president asked House Speaker Mike Johnson to “get rid” of the subject act.[190]
* https://en.wikipedia.org/wiki/CHIPS_and_Science_Act#Subseque...
What has much more commonly produced good outcomes in such situations is robust public-private partnerships like the ones that produced the semiconductor industry in the first place. Run the list of innovations in strategically key technology and what will you find at one remove in every instance? The DoD, NASA, the Labs and ATT more broadly, the university system.
It's always a public/private partnership during periods of explosive value creation when the stakes are high, and it's always a private sector capture orgy during periods of extractive stagnation like the present.
That era of American history has passed. Innovation gives way to consolidation and cronyism. Think Mussolini’s Italy.
A lot of cash is going to be redeployed towards tilting at the windmill of domestic ballistic missile defense… while we’re watching a war in Ukraine where cheap drones are demonstrating that most major weapons platforms are functionally obsolete.
Forums like HN full of senior technologists and future founders are disproportionately high impact. If the tone around here shifted a little to stop excusing what YC has become and start embracing how it all started?
Shit like that adds up. geohotz had that post a few weeks ago about this late capitalism internet shit, he was pretty deep in with the Effective Altruists and he got it together. I said at the time and I'll say again, you get a few more people like that to sober up? pmarca and lex and people? Maybe even pg?
Real change happens that way.
The current board is a pack of cargo-culting epitaph writers.
I forget the name of the speaker guy who has this turn of phrase, but whatever the merits of his overall platform this hits perfectly: "People doing well today are using every means at their disposal to decrease their accountability while increasing their compensation. If you don't compensate people based on the responsibility they are willing to undertake, you will get a world run by people like this and it will look like the world you live in right now".
It's also noteworthy that GAAFET being a complete redesign of major parts of the manufacturing process levels the playing field significantly. A big example of this is Japan's Rapidus which was founded in 2022 and has managed to invent (and license) enough stuff to be prototyping GAA processes.
Intel's 18a process seems to be quite good. It's behind TSMC in absolute transistor density (SRAM density seems to be the same as N3E), but ahead on hard features like BSPD and maybe on GAA too. I suspect that they didn't push transistor density as hard as they could because BSPD and GAA tech were already big, risky changes.
We'll have a much better idea of Intel's fab future with 14a and 10a as they should show a trend of whether Intel's fabs can catch up and pass TSMC or if they run out of steam after the initial GAA bump.
TSMC by collaborating with many different customers with different needs had a lot of insensitive to not just create powerful tooling for one kind of CPU design approach but also being very flexible to allow other approaches for other needs. And AMD has repeatedly interrelated on their whole tool chain and dev. processes for many years while Intel was somewhat complacent with what they had.
And a bunch of the recent issues with CPUs internally dying sound a lot like miss-design issues which tooling should have coughed (instead of looking like fundamental tech/production issues).
They were wasting a ton of time and effort eagerly trying to convince Apple to put IA into phones despite obvious failures to deliver power-effective chips (Atom being the result of these efforts from what I understand). They were spending a lot of time and money trying to start up like a junk ware app-store thing for PCs that they could use OEM relationships to peddle, as if the PC ecosystem belonged to them the way that Android did to Google or Apple's ecosystem to Apple, not realizing that if anyone has that power it's Microsoft (but they also don't).
It was pretty shocking coming from a hacker/cyberpunk culture where everybody had been dunking on Intel designs for over a decade. (I personally had been waiting for an ARM laptop since around 2000.) A lot of leadership I got to interact with were business/people-people types that truly seemed to believe that the best product boiled down entirely to social perception of status and has zero basis in reality. Basically the company seemed to be high on the Intel Architecture's accidental monopoly over personal computing thanks to PC-WinTel becoming so dominant (and Apple's later capitulation) and seemed to believe that it was all because of their "genius" Intel Inside marketing campaigns (which were pure social status signaling, but with an effect of avoiding price competition with lower-cost IA rivals AMD,Citrix,VIA and holding power over OEMs rather than being responsible for the market situation around IA in the first place).
Maybe something in the Hillsboro/Beaverton area's water? Both they and Nike seem to entirely consist of a diet of their own farts.
Of course, one could have done an ARM Linux device at any point in that timeline, but using efficient software is apparently cheating.
agreed, but that was often not necessary a hardware issue but a ecosystem issue and Intel executives maybe not seeing/realizing that is pretty incompetent
On one side you had the whole windows was absolute garbage on ARM until very recently, and needed Apple to show them how to have a low friction support extension/transition. And if you instead shipped it with Android or Chrome OS it supposedly didn't count anymore (except a lot of non tech afine consumers have replaced home desktop/laptop with a tablet anyway (cheaper and does everything they need)).
On the other side there was a best technical fit/best customer fit mismatch. Best customers where tech enthusiasts which want to try out new things and can live with a bit of friction (if it's small enough) and are also often willing to pay _slightly_ more. But the best price/product fit is the low (initially, then to mid) end devices except they aren't really that interesting for enthusiasts and due to low (initial) production quantity also not necessary that cheap either and for the people which normally buy this devices buying a similar priced android tablet is most times just better and with a bit of effort you can get an even better x86 PC, through with many 2nd hand/hand me down parts.
and outside of 1) means to pressure MS for better deals, 2) Steam Deck/OS, there just weren't any meaningful large/well known hardware producers shipping with Linux (yes Lenovo and Dell do care (do they still? idk.) for Linux compatibility in _some(few)_ of there expensive business focused lines. But outside of exceptions in 1) don't ship with it so no "normal" consumer pics it up, and Linux shipping ORMs are on the larger consumer market picture just too small to make a big difference. So ARM Linux stayed relegated to niche, too.
The fact that they can't use their own fab for 30% of their products [1], all of which are those that require power efficiency and compute performance [2], suggests it is not overstated.
[1] https://www.tomshardware.com/tech-industry/intel-will-keep-u...
[2] https://www.extremetech.com/gaming/intel-is-using-tsmc-4nm-f...
There is talk about the next version of Arc using 18a. If it does, I'd expect Intel to move that generation's compute tiles to 18a as well.
Apple farms out its displays to Samsung, a competitor. It's just how business is done.
Apple just recently moved back into the hardware space after farming everything out since the iMac gen2 days. Hell, I remember the Mac clones. I miss Power Computing.
Samsung's competition is Huawei, Xiaomi, Oppo, etc.
18A is canceled for foundry customers, it's not going to save them. If they can't get it together for 14A, they are toast.
It seems difficult to figure out if they are getting back on track, though. They always seem to just be a couple years from finally catching up to TSMC.
But so far nothing of the sort has happened for a long time. If feels like ever since Ryzen landed, they have been desperate to catch up but keep tripping on themselves. Losing Apple, while inevitable, has made them look even more irrelevant. They still do decent stuff for the most part but there isnt anything really exciting.
I do like what they are doing with Arc GPUs but it is clear those are loss leaders and it isnt really gaining that much traction.
Alas, this is a story where we will have a better understanding in five years from now.
I would be very surprised if 14a and 10a comes out soon enough to be competitive with TSMC.
It's not about how soon 14a and 10a come out, but rather about how good they are when they arrive. 14a will be competing against TSMC A16 in late 2026 and 10a will be competing with TSMC A14 in late 2027. The measure of Intel's success will be whether they are gaining or losing vs TSMC.
On the customer front, I think customers are probably necessary to offset the ever-increasing R&D costs and an extra year or two to work on making their libraries more standardized may be best for everyone.
I dont know if I count, but at least I wrote about TSMC before most if anyone knew much about TSMC. Which is when Apple brought them to spotlight.
It depends on how you define or count as being able to compete with TSMC?
If Intel technically leapfrog TSMC and their 18nm is better than TSMC 20nm this year but;
It is 30 - 40% more expensive.
It has lower Gross margin, or even negative margin.
It has much lower volume and capacity.
It is slower in ramping up capacity for future capacity planning.
It has limited IP range for its foundry.
It has less packaging options.
It does not have other high speed, low power or analog node options.
At what point does it count as competing? Because right now there isn't a single metric that Intel Foundry is winning. And they are feeling exactly the same as Global Foundry or AMD when Intel Foundry advancement is getting all the oxygens.
And even if they did, with a magic wand got them to compete with TSMC on every single one of the item above, in medium to long term there isn't a single chance Intel could compete with their current board and management.
TSMC leadership and management team is Nvidia's level great. I cant think of any other tech company that could rival them. Their only risk is China.
The first year of TSMCs latest process goes to Apple. And the second few years are booked completely full. There is room for Intel if they can just get in the ballpark of TSMC.
Think you mean 1.8nm, aka 18A. We're way past 18nm and 20nm.
"For a long time, gate length (the length of the transistor gate) and half-pitch (half the distance between two identical features on a chip) matched the process node name, but the last time this was true was 1997"
Or not. Sometimes it if looks like terminal decline, it simply is terminal decline.
These days iGPUs run pretty much any game I care to play so it doesn't matter.
Now we are in a different situation. There are several big competitors using ARM instead of x86. The software world is actually transitioning away from x86 in masses. Apple does their own CPUs better than Intel. AMD outsourced production already. Everybody is pumping money into TSMC who are are already ahead of Intel and they are moving faster.
Either Intel gets a really really lucky run with their new technology or they need to split off the foundry business. The government may put it on life support until TSMC themselves may run into serious problems.
The better way into the future may be to split up TSMC in multiple redundant and competing companies.
Indeed that's how they were marketed where I worked (Office Max) and were priced and spec'd comparably to the Celeron based offerings from IBM, HP, and Packard Bell.
Another issue with the K6 line was they were always a generation behind at a time when Intel was rapidly rolling out technologies like MMX and SSE. Intel coordinated with software manufacturers and had launch day examples that presented significant performance gaps between the CPU lines.
The K6 also had a shorter execution pipeline than Pentium so it struggled to hit 400mhz when Intel was approaching 500mhz. That's why the Athlon was such a shock because it arrived at 700mhz and stomped everything.
Looking back at the K6 line now, they likely perform far better then they did at the time because software eventually got around to supporting the hardware.
The 650MHz came two months after than, and 700MHz another two months later. 6 months later 1GHz! It is easy to forget just how rapid performance increased in the late 90s.
I remember it was a K7 700 because it was the first from scratch PC that I ever built. Everything before and probably since has been a Ship of Theseus.
"AMD Athlon 500-600MHz (bulk) price display. The product is scheduled to arrive in mid-July, and reservations are being accepted. However, there is no specific arrival schedule for compatible motherboards yet."
"the K7 revised "Athlon" has been given a price and reservations have also started. The estimated price is 44,800 yen for 500MHz, 69,800 yen for 550MHz, and 89,800 yen for 600MHz."
Those were Pentium 3 450-550MHz prices.
A week before official AMD shipping date retail Athlons arrive in Japan https://akiba-pc.watch.impress.co.jp/hotline/990813.html
"AMD's latest CPU "Athlon" will be sold in Akihabara without waiting for the official release date on the 17th is started. All products on the market are imported products, and 3 models of 500MHz/550MHz/600MHz are on sale. The sale of compatible motherboards has also started, and it is possible to obtain it alone, including Athlon"
https://akiba-pc.watch.impress.co.jp/hotline/990813/p_cpu.ht...
~$380-800 depending on speed.
https://akiba-pc.watch.impress.co.jp/hotline/990813/newitem....
Picture of one of the Akihabara stalls full of CPUs being sold retail before official AMD launch date :) https://akiba-pc.watch.impress.co.jp/hotline/990813/image/at...
For reference in US 4 days later on August 17 Alienware was merely teasing pictures of Athlon system https://www.shacknews.com/article/1019/wheres-my-athlon According to Anand "OEMs will start advertising Athlon based systems starting August 16, 1999" https://www.anandtech.com/show/355/24
By January 2000 prices corrected to saner bus still delusional levels https://akiba-pc.watch.impress.co.jp/hotline/20000108/p_cpu....
K6-III/450 14,550 $140
K6-III/400 8,980 $85
Celeron 300A $57
First time Duron shows up in Akihabara is June 17 2000 https://akiba-pc.watch.impress.co.jp/hotline/20000617/p_cpu....
Celeron 533A 10,570 $100
Duron 600MHz 9,990 $95
K6-III/450 24,800 $236 haha whats up with that price? Either AMD stopped shipping already and its leftovers or its a sucker tax for ss7 owners wanting to max out.
K6-III/400 14,800 $140
K6-2/550 7,949 $76
K6-2/533 5,970 $57
K6-2/500 5,350 $50
Week later https://akiba-pc.watch.impress.co.jp/hotline/20000624/p_cpu....
Celeron 533A 9,980 $95
Duron 600MHz 9,480 $90
K6-III/450 24,800 $230 AHAHAHAHAHA
K6-III/400 15,800 $150
K6-2/550 7,940 $76
K6-2/533 6,700 $63
K6-2/500 5,300 $50
Looks like by the time Durons showed up nobody was bothering to stock K6-3, only 3 vendors in Akihabara had them. Those crazy prices werent limited to Japan, Poland September 1999:
Pentium III 450MHz 1260 $308
Pentium II 400MHz 943 $231
Celeron 366MHz 348 $85 (300A missing from the list, but was still available and selling cheaper)
K6-III/450 1108 $271 HAHA
K6-III/400 877 $215
K6-2/400 397 $97 haha
K6-2/350 230 $56
For a brief moment in 1999 AMD pretended K6-3 was equal to Pentium 2/3 and tried to price it accordingly but market corrected them swiftly. There was a 1/3 performance gap between K6-3 and overclocked Celeron.
https://web.archive.org/web/20080418185205/http://arstechnic...
https://web.archive.org/web/20070918073530/http://arstechnic...
https://web.archive.org/web/20070918135927/http://arstechnic...
Yes, building top-of-the-line CPUs is hard and it's impressive that we saw the dominance flip in the course of just a few years.
But I think frontier chip fabrication is a bigger juggernaut than "mere" CPU design.
(Your conclusion could still be correct, but I don't know if I buy the high-level reasoning).
“If Intel can just get this next node they’ll be sitting pretty” is what people have been saying for over a decade isn’t it?
Just getting the nodes working and producing enough chips has been a huge issue for them, let alone having good chip designs on top of that.
“No one got fired for choosing Intel” has stopped applying. They’re even losing server marketshare, which was their rock.
My most recent computer is AMD Ryzen based, but we just bought an Intel-based Dell for my partner because the price/performance was better than comparable AMD machines at the time, possibly due to a sale. But the Intel chip is a lot faster than my laptop, so now I'm a little bit jealous of the Intel machine.
it's almost apples to oranges in most cases.
Maybe it is just bad luck on my end, but I have not had those issues with Intel in the past or currently.
I have 2 intel/dell laptops and thinkpad/amd 14s laptop. Both Dells (a workstation-class 22 core cpu and a more power-efficient one) suck massively when compared to amd ai-something-something-ryzen.
What's worse, intel drivers are a mess on linux right now. Dell xps 13 plus is the worst laptop I had in a decade, and that's after owning every Linux-preinstalled Dell XPS 13 ever released.
Not really sure what you mean by that.
Both our Intel and AMD computers are doing great. Nothing "weird" about it.
No problems at all. YMMV.
FWIW, Apple M4 Max 16 Core scores 43,818 on passmark and runs at 90 Watts TDP, so Intel certainly is competing on speed, as well as TDP.
Intel as a brand may survive in some shape or form but it's not looking good for the foundry.
I think the big fear here is that if Intel does the same, there won't be much competition left in the fab space.
Is Samsung still competitive with TSMC?
Part of it, sure, but they were still fabless and in the ditch before Zen. Unless you're referring to going with TSMC instead of GloFo as going fabless.
What helped them is putting the right people in charge of Zen design and intel fumbling 10nm due to their own hubris.
AMD was fine with having GloFo as a fab until 20nm process. They were already behind, but not terribly.
AMD even used TSMC for their CPUs and GPUs before Zen. Ontario was fabbed at TSMC in 2011.
Point is AMD as free to shop around. Only in 2016 the agreement was amended that GloFo would be preferred for 14nm and 7nm, but since they decided not to work on 7nm, it freed up AMD.
Zen 2 is also where Ryzen went from "exciting and competitive, but not top of the line" to actually giving Intel a run for their money in more than just highly multithreaded workloads.
Improved architecture put AMD within striking distance of Intel and the move to TSMC allowed them to pull ahead.
AMD used TSMC for their CPUs (not IO die!) at least for one generation, and TSMC actually dropped the ball that time and AMD went back to GloFo.
Once again, AMD was fabless since 2009 it's a fact, and it's also a fact that it didn't help them at all.
Nowadays, there will be another process node from TSMC. If AMD doesn’t pay for the R&D, TSMC’s other customers (like Apple and… actually, Intel) will instead.
He later noped out of Intel shortly after joining. Whatever he saw, either in leadership or product, had to be pretty bad in my opinion. AFAIK there's been speculation, but nothing really concrete.
They don't want to be competitive they want to bleed the company dry.
Intel outsourcing their core product line is also a massive liability. It's just a different kind of liability.
I personally think the world's reliance on TSMC indicates that fabs are critically important infrastructure. And operating a world class one provides a company with a ton of leverage with governments and other businesses.
Intel's fab would be doing much better if it spun it out a while ago and was making Intel, Nvidia, and Apple chips right now.
What Intel process from the last decade would have been enticing to Nvidia or Apple?
Intel is doing the same. IDK if they are working on new fabs at this point, but the last few generations of chips from intel have used TSMC.
My expectation is that Intel might still run fabs, but they'll be mostly contracting them out to people who want cheap ASICs and 10 year old fab tech.
Yes, they are.
https://www.tomshardware.com/tech-industry/semiconductors/in...
Definitely struggling, but still in the game.
Samsung comes in a close second in terms of tech. GloFo is also still floating around though lagging pretty bad AFAIK. Micron has it's own fabs that they are actively developing (in fact, they are building new facilities right now).
What TSMC is is cutting edge. That's why everyone that needs top performance uses them.
In this case...TSMC is holding all the cards, not Nvidia
Intel fabs have never had to be as cost effective as others. They were selling top end chips for top dollar for decades. I bet there are 10 other companies that can make 45nm chips cheaper than Intel can on their old equipment. I could be wrong.
If everyone chases higher margin and ditches their fabs what kind of industry are we left with? One giant fab company like TSMC? That sounds healthy!
How many of them develop high performance x64-64 cores?
But if Intel joins the fabless club, all of the sudden the playing field gets much more level.
AMD would disagree?
Maybe if you ignore they're the only player with remotely competitive discrete GPU IP for graphics and AI, after the Nvidia and AMD duopoly.
The problem with Intel is that they are so short sighted and they change direction and focus very quickly. Intel will adopt these seemingly great ideas that require 10-20 year strategies, invest heavily in them, and then abandon them 5 years later. They always measure initiatives against their core CPU line and if they don't show similar profitability in the short term then they defund and eventually cut the programs entirely.
That's the precious leader. The new CEO is not from a semiconductor manufacturing background. His main claim to success is leading a company that built EDA tools.
So we're just going to hand control of the US supply of semiconductors completely over to TSMC, Samsung, and the Chinese fabs in the works? That seems incredibly short sighted and reckless.
global* foundry
https://news.ycombinator.com/item?id=14588429
The top of the market will go GPU and the bottom will go ARM, and the middle will be an ever shrinking x86 market share. The few places that will need heavy CPU resources will be the same people who can apply pressure to Intel's margins.
The process of chip making will look very similar in the future, but the brand of the CPU will matter less every year. Intel's not "dead in five years", but Intel will definitely cross the point of no return in that timeframe. Shifting a big company's focus is more difficult than growing another company who already has the right focus.
I think they have a lot of potential in the dedicated GPU space, but that is a consumer market so profit margins are smaller and they have potential in the low-to-mid-end market so even less margin. It's really sad as the competition there would help consumers.
the sad thing is, it was predictable. Wintel and other monopoly-like deals/situations had removed the need to compete/stay on edge from Intel. They then noticed it too late and made mistakes when trying to course correct/having to much innovation dept to effectively course correct screwed them up big
At the same time AMD again and again re-invented and optimized their development flow and experimented with alternative approaches and did not shy away from cooperating with TSMC and implicitly through that Nivdea and other (sometimes also Intel). Intel on the other hand AFIK got stuck on a approach where they had a edge over AMD but which was seem to have turned out to be somewhat of a dead end.
what is interesting is how TSMC has so far avoided the same kind of trap
- by having competing customers and having deep research co-operations with all the customers they brought competition and innovation back into a monopoly in a round about way like position
- having limited capacity of the newest tech which their competing customers bit for bring in monetary insensitive to innovate
- and them being somewhat of a life line for their country put a lot of pressure onto them to not break their own innovation machine for greed (e.g. by intentionally not expanding the availability of the latest node even when they technically could)
I think dedicated GPUs will be dead soon. AMD will beat nVidia with APUs that compete with midrange DGPU in performance with lower system cost. With AI using GPUs we want the shared memory of the APU rather than splitting RAM into two mutually exclusive areas - witness boards starting to use soldered ram in 64 and 128GB configurations. nVidia can't compete without x86 cores and Intel just cant compete for now.
I mean for gaming there is already the Ryzan Max+395 which already is beyond the level of low end graphics (at least if placed in a desktop where it's not heat/power throttled). But it's a bit of a unicorn (especially if you look for a system where it can run full throttle).
but I'm not sure about the beat nVidea part, nVidea has some experience with putting ARM CPUs on their graphic cards and as far as I remember on for their server center solutions there is one which pairs up graphic cards (and their RAM) over PCIe and mostly cuts out the CPU
"Over the past 10 years, Intel engaged in financial engineering, primarily through significant stock buybacks ($53 billion in 2011–2015) and stock-based executive compensation, which diverted resources from innovation and contributed to its lag in semiconductor fabrication. This financialization, as critiqued in the 2021 report, is a long-term factor in Intel’s weakened competitive position"
https://semianalysis.com/2024/12/09/intel-on-the-brink-of-de...
https://www.ineteconomics.org/perspectives/blog/how-intel-fi...
https://www.intc.com/news-events/press-releases/detail/1726/...
IMHO the whole user-visible p-core/e-core thing on desktop CPUs is one of the worst decisions in the history of microprocessors. My gaming machines need to do double-duty as as build boxes, so they're just utterly unusable for me.
They're also very unpopular online so it's tough to find solid unbiased info about them. Like is the stink about 18A true or do people just want to hate on Intel?
But maybe it was more of an early foreshadowing. I had a housemate that worked on their internal CAD tools and it also sounded like a bit of a mess with NIH syndrome. (20+ years ago)
I live in the area and know a LOT of intel fab workers.
The issue is not the workers: Intel has been captured by corporate raiders and toxic management.
They aren't interested in making chips or an innovative company. They just want to squeeze the juice out of the company until it is dry.
That is why it is so bad.
Instead of investing in the future and paying top dollar for top employees, the Board paid the shareholders (even 20 years ago). They never even tried to compete for the best employees, and instead let them all go to Alphabet/Apple/Amazon/Meta/Microsoft/Nvidia/Netflix.
This includes the employees in management.
Could you explain to those of us who don't understand how corporate raiders have influenced Intel's strategy?
Listed companies are too vulnerable to short-term thinking, thanks to the meaningless “fiduciary duty”.
Smart people know to choose AMD. OEMs heavily favor Intel for the brand recognition. It's the same on the workstation side, though AMD's market share has been rising quite fast (it's apparently at a 36.5% share) so I'm unsure if system integrators will keep pushing their Intel SKUs so heavily.
So they're not cooked, but they're certainly not doing well and barring a massive jump in performance or efficiency, they're not going to be making a recovery any time soon.
I was disappointed with their offerings and went with AMD for my latested build. I don't know too many people who have built PCs recently, but the few I do know who have or are planning to, everyone is planning to use AMD. Similar to the GE example, it seems many people would recommend LG or Samsung appliances over GE.
A friend used to send me articles regularly from Semiaccurate in the mid 2010s. I thought it was "alternative truth" but it turns out to have been more, uh, accurate than I thought.
Quite simply imagine being dropped in as CEO of Intel in 2015. Could you have prevented the malaise of today?
https://www.nytimes.com/1997/05/14/business/suit-by-digital-...
There were much more discussion of Intel stealing things in the 90s than today.
I feel like x86 itself is kinda legacy tech. So while AMD has made advancements, they're somewhat in the same boat as Intel.
It seems like NVIDIA and Micron are the real "crown jewels" of US tech
Intel could make exciting RISC-V relatively quickly if they wanted to; what stops them and other companies like this is the strategic asset they perceive their existing ecosystem as.
https://www.computerenhance.com/p/an-interview-with-zen-chie...
It's like saying that programming language syntax/keywords are better than the other.
Everything is about compiler, lib, runtime, etc.
https://chipsandcheese.com/p/arm-or-x86-isa-doesnt-matter
Also some people say that RISC-V is the way to go
When people say "ISA doesn't matter", they mean that the "legacy cruft" in x86 doesn't matter (that much) and that x86 remains competitive with other similar ISAs. It doesn't mean that the difference between VLIW and traditional ISAs doesn't matter. ISA paradigm still matters, just not the "syntax".
But let's be clear: Of course ISA matters. It's just as trivial to make a bad ISA as it is a bad syntax. But does the ISA of modern superscalar processors matter? Probably a bit, but certainly not a whole lot.
In this particular case: ia64 leaned hard into wide VLIW in an era where growing transistor budgets made it possible to decode and issue traditional instructions in parallel[1]. The Itaniums really were fine CPUs, they just weren't particularly advantageous relative to the P6 cores against which they were competing, so no one bought them.
[1] In some sense, VLIW won as a matter of pipeline architecture, it only lost as a design point in ISA specs. Your Macbook is issuing 10 arm64 instructions every cycle, and it doesn't need to futz with the instruction format to do it.
Isn't having fixed-size naturally-aligned instructions (like on 64-bit ARM) enough to get that advantage?
Really VLIW is a fine idea. It's just not that great an idea, and in practice it wasn't enough to save ia64. But it's not what killed it, either.
And by the way that's why open source makes such migrations much cheaper.
ld r1, [r2 + 10]
st [r3 + 4], r4
Either have a stupid ISA and do all the work ahead-of-time with way more compute time to optimize or don't optimize and have a higher level ISA, that also hs concepts like pointer provenance.
The current state seams like a local minima with both having ahead-of-time optimization, but the ISA does it's thing anyways and also the compiler throwing much of the information away with OoO analysis being time-critical.
Meanwhile the CPU probably can't do a loop invariant hoist in a reasonable way or understand high level semantics.
The impact of ISA is overrated, it's much more important that the ISA continues to grow and adapt as CPUs get larger.
No, it's not. In modern high-speed CPUs, many instructions are decoded directly, without going through the microcode engine. In fact, on several modern Intel CPUs, only one of the instruction decoders can run microcode ("complex") instructions, while all the other decoders can only run non-microcode ("simple") instructions.
It would be more precise to say that it's at the "front-end" part of the core (where the decoders are) that the ISA lives, but even that's not quite true; many ISAs have peculiarities which affect beyond that, like flags on x86.
nVidia market cap is 4T or about 40x Intels. Im not sure who those smart people are.
Intel fired the one CEO that spoke both engineer and business, and Gelsinger could have been their Lisa Su. They fired the only talented CEO they've had for years.
This will be fatal.
Gelsinger was the scapegoat for 20+ years of inability to compete with foreign companies, no matter how much money was poured into them. They used American exceptionalism as a cover to defraud shareholders and any government that invested in them. They used the relationship of AIPAC and Congress to build a fab and R&D lab in Israel (inserting yourself into global politics to make a buck is always spicy) at low cost to them.
Taiwan became the capitol of electrical engineering in the world, and is a shining example of how to survive and thrive in a post-war era, and it absolutely shows. They caught up to Intel and zoomed right past.
Gelsinger's crime was try to do what AMD did: they didn't have a fab that could make their chip BUT they had a fab that made chips that people wanted AND the foundry could take that work and survive if they legally split. GloFo is now the third largest semi foundry in the world today, and when it was part of AMD, it very much wasn't; I can't quite remember, but 5th or 6th? Something like that. GloFo is #3, TSMC is #1, Samsung is #2, and Intel could very well be that #4, and push out UMC (#4) and SMIC (#5) in the secondary chip foundry market.
Gelsinger could have split Intel into Intel and IFoundry or something, and Intel could have profited on IFoundry taking off and taking external work. Right now, IFoundry can't compete on top nodes, but _could_ steal work from all other fabs for secondary larger nodes. Having a working 12 nm competitor as well as a working 7nm competitor is big business, which Intel currently has _ZERO_ of (since they don't take external contracts). Gelsinger was big on this potential revenue stream.
Gelsinger's other crime was being part of the negotiation between TSMC and the Biden administration for the CHIPs act money: part of what built the TSMC fab right next door to Intel's in Arizona was Biden and Intel money. Intel was investing in it's future by playing the American exceptionalism card again, but now in everybody's favor. We _all_ benefit from this. Gelsinger wanted to have _somebody_ fab the chips, and if its good enough for AMD, Apple, and Nvidia, its good enough for Intel.
There is zero indication that GAA 20A is ready, and Intel has a history of having leadership that says such-and-such is ready for it to either come out several gens later, or just vanish off the roadmap. Gelsinger's other OTHER crime is admitting to this and changing the direction of the Titanic before it hits the iceberg, for the CEO that replaced him just to steer right back into the iceberg.
I have _zero_ faith in Intel's leadership if they can't bring Gelsinger back. Tan, Gelsinger's replacement, is a former board member. I have no reason to think he is not just going to further poison the company. Tan has not spoken about any plan that indicates he understands Intel is not competitive, Intel cannot competitively make 100% of the tiles, that Intel's Foveros tech stack is extremely valuable because the only truly comparative alternative is TSMC's CoWoS tech family and superior to it and people are willing to throw money at that problem but they can't license it as long as IFoundry is part of Intel.
Intel is cooked imnsho.
At the time, EPROM tech was Intels most profitble product until the 8088 and 8087, which were designed in Israel at the dev center (along with many of their chip designs).
Edit: Look, to whoever is out there on a downvote spree, I don't care if I get downvoted, man, but wild you'd just downvote people talking about a guy whose won multiple IEEE awards, has patents to his name, and has left his mark on EE, and isn't even the focus of the discussion at hand.
His over-optimism gave the whole "5 nodes in 4 years" supposed path to leadership a weird flavor, like it must be somehow a bit of a con even if it gets technically achieved.
Also, I looked into the claim when he had said it, apparently he was being intentionally misleading about it, and the press tried to ask what he meant: he was speaking tensor performance on future enterprise Arc card products at datacenter scale, ie, AI bait.
In early 2021, Nvidia's compute flagship was the A100, 19.5 FP32 MAD TFLOPs, but the misleading number they quote in marketing is the tensor performance of 312 TFLOPs of FP16 accumulates. That would be about 3.2 million of these at tensor perf.
Skipping H series, in late Nov of last year, their new flagship is the B200. 124 FP32 MAD TFLOPs, 2250 tensor FP16 accumulate TFLOPs. That is now 445k cards to reach zettascale if using tensor cores. You won't be fitting ~1400 GPU-laden machines in a single datacenter, but the number is becoming more manageable.
They improved, in 3.5 years, 7.2x.
Lets say Nvidia does this again. 3.5 years, again, would put you in early 2028, and they manage another 7.2x win: that could be 62k cards across 7.7k. That absolutely is doable in a single datacenter.
The problem is, and this is where the prediction actually falls apart, not that its impossible: We don't know what future Arc cards look like, nor enterprise ones. Battlemage is an improvement over Alchemist, so the tech *is moving forwards at, but either Celestial or Druid was supposed to introduce the enterprise compute card variants, but that seems to be dead, and no indication either of those lines will even see the light of day now. The new CEO seems to be hard set on making Xe for iGPU only.
I can't find any hard numbers on Intel's tensor units, but apparently they're actually competitive. I can find the normal FP32 MAD numbers, and it ends up that Intel is 13.5w per TFLOP and Nvidia is 8 and both companies have equal efficiency in transistor usage. Assuming Intel made a B200 competitor, and assuming the higher power usage is due to voltage (Intel B series voltage is similar to Series 40's voltages, which is a lot higher than equivalent enterprise/pro series cards), Intel could be making a card that's somewhere in the ballpark as 2/3 as good for the same power usage.
So, in the end, yes, I don't agree with his claims of future Zettascale at Intel by 2027. I don't think he was wrong for the industry as a whole, however. If he would have said, say, 2030, I don't think we would be discussing this, that certainly would have been doable if he was at the helm and they kept doubling down on Arc every gen and everything went according to plan.
Regardless, it seems like the company leadership should be gutted (the same could be said of Boeing) and the company given over to a new technically-grounded leadership team.
Betting some on Intel is very wise when the alternatives are, as I see it: 1) investing in TSMC building fabs and creating more of an employee knowledge base and skill base on shore, 2) hoping a US-based startup gets enough traction to grow.
Agreed on leadership. But selecting leadership teams, especially technically-grounded leadership teams is extremely difficult. Which is why companies revert to non-technical leadership so often.
Since Intel has been mismanaged for so long I don’t know how many good lower level employees they managed to retain, I doubt much would be left if they properly cleaned house.
TSMC is making fabs in the US, but they are not SOTA fabs. Those are kept in Taiwan.
Building a fab is no mean feat and loss of infra is a major blow, but it's certainly not impossible to build these fabs in the West, just not economical. You are not starting from scratch.
That being said the government will likely not allow them to fail completely out of the foundry business for geopolitical reasons
There wasn't any bailout on them, what do you mean?
google://Intel chips act billions
Intel was looking bad but not the dire state they’re in now.
>The CHIPS Act primarily benefits semiconductor manufacturers and related industries by providing substantial funding for domestic chip production and research. Companies like Intel, TSMC, Samsung, and Micron have received significant grants and loans to expand or establish new manufacturing facilities in the United States.
>The act authorizes roughly $280 billion in new funding to boost domestic research and manufacturing of semiconductors in the United States, for which it appropriates $52.7 billion
>The act includes $39 billion in subsidies for chip manufacturing on U.S. soil along with 25% investment tax credits for costs of manufacturing equipment, and $13 billion for semiconductor research and workforce training, with the dual aim of strengthening American supply chain resilience and countering China
That is the political calculation, not "throw good money after bad" kind of economics 101.
But I guess "too much socialism"
The problem is that they are far too incompetent and have zero clue about tech, and only understand real estate, that simplest of business that can be executed with mere lizard-brain intelligence.
Tech is also about small startups disrupting large giants, which is completely antithetical to current Republican leadership ideals, where the wealthiest get all gains, regardless of who does the work.
It will take many years of full-on Democratic leadership to reconfigure the Republican Party back to a somewhat innovation-friendly business party. Meanwhile the Democrats, under Biden, were by far some of the most business-friendly politicians we have seen in perhaps a century, spurring massive investment in factories and industry, mostly across red states. But because it's a politically incorrect fact, it never gets reported.
This neat little dichotomy between "free market capitalism" and "centrally planned socialism" is a cute story but also complete fiction. In "capitalist" countries the government basically always runs R&D during any period of time when the stakes are high, and in "communist" countries there are always markets, and they are always sanctioned to some degree.
All of the foundational progress for American leadership in high technology was centrally planned and administered, all of it one way or another: through ATT, through NASA, through the DoD, through the universities. Value creation occurs under the watchful eye of the DoD.
Once in a while we go on an orgy of extractive wealth transfer like now, instead of creative innovation like usually, and the top industry guys always fuck it up. And on cue, yeah this is going great.