In today’s Intel Accelerated event, the company is driving a stake into the ground regarding where it wants to be by 2025. CEO Pat Gelsinger earlier this year stated that Intel would be returning to product leadership in 2025, but hasn’t yet explained how this is coming about – that is until today, where Intel has disclosed its roadmap for its next five generations of process node technology leading to 2025. Intel believes it can follow an aggressive strategy to match and pass its foundry rivals, while at the same time developing new packaging offerings and starting a foundry business for external customers. On top of all this, Intel has renamed its process nodes.
Counting Intel out because they’re facing some really tough years is not very smart. I obviously have no idea when they’ll be on top again, but this industry has proven to have its ups and downs for the two major players, and I have little doubt the roles will become reversed again over time.
I agree with you Thom, that it would be a mistake to count Intel out because the road ahead is bumpy. But, I’d be cautious not to over-estimate what to expect as well. I’ll point out the obvious, that Intel is renaming their product line in a way to intentionally mislead & confuse consumers. That means those products fall short on technical merits. It’s not like that behavior is exclusive to Intel, I just didn’t read anything to be excited about. We’ll see where things stand when people have hardware-in-hand since that’s what actually matters.
friedchicken,
Absolutely. I don’t like taking claims at face value because people have a tendency to get over excited and/or exaggerate. I always wait till hardware is available to the public or at least a couple 3rd party reviewers before I take claims too seriously. Many companies sugar coat things, and their fans are even worse. That’s just in their nature. So I’ll wait for real world benchmarks before agreeing with the claims.
Remember the Athlon XP rating ? An AMD 1700XP at 1.4GHz was considered “equivalent” to the 1.7GHz Intel offering.
But that had actual merit as the GHz number didn’t take into account IPC and the AMD was doing much more work per cycle. Compare a 3Ghz P4 to a 2.2Ghz Athlon from the same period and even though the P4 has a faster clock thanks to its large pipeline slowing it down it does less work per clock.
ironically it seems Intel has the same problem today they did then which was and is heat, looking at the latest performance per watt charts while Intel and AMD trade blows on performance the temp difference is just nuts. Unless they can fix the temp problem (and the big.little design is frankly a bad idea, if your design is such a power pig you gotta stuff an entire Intel Atom CPU in your Core chip just to try to lower the temps? You need a new design that can scale down as well as up and up) they can measure their chips in gnat farts, isn’t gonna matter as they are already getting close to FX-9590 heat levels.
It’d be more accurate to say that other companies (e.g. TSMC) have been intentionally misleading people for about 10 years (attempting to pretend that their process is “better because they made up a smaller number” than the technically equivalent Intel process); and now Intel is merely adopting similar naming to other companies to make things less confusing.
I see, so Intel purposely muddying the waters of their own products actually makes things less confusing for consumers. That’s very thoughtful of them, we should thank Intel for making things as clear as muddy water.
No.
Other companies muddied the waters, causing non-technical consumers like you to see 2 numbers (e.g. “7 nm” from TSMC and “10 nm” from Intel) and incorrectly assume that “7 nm” is better when it’s not; so Intel finally did what technical people have been expecting for the last 5+ years and changed their own numbering to unmuddy the waters.
Note that even a non-technical consumer would’ve known this is what’s happening if they bothered to read the article.
Finally; I guess someone should inform you that a process name (e.g. “10 nm”) is not a product name (e.g. “Xeon Gold 6354”), and that Intel are not changing product names.
I understand the old metric of die size became irrelevant a few years ago and things became more about how they were put together. TSMC had a grasp of the processes required to push things forward whereas Intel did not. This is an open secret in the industry. I also understand this was pushed because “consumers” wouldn’t understand this (which is fair point and one Intel played along with) which is why they continued with advertising die size. It’s also not much of a secret that more can be squeezed out things whether by more shrinkage or more clever techniques at the gate building level. People were saying the limit had been reached (much like diffraction limits with caera lenses set a limit) but we all know now this is not true. Materials science ad engineering is getting very clever at the nano scale.
I wouldn’t go easy on Intel. Intel were the ones who pushed the MHz wars when we all know that MHz (much like camera manufacturers resolution wars) wasn’t the important factor.
It’s pretty much standard practice for industry bodies and companies to post long term roadmaps so clients and others in the ecosystem have a clue where to direct investment. This is rarely published raw by media as they like to appear to be on the “inside” and have “exclusive” information and be on “favoured terms” with particular players.
I’m disappointed with Anandtech. More and more they seem to exist to publish puffery by the usual suspects. Intel has swallowed whole the book by Jack Trout. Anyone who doesn’t think Anandtech’s article isn’t an infomercial and Intel aren’t packing their corporate statements out with marketing is day dreaming. Intel just want to get attention, attract talented employees, and divert investment attention in Intel’s direction. In a sense Intel are trying to create the future they want everyone to believe and if that fails put a spanner in the works of the competition much like Boeing pulled with Airbus.
Between Intel suddently giving a hoot about energy consumption and Microsoft concerned about security I smell a very big rat. They just want to sell stuff and are twisting and turning to hide A.) Intel is not the market leader and B.) If you don’t buy their latest greatest thing they’re throwing your stuff on the scrapheap. Anandtech is notably not covering any of this.
@Brendan
Have I wrongly assumed your replies are meant to be read as non-fiction or do you actually expect people to take the foolish assumptions and remarks seriously? I hope for your sake it’s the former and the Intel cheerleader costume you’ve got on is for entertainment purposes only.
Does Intel make mistakes? Definitely – Netburst was a disaster, failing to bring memory controller on die sooner was stupid, the change to “10 nm” was possibly the largest blunder in the company’s history.
Is the Roadmap primarily marketing? Definitely – they always have been (and not just Intel’s).
Do “for profit” companies (Intel, Microsoft, AMD, ARM, Apple, Nvidia, … ) just want to make more profit? Of course. It’s been the same since capitalism was invented. Anandtech reporting this would be like a physics journal reporting that water is wet.
Does any of the above have anything to do with Intel changing (“correcting”) their process naming? No.
Is Intel changing (“correcting”) their process naming fair? Yes. Feel free to compare “transistors per area” yourself – you’ll find the same as everyone else (that the density and feature sizes of Intel’s processes are equivalent to the density and feature sizes of other manufacturer’s “next smaller by made up number” process).
Will Intel changing (“correcting”) their process naming confuse people? No. It’ll result in less confusion.
Intel’s obvious marketing ploy is not some attempt to do consumers a favor. It’s weird you think naming something intentionally misleading, regardless of who else does it, gives people clarity. It’s like someone claiming the moon is made of cheese and you claiming `the moon is made of gruyere` makes it crystal clear. Then again, you also think water is wet. I think the lack of even a single person agreeing with you on this says it all.
It’s weird that you’re still attempting to call it “intentionally misleading” (and still failing to understand that the confusion caused by other companies puts Intel at a marketing disadvantage, and that by changing process names they avoid confusion and therefore avoid the marketing disadvantage). The first time I explained this I was willing to believe you were merely ill-informed. Now it’s obvious that you’re intentionally ignorant.
The lack of anyone agreeing with either of us is an indication that, out of the relatively small number of people who comment on OSAlert, nobody cares enough to get involved.
Anandtech has a larger audience. You can find essentially the same discussion there – stupid people ignoring facts and implying that Intel is being “misleading” without providing any valid reasons (like you), and other commenters trying to educate the fools using sane arguments backed up by actual statistics (like me).
The reality is that if Intel actually was trying to mislead they would’ve tried to pass their “10 nm” off as “5 nm” (to make it sound better than TMSC’s “5 nm” when it’s not); but Intel can’t mislead like this because the majority of their profits come from customers who are smart enough to tell the difference between “fair” and “misleading”.
@Brendan
Yawn… Let me know when you’re done digging your hole.
@Brendan
I’ve indicated the industry forums have all the roadmaps. I’ve also indicated sharp busineess practices and the use of marketing by quoting one notable author on positioning. You can read it all yourself if you hit the search engines.
There is no big secret here. Everyone i the industry knows die sizes are not die sizes anymore because it is more complicated than that and I explained why. Intel knows this too. I would also be deeply surprised if Anandtech didn’t know. The whole Intel announcement and Anandtech article utterly reek.
This:
And this:
Neither of these statements is saying much if anything at all. Intel are just shuffling words around to talk themselves into being number one by their definition. It’s complicated more by architecture in any case. Basically Intel swallowed Jack Torut’s “Positioning” and want to position themselves as “number one” and that’s what this is all about. It’s marketting, nothing else.
Again, nothing new is being said here and TSMC won’t be standing still nor will ayone else.
Yes.
Well, yes because die size and propriatory technology vary company to company and you can’t give an exact same definition. Intel is just trying to wade in with their own definition which, conveniently, makes them sound like number one, and have everyone nod along.
Uh, huh. So the journalist writing this has a vested interest in the topic. No wonder he is pushing this so hard.
Ok, sure. That journalist, plus all his peers, plus anyone that ends up having to explain it on social media, plus Intel, plus any consumers that try to compare products; have a vested interest in not “having to extensively explain differences in a simple name” (avoiding unnecessary confusion).
Can we say that the government should take legal action against Intel (for “intentionally misleading” marketing) because a true statement was made (by a journalist with a vested interest in avoiding confusion)?
@Brendan
You’re deliberately chosing to ignore the bigger picture including regulation and market abuses and various nefarious practices. Fundamentally this is about corporate arrogance and unilateralism and greed and lies. You chose your side and that’s all I’m going to say.
I think you’re right. It’s not about people being wrong because they’re misinformed; it’s about people being deliberately wrong because they want to punish Intel for unrelated nefarious practices that have absolutely nothing to do with the process name change whatsoever.
@Brend
No. It’s marketing and I explained how and provided the relevant quotes but you’re chosing to ignore this.
Intel are not just shuffling words but also conning people into buying their old foundaries and using the money to build new foundaries. They also want to build foundaries in Europe and suck up European talent so Europe gets cut off at the ankles before the Horizon initiative goes anywhere. That is what it is all about.
Another weasel words argument from Intel is this is about power consumption. Well, duh. That’s been the case since forever and not a unique and special insight Intel has.
https://asia.nikkei.com/Business/Tech/Semiconductors/Taiwan-gives-TSMC-green-light-for-most-advanced-chip-plant
Like I said early the competition is not sitting on its hands.
As you note the competition also prioritises national self-interest as well as farming out its lower tier manufacturing to other countries. The environment is an issue too so water and water reuse are now play a bigger role in decision making both for capacity reasons but also being responsive to environmental regulation.
The days of America claiming to be “leader of the free world” and walking in the door with big smiles and promises of making you rich are long long long gone. This isn’t about dick measuring but about self determination and national wellbeing and international relations. It’s about everyone having a fair crack at the whip.
@Holly
Yes, it’s marketing. For marketing, Intel has 4 choices:
1) Keep using the old process naming, and continue being victims of other company’s deceptive marketing that makes Intel products sound worse than they are.
2) Change the process naming to match competitors. This is 100% fair (and backed up by density statistics, and understood/accepted/suggested by everyone familiar with the current/past process naming mismatch). This is what Intel are planning to do.
3) Change the process naming to something completely different (e.g. maybe use a “month-year” format like “the Intel 9-2021 process”; maybe use “major.minor” version numbers like “the Intel 2.1 process”, …). This would also be 100% fair; but makes it harder for people to compare (and doesn’t align well with a “sell fab capacity to third parties” goal).
4) Change the process naming to make it sound better than competitors (e.g. rename “10 nm” to “5 nm” so that it sounds better than TMSC’s “7 nm” when it is not). This would be intentionally misleading. This is not what Intel are doing (despite friedchicken’s deluded claims).
Every thing else is irrelevant. Specifically:
a) What Intel are doing to encourage forced obsolescence (your pet peeve) is irrelevant to this conversation.
b) Intel’s annoying “product differentiation” tactics (one of my pet peeves – forcing people to pay a “Xeon tax” just to get ECC support) is irrelevant to this conversation.
c) Intel’s past blunders (e.g. the woeful switch to 10 nm, the frequency wars, etc) are irrelevant to this conversation.
d) Whether Intel will/won’t achieve other parts of the roadmap (5 processes in 5 years is probably too ambitious) is irrelevant to this part of the conversion.
e) Where and how Intel (and their competitors) build new fabs is irrelevant to this conversion.
f) The ecological impact of semi-conductor manufacturing (while important) is irrelevant to this conversion.
g) Whether Intel’s products are better or worse than other manufacturers products (including “Intel vs. AMD”, “80×86 vs. ARM” and “Intel 80×86 vs. Apple ARM”) is irrelevant to this conversion.
h) Whether any specific journalist, or any specific person, has a “vested interest” is irrelevant to this conversion (whether a statement is correct or not depends on the statement alone, and true statements don’t become false merely because someone you don’t like says it).
i) Whether USA is “the leader of the free world” is irrelevant to this conversion.
Essentially; my original comment (and all comments since) are focused on Intel’s process name change and whether it’s good/fair (or bad/unfair/misleading/confusing); and I have no desire to be distracted by “chaff”.
@Brendan
I make my own mind up about things without running it past you. I also decide for myself what is relevant and not relevant. Your opinion is simply your opinion. And what you call “chaff” are matters of public policy and regulation and law. Those don’t go away because you want to bury your head in the sand.
As I have exposed this Intel scheme is a marketing dodge. They are trying to repeat the success they had with the “Intel Inside” marketing campaign and it is marketing and it is a campaign.
Arstechnica have a few questions to ask of themselves and of the journalist who peddled the article. They have a fact and ethics problem. But like I said: you chose your side. That’s on you for good or ill.
@ Brendan
Nobody is misleading anything. Process node names are basically shorthands to refer to a specific process, the definition of which literally takes gigabytes of data. The customers of the Foundries are the system designers, and they are aware of the details.
The main mistake has been exposing node names to the end consumer. Which is something that was always going to lead to more confusion.
Consumers should only care about the performance metrics to make an informed decision for the use case for the actual device they will be purchasing. Details about the manufacturing process are irrelevant for the end consumre.
The bottom line is companies started using process node names in their marketing and people naturally & predictably started associating them with the other generalized marketing bullet points (lower power, speed/efficiently boost, etc). And thus `smaller is better` was born.
Once it became obvious the marketing tactic worked, in spite of the consumer confusion it caused, companies doubled-down with it. This is the point where you can argue companies were intentionally misleading consumers because they chose to run with it rather than better inform people.
Intel’s fancy new renaming campaign does nothing to improve anything. The pendulum hasn’t been swinging their way and in a matter of self-interest they’re trying to do whatever it takes to change that. Regardless of the silly nonsense @Brendan is pushing, Intel is simply trying to improve their public perception in hopes it will translate into more sales. Better informing consumers is neither their goal nor a side effect. Taking the `if you can’t beat em’, join em’` approach doesn’t magically make everything clear and create an educated public. It’s nothing but a concession you’re losing the marketing fight.
nevermind
5 new process node generations in 5 years? That seems a bit… ambitious, considering their current pace.
I read it as bragging. Parly a roadmap. Partly marketing. How much is deliverable really is another question. They may have the potential if everything goes right. They may come unstuck. We don’t know and are not being given the information to know for certain.
Taiwan pretty much bet their house on TSMC. Given its national strategic importance both economically and from the point of voew of national indepedence they are not going to stand still. Nor is anyone else. The European Union Horizon’s project is intended for the EU to pull its socks up and develop domestic production capability. Taiwan is almost certainly aware of how Europe’s chip manufacturing industry collapsed and then collapsed again after the US wrecked it by gutting the phone industry.
The story of Rolls-Royce (the engineering company not the car company) and Lockheed in interesting. Lockheed suffered from bad timing with the Tri-Star or might still have been in the civil aviation business as the Tri-Star was then the best plane of its type. Rolls-Royce went bust developing the RB 211 engine and had to be nationalised. As part of this exercise the UK goverment leaned on the US goverment to bail out Lockheed who were about to go bust themselves. Lockheed was a fine company worth saving although exited the civil aviation business as Boeing cleaned up. Rolls-Royce went on to dominate the jet engine business.
I think the point I’m making is you don’t **** and **** on your friends. Plurality and partnership has benefits. Monocultures can be very wrong in lots of ways. I don’t think this unilateral move by Intel is winning them friends.