Europe completes first phase of silicon independence project The European Processor Initiative (EPI) has concluded the first phase of its efforts to create made-in-Europe chips, an effort it is hoped will reduce reliance on imports, improve sovereign capabilities, and create the continent's first exascale supercomputer. The EPI’s end-of-year report, published this week, notes a few …
Interesting. Is this going to fracture the CPU market ? Are we going to find ourselves in a world where there's Intel, AMD and EuroChip ?
And what about other countries ramping up their own fabs ? Are we going to find a dozen choices for CPUs in the next decade, or is everyone simply going to make Intel or AMD designs ?
I'm not strongly against the idea, mind you, but choosing between the two incumbents is already a headache, I'd hate to think of having to choose between four or five different platforms.
The proof is in the pudding.
There is already a made-in-Europe chip, the Russian made MCST Elbrus-8C processors. Unfortunately Sber, Russia's biggest bank, has evaluated them and found "Insufficient memory, slow memory, few cores, low frequency. Functional requirements not been met at all."
One hopes the EuroChip can do better but until it exists and can be tested, hope is all we have.
https://www.tomshardware.com/news/russias-biggest-bank-tests-elbrus-cpu-finds-it-unacceptable
Yes, but you have to remember that the EU is “The One True Scotsman”.
Russia isn’t in Europe according to them.
Although, the EU also include the Israeli Tower Jazz foundries in their official statistics of “European fabs”. Tower Jazz participated in the EU Horizon 2020 R&D program and have been quite extensively funded by EU.
There are two aspects to this: one is chip fabrication capability (which has nothing to do with ISAs), the other is design capability.
We're unlikely to see cutting edge fabs in Europe any time soon - the investment required is enormous simply as a form of insurance. However, the current crisis in vehicle manufacturing demonstrates there may well be an economic benefit in being able to produce more commodity components such as industrial microcontrollers.
There's clearly also some interest in increasing design skills and having input to newer initiatives like RISC-V. In the controller area it's more likely there'll be a convergence of instruction sets (on something like RISC-V) as opposed to the eclectic range we currently have. The EPI obviously has ambitions in high-performance computing, but that's going to be a somewhat bigger challenge.
@aki009
Yes, but I wonder why you state the obvious, yes it's a state run project. I prefer to call it "state run" as governments come and go quite often while the projects proceed.
Take ITAR and so many more.
The goal and the only reason those project are started (by intelligent) states is that companies, the private sector, haven't got the means to do it alone, the end goal is that the industry will eventually pic it up and prosper from it in those countries that invested in the project.
(Universities are also often for very good reasons part of such projects).
How difficult is that to understand.
Two CPUs? I dont think so, if you limit yourself to AMD and Intel you still have an array of chips to choose from. But look inside a mobile phone, how many sport AMD or Intel CPUs? Not many. Not many tablets either. Nor those pesky units in your car, TV etc etc
You should be choosing the best from a multitude, it certainly makes sense for europe and the UK to have not just capability to design but capability to produce because today we need to have the ability to defend ourselves because Russia is poised to start another war... sanctions? How much does Russia trade with the USA? Can europe apply effective sanctions when so much gas and oil come from Russia? Does Russia care even if europe managed to do without Russian gas when the Chinese will just buy any spare because they dont give a damn about the global warming that everyone else is panicing about. What happens when Russia invades and installs a puppet in Ukraine, nothing, it the sanctions are too great then China who is still friends with Russia turns round and turns off the supply of cheap tat to USA and the USA caves in again (as it usually does). Russia is then emboldened and wont stop until it has absorbed all the old eastern block and maybe even Germany
"We probably aren't in there, processors don't work well when you put borders around your parts of the chip design and waste valuable silicon space decorating them with flags and royal insignia."
Comments like this are as prejudiced as the Daily Mail nonsense that prompts them.
As for this initiative, it's not clear how deliberate it is that the ability to design is being confused with the ability to manufacture. We (the West - Europe, UK and the US) are short on skills in both, with demand outstripping supply of people with relevant experience. It's relatively (in terms of industry scale) easy to increase the design skill base, but incredibly expensive and difficult to improve the manufacturing capability.
This isn't just the issue of the astronomical cost of a modern fab, but also the very limited number of people with the skillset to utilise it. Remember that Intel struggled with moving from 14nm for years. Unfortunately for many academic centres, the theory of design and manufacture is taught, whilst the practicalities are left to technicians or simply treated as a 'virtual' problem. In short, how are we teaching the next generation how to run a fab when we don't have a current generation with this knowledge?
The expectation has to be that after great strides in design, the second phase will see relatively small numbers of 'proof of concept' parts produced, with as much reliance on foreign fabs as we currently have.
Supply of people with relevant skills....
This is a problem highlighted in a very good book 'how the west was lost'. (Dambisa Moyo)
If you have a kid today what is your advice to them...
a) Study and work your fingers to the bone on science and engineering in order to be paid a pittance
or
b) Go into banking and earn a shed load
or
c) Get your tits out (sorry girls) or kick a football around / take up ham acting and become famous, rich beyond avarice and have whole pages and pages devoted to how wonderful you were when you finally pop your clogs after a couple of forgettable football game or films.
I wonder why we have no engineers or scientists any more? Sadly for me no one wants to see my dazzling football skills now I am fair fat and near sixty but if only my dad had given me the correct career advice I could have been retired in the sun on one of my choice of private ships with a queue of teenage hopefuls willing to trade anything for my autograph
Not sure why you got one downvote.
The Johnson is an idiot, and Brexiteers will probably scoff and deride anything European, despite it being advantageous to the UK. Just look at the Daily Express comments section.
If the UK were involved, then any fabrication here in the UK could be expanded upon using the gained expertise. That, and the given security of design, and the security of supply chain, is a bonus.
The bluff, bluster and claimed patriotic slogans are just superficial con tactics for the electorate. The Tories only see business opportunities that pay back in the first year or so. They are idiots, and the claimed techno revolution for the UK will fail. The Truss trade deals are a farce.
I despair at the ignorance of the UK politicians and their continued corruption in back room deals with companies to sell off the UK assets and select them as non tendered favourites.
It's a lot deeper than that, no country should be run by a goverment of just one party, those systems are for countries like North Korea and China. Sadly also for most English speaking countries too, and it certainly shows.
In coalition governments there is better internal control, less sleaze, and more internal competition.
What is opposition in a two party system like the British more than a guy shouting across the dispatch box waiting for the next election.
Indeed sometimes it takes time for political parties to agree on what to do and how to do it.
The advantage is of course that during the much needed time for them to actually agree the country is run by those who actually know what to do and howto keep the country running, the so called unelected bureaucrats.
Also when a coalition government goes shit and cannot cooperate they simply resign and let the parties try again.
Have you ever heard of a one party government asking the opposition to take over.
The two party system is a disaster, just look at the Americans where half of the population feel that the other half are their worst enemies. The British are a bit better due to a better education I think.
Nothing is perfect but lets try to have a happy Christmas all the same.
(and the James Webb is soon on it's way, happily too, I hope)
Two party govt means that you can replace them on a regular cycle, without a wave of executions and famine. The "opposition" merely has to exist for this to happen, not much more. That is really the only thing our system has going for it, but it's a big plus.
They don't have a lot of good ideas to begin with, and after a couple of terms they have certainly used up whatever they had.
Coalitions - if you could make it work stably for long - would be a one party govt, weighed down by the long tail of backroom deals, and the inability to have a proper cleanout.
And yet all around the world we find stable, economically functional nations run via 'coalition' generating electoral systems. Wow, they are clearly defying 'laws of nature'. Obviously reality must be changed, the 'law' cannot be wrong.
Seriously though, the 'two party' adversarial system is basically an historic throwback to a different age---principally the 17th-18th Centuries, where 'Britain' was still getting over the Wars of the Roses, which were themselves the final gasps of the tribal/feudal system. It's not an unworkable system, but in a modern 'democratic' society it has massive weaknesses, which are on display daily in Britain at the moment.
The UK is the home of ARM, but let SoftBank, a Japanese company specializing in Ponzi schemes like Uber or WeWork, acquire its tech crown jewel.
As for Intel’s plans to open a cutting-edge fab in Europe, their CEO clearly stated the UK is not in the running because of Brexit.
@Fazal Majid
"The UK is the home of ARM, but let SoftBank, a Japanese company specializing in Ponzi schemes like Uber or WeWork, acquire its tech crown jewel."
A private company sold privately and currently being blocked from privately selling privately to another private business. Private used to mean something.
"As for Intel’s plans to open a cutting-edge fab in Europe, their CEO clearly stated the UK is not in the running because of Brexit."
Thankfully. Look at the subsidies needed to make these fabs in expensive countries. We have high costs in labour, bureaucracy, health and safety, electricity, etc to produce something we cannot compete on price with. Better let the US and EU pay it out and UK benefit like the rest of the world from cheaper chips.
"Cheaper Chips" seems to be a phrase that highlight a very profond lack of knowledge on the semiconductor sector.
And it's all sunshine & rainbows until a country (let's say Korea) puts its industries (let's say Automotive) to the front of its silicon supply queue and everyone else has to start scrabbling around in a panic because they have no domestic production themselves.
@AC
""Cheaper Chips" seems to be a phrase that highlight a very profond lack of knowledge on the semiconductor sector."
The rest of your comment doesnt seem to disagree with cheaper chips as the reality though. You call it a profound lack of knowledge but then accept its true.
"supply queue and everyone else has to start scrabbling around in a panic because they have no domestic production themselves."
That instead is a supply issue. One which seems to be less of an issue if the US and EU are paying to produce there so there is 'competition' with Asia. The reason its still uneconomical to manufacture the chips here is because even when demand is higher than supply the things still just cost too much to produce in the expensive countries.
There are a few, but all older technology AFIK. Also as companies change leadership they get shut down if not profitable enough (e.g. the Motorola in South Queensferry, TI fab in Greenock was being closed but last minute bought over by Didoes, Inc, Didoes bought the old Zetex group (formally Ferranti) but moved some of those part manufacture to far East, etc).
The car companies did it to themselves, by leaving the foundries holding the bag and cancelling the orders, as they always do and screw their suppliers.
Well, this time the foundries found other clients quite keen on using spare space.
When the car companies realised their mistake, the foundries were not happy with their clients, having put them on a critical situation for a second time. Can't blame the foundries for not giving the car companies priority.
Thankfully. Look at the subsidies needed to make these fabs in expensive countries. We have high costs in labour, bureaucracy, health and safety, electricity, etc to produce something we cannot compete on price with. Better let the US and EU pay it out and UK benefit like the rest of the world from cheaper chips.
This is an example of the massive ignorance and short-sightedness, blinkered vision that has led us into the mess we are in. Or it's a desperate barrel scraping attempt to save face when you are looking at being part of the most embarrassing fuck up ever.
It's not an unreasonable comment given the last attempt at building a home-grown fab involved a billion pounds worth of subsidies and resulted in a fab that came on line just as chip prices collapsed. You don't hear about it much because it was never fully built out, never made a penny in it's original form and was almost immediately out dated.
It's ignorance to not be aware of the history, or to assume that the only reason we don't have a fab is because we haven't wished hard enough.
Which factual points do you disagree with? Other than just throwing random insults.
EU labour costs more than Taiwan. That is just a fact. You are entitled to alternative opinions, but not to alternative facts.
Fabs are very power-intensive. The wholesale price of electricity in EU has more than x5 over the past couple months. Much of East German heavy industry is currently in shutdown, and Kosovo has rolling 2-hour blackouts. There are reasons, and it seems pointless to debate why. But again, those are just *facts*.
Better let the US and EU pay it out and UK benefit like the rest of the world from cheaper chips.
If Taiwan supply gets interrupted then the UK would be last in line. Although is suppose depending how far you are willing to bend over for Xi that might not matter.
So, your statement is not falsifiable.
In your view, either:
The EU is first to do a deal with China for Taiwanese silicon because it is so big and muscly and important, and the UK is so way behind we are puny and irrelevant.
Or the UK is first, because we bent over so hard and got screwed in the negotiations.
Either way the facts turn out, proves you were right? Does that….suggest any methodology problem to you?
And on the subject of nothing at all. How did the commercial negotiations for early access to vaccines go? Did it in fact turn out that a governmental negotiator with skin in the game (his fellow citizens *die* if he fails to negotiate a quick deal) outperforms a “skilled EU negotiator” who focused on getting the lowest price? Who *told* the EU negotiator that was his goal? Was that *your* goal? Do *you* think that it was more important to buy at $2 per dose than $2.50 per dose, as a trade off for saving 1% of a life, or a 10% probability of a stay in intensive care that would cost £40k? Because those are the actual numbers. And only a negotiator with no skin in the game, who wasn’t within a thousand miles of bearing the cost of a £40k stay in intensive care, would think that *price* would be what they should be focusing on. Or insisting that while it could be made *in an existing working factory*, instead they should make the purchase conditional on the production being in a factory with no experience of vaccine production at all, purely because it was in their country.
This is the reality of doing things your way. And now, a year later, fortunately we all have vaccines. And in a very real sense, this was never a moral issue, because each vaccine dose goes to save some human life, and I don’t care if it is a Belgian or an Indian. But my god, I don’t understand how you watched somebody negotiate and say explicitly that your life was worth less than a dollar, and not think
“Are these my bros? Do they have my back?”
Back to topic. How do you think an EU negotiation with China for access would or will go? My view: I bet they will play hardball, it’s all they know. And they will get a good deal in the end on price. But in the three years it will take to get there, EU industries will have been blasted into a smoking ruin. But the EU negotiator will think they did a great job.
There are certain things going quite well in the UK (ignoring the ineptitude of COVID and every single political party, besides the monster raving loonies).
The offshore and onshore wind power revolution is stepping up a gear (we will soon have 40GW of wind power by 2030, a staggering quantity, with rumours of 100GW by 2050).
The hydrogen manufacturing plants are building up (ITM power in Sheffield leading the way). We may soon be a major hydrogen exporter (we have 40% of all EU offshore wind resource).
Syncona is creating a significant amount of UK biotechnology firms. Autolus, Achilles, Gyroscope etc. Allowing major breakthroughs in cancer treatments.
It has been a long time since the UK was at the forefront of a technological revolution. But we are a windy island with shallow seas, so offshore wind was always there for the taking.
I suspect most Brexiters would be fine joining an EU CPU initiative. It's not exactly a 'red flag' hot topic that anyone is likely to have heard drunks in Wetherspoons uttering "dey tooook oour silicons!"
Honestly, I thought it was a good idea to sell ARM, when RISC-V started becoming a major threat on the horizon. £24bn is a significant sum.
I suspect most Brexiters would be fine joining an EU CPU initiative. It's not exactly a 'red flag' hot topic that anyone is likely to have heard drunks in Wetherspoons uttering "dey tooook oour silicons!"
All covered by "Bloody EU".
It's a spectrum - there are a few "I didn't like how it was going but I'm happy to trade with them, work with them" but a whole lot more with an "It's evil! We must have nothing to do with them" perspective.
There's no value in reasoning with those who believe Farage ought to be, not just knighted, but made a saint, that the UK's future is as the 51st State with a Trumpian lording over us.
The real problem is we have a government which seems beholden to such voters, is too afraid to upset them lest they lose power.
The first grant agreement is implemented under the European Commission program Horizon 2020 (FPA: 800928) in the December 2018 to November 2021 time span. The second agreement will be implemented afterwards under the EuroHPC Joint Undertaking which issued a call answered to in January 2021 by the same consortium (H2020-JTI-EuroHPC-2020-02 FPA in EPI (phase II)).
https://en.wikipedia.org/wiki/European_Processor_Initiative
This post has been deleted by its author
About the EuroHPC.
"Public members
As of January 2020, public members of the Joint Undertaking include, the European Union (represented by the European Commission), 26 of the 27 EU Member States (Austria, Belgium, Bulgaria, Croatia, Cyprus, Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, the Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain and Sweden), and five non-EU associated states of the EU's Horizon 2020 programme (North Macedonia, Norway, Montenegro, Switzerland, and Turkey).
Other EU Member States or countries associated to Horizon 2020 are able to become members, provided that they accept the Statutes and financially contribute to the achievement of the objectives of the Joint Undertaking.
Observer states
Malta has been given "observer" status for the EuroHPC JU, allowing it to participate in deliberations of the Governing Board, but not receive a vote.[28][29] The United Kingdom lost its observer status following its departure from the EU on 31 January 2020.
Why would UK need a “say”?
This is one of the weirdness of the whole EU project. Is there some special German instruction that they want to have a say in inserting into the CPU? Maltese interest in the optimum data buswidth? A French cache-size?
Then WTF is the reason to have 27 different RISCV cores on the silicon, each one proposed by a separate EU country “having its say” for some “National Project”.
Pure technical insanity.
Fortunately, not our circus, not our monkeys.
"That reminds me of SECAM."
Not sure about what you want to say there.
"SECAM was later adopted by former French and Belgian colonies, Greece, Cyprus, the Soviet Union and Eastern bloc countries (except for Romania), and some Middle Eastern countries. However, with the fall of communism, and following a period when multi-standard TV sets became a commodity, many Eastern European countries decided to switch to the West German-developed PAL system.
Other countries, notably the United Kingdom and Italy, briefly experimented with SECAM before opting for PAL. "
https://en.wikipedia.org/wiki/SECAM
https://en.wikipedia.org/wiki/PAL#/media/File:PAL-NTSC-SECAM.svg
"That reminds me of SECAM."
Not sure about what you want to say there.
I'd guess he's saying that France opted for the French-developed SECAM system primarily because it wasn't American (NTSC) or German (PAL). It wasn't entirely surprising in post-war De Gaulle-era France, any more than European nationalism is particularly surprising in Macron-era France.
"De Gaulle-era France"
Phil O'Sophical, I am sure you will smile at this funny guy with his mad claims, right.
https://www.youtube.com/watch?v=4orQgi1avP4
I get a rather similar feeling when you associate de Gaulle with SECAM
Companies invent and compete regardless of who ever President.
Why don't you read about SECAM and PAL and the history behind it, and the time when invented.
https://en.wikipedia.org/wiki/SECAM
https://en.wikipedia.org/wiki/PAL
I have listened to Brexit Britain now from the very beginning and still you cannot avoid the "European nationalism isn't particularly surprising in Macron-era France".
Just more silliness, what is it with you, still sour about having been kicked out of the continent or is it just that you know they are as good as you are, and that is so annoying.
Companies invent and compete regardless of who ever President.
Why don't you read about SECAM and PAL and the history behind it, and the time when invented.
Might I suggest that you read the link you provided, where you will find the statement "in the post-war De Gaulle era there would have been much political resistance to dropping a French-developed system and adopting a German-developed one instead".
I've studied, built & repaired PAL and SECAM decoders for some 30+ years now. The RCA system that was adopted by the NTSC was an ingenious solution to the problem of retaining compatibility with the previous black & white system. As always, though, being first had disadvantages; once the system was in common use its sensitivity to phase errors became evident. Both Thomson's SECAM and Telefunken's PAL were updates to NTSC which resolved that, and there were technical pros and cons to each. Simple-PAL decoders could handle an error of 20 degrees or so, SECAM could do better but was harder to handle in the studio and required an expensive (at the time) precision delay line. As European countries came out of their post-war period in the 1950s/60s and started to spend on more 'frivolous' things like TV, they all tested all three systems. Faced with the political resistance to adopting a German-derived system (PAL) over a French one, France went for SECAM. Rumour at the time was that the French government subsidised Thomson's delay line production to make SECAM TVs as affordable as the early PAL ones.
I have listened to Brexit Britain now from the very beginning and still you cannot avoid the "European nationalism isn't particularly surprising in Macron-era France".
That has little to do with Brexit, and more to the fact that I've been watching French politics for 20+ years. It's clear that Macron is determined to create an EU as a nation that can outdo the USA, and for that he is keen to drive more integration at the political, fiscal and military level. He is by far the most nationalist of the mainstream EU politicians. If he gets re-elected as President next year then I would not be surprised to see him push to make the role of EU president one that has more executive power, along the lines of the USA, so that he can then move on from France to become EU president post 2027. It's all about power & control.
what is it with you, still sour about having been kicked out of the continent or is it just that you know they are as good as you are, and that is so annoying.
Why do you think one group is "as good" or "better"? We're just different, with different approaches to solving problems. Personally I prefer the freedom of the EEC model of co-operating countries in a free market, with shared aims, over the EU model of one-size-fits-none centralised control. It's a pity it reached the point where we chose to leave, but we did and that should not be any reason why we can't still cooperate where it benefits both groups.
@Phil O'Sophical
I suppose you know that text is written by a Brit.
I would claim the Swedish version is more to the point. Why would the French not use a French system (and why would the Americans not use a American system) when there is business involved.
The Swedish through GTransl.
"Some have argued that the main reason for the development of SECAM in France was to protect the domestic television industry. In any case, incompatibility had already begun with the unusual decision to adopt positive video modulation for French broadcasts. The old British System A was the second system to use positive video modulation. Just to mention, the SECAM development was a precursor to the PAL development. NTSC was evaluated as inappropriate in Europe due to its nuance problems, which required additional adjustment control, which SECAM and PAL solved without an adjustment control. PAL was developed by Telefunken, a German company, and in the post-war de Gaulle era, it had been and has become a major political resistance to abandon a French-developed system to adopt a German-developed system instead. Unlike some other manufacturers, the company that developed SECAM sold Thomson TVs worldwide under different brands. Thomsom bought the company that developed PAL, Telefunken and today also owns the RCA brand - RCA, the developer of NTSC. Thomson was also involved in the development of the ATSC standard used for American high definition television."
The French version has this:
"The political choice of part of the television standards in France rests with François Mitterrand (Secretary of State) as well as Charles de Gaulle. For reasons, on the one hand, strictly to protect French borders against the real "influence" and competition that already existed in radio (Anglo-Saxon and pro-American) during the Cold War, but also for the influence of technological mastery French. The government has also advocated the adoption by other countries of this innovative technology since it is theoretically better than its two competitors (German PAL and especially American NTSC)".
As for the British are all similar and like this and all the French are similar and like that, that is a lot of rubbish with deep roots not only in Britain.
But you are damned good at it, learn about the Scots by looking at Dad's Army, for proper information.
But all the best for the New Year to us all.
"A French cache-size?"
Well, that would gave a certain caché.
NB. Just for clarification, cache and caché are different words, with different meanings and, most importantly, different pronunciations. Cache rhymes with cash! Caché does not.
(I've noticed this problem especially but not exclusively with some of our North American colleagues)
Justthefacts 'Then WTF is the reason to have 27 different RISCV cores on the silicon, each one proposed by a separate EU country “having its say” for some “National Project”.'
It would be a misconception to believe that large ASICs such as multi-core, multi-processor arrays were developed with a single organisation or team. Multiple teams perhaps even using multiple partners and/or subcontractors, based around the globe, may be used in large scale ASIC design projects.
The idea that there would be 27 teams working on 27 different core implementations is also not correct. There may be a couple of different variants to deal with the physicality of the overall system floorplan, but otherwise these would just be multiple instantiated instances.
ASIC project Implementation tasks on large scale, global projects are usually split by hierarchical blocks and sub-systems. Not by duplicating work.
Of course it is true that SoCs assemble IPs from many sources. I was in the ASIC game for twenty years, and managed several complex ones. But look, here’s the top-level details of this actual SoC project. Have a read, see what you make of it. Technically. As a SoC that you were considering buying. Google the companies IP referenced, and have a read of their white papers.
https://www.hpcwire.com/off-the-wire/epi-announces-successful-conclusion-of-european-processor-initiative-phase-one/
I particularly enjoy the separate power controller running in software on a RISCV, using “advanced AI algorithms” co-designed between University of Bologna and ETH Zurich. Not the Hardware People who were involved in the clock tree architecture and layout, or understand the IPs being managed at the gate-level, or with experience of the silicon libraries. Unrelated academics. And actually, neither gate-level nor layout is available yet, let alone EDA power estimation on vectors. Just the RTL. But they’ve already completed the power controller design and algorithms.
"So, basically, UK has no saying on the project, right?".
Well, Britain wasn't kicked out of the EU by the EU.
I Britain wants to take part then sign up to take part of the financing and the rules.
Most people in the EU would support such an decision without doubt, perhaps.
"A processor designed by committee?"
Just a very stupid thing to say.
It's like claiming Airbus is rubbish because it's partly owned by several countries and has production in several countries too.
Resources and costs have to be shared in the EU in order to compete with the rest of the world.
Rubbish. The process parameters have *everything* to do with the design and fundamental architecture. It determines, for example:
How many pipelining stages are optimal.
Relative Power consumption of different operations, which in turn determines how many gates you want to spend on cache vs branch predictors vs execution units vs OoO
Layout delays which in turn determine things like the best Network On Chip architecture.
The ISA is (maybe) the first 10% of the total design work of the CPU.
Honestly, I personally could knock up a “high-level” rival to a RISCV on my own, from absolute blank sheet zero IP, in twelve months. Making it perform in practice like an Arm Cortex A72 on that power budget? Ah. Well you’ll probably have to give me a team of a thousand talented engineers and five years to do that.
The budget to port a *fully working and optimised CPU design* from 28nm to 14nm? $50m. But to make it optimal on 14nm to compete with designs originally optimised for 14nm? $500m ante, plus a large risk budget.
Global and interdependent, sure, but also pretty much a monoculture (ASML and TSMC). Given the relationship between China and Taiwan at the moment, being concerned over this very brittle industry structure is not jingoistic.
We have single points of failure that would have a profound effect on the world - this isn't just a question of whether AMD can produce enough mining rigs, but whether factories, automated production lines, transport and communications systems can be built and maintained.
We are perhaps 6 months away from solving that issue.
It didn't help that when the economy tanked in 2020, most semi fabs decided to cut production capacity, only to find chip sales essentially bouncing back quicker than expected. Followed by a blind panic of factories ordering far too much, when faced with existential risks and 12-24 month delivery leadtimes.
For sure, I know our purchasing team severely crapped the bed when orders went from "confirmed delivery next month" to "oops! 12-18 months". We now have orders for 24 months of critical components and hoovering up any vaguely grey market stock we can find.
Exactly. ASML Making the equipment in Netherlands is an excellent use of capital. They have a competitive advantage moat, the skills, the processes, the surrounding R&D landscape, and an excellent reputation for both quality and as close to on-time execution as possible in a monstrously technically challenging area.
Whereas investing in silicon fabs in the EU is just dumb, and jingoistic nonsense.
A less arrogant EU would shovel investment to ASML. But we know perfectly well why they don’t. Because France insists on pushing money to their fab in Crolles, and spends vast lobbying budget to prevent investment going to Netherlands. That’s the way it works. Helps nobody.
Edit: For those who don’t know, France’s *annual* lobbying budget to the EU directed specifically to *killing* investment in semiconductors in Netherlands is 85MEuro. That’s a lot of liquid lunches. France invest heavily in trying to starve ASML, because they know they can’t build up a French competitor while ASML exists.
"Edit: For those who don’t know, France’s *annual* lobbying budget to the EU directed specifically to *killing* investment in semiconductors in Netherlands is 85MEuro. That’s a lot of liquid lunches. France invest heavily in trying to starve ASML, because they know they can’t build up a French competitor while ASML exists."
I think you need to back that up with some references. Otherwise it's thealternativefacts.
"A less arrogant EU would shovel investment to ASML. But we know perfectly well why they don’t. Because France insists on pushing money to their fab in Crolles, and spends vast lobbying budget to prevent investment going to Netherlands."
I don't get that at all. You mention a semiconductor fab in France which is supposed to be in competition with ASML.
But ASML isn't a fab - they make equipment used in fabs. They may be a supplier to the Crolles plant but they certainly aren't a competitor to it.
Perhaps I misunderstood you - could you explain your point?
"A less arrogant EU would shovel investment to ASML."
Why? ASML is a v successful company which doesn't need public sector investment. The rise in their share price indicates that private investors are very happy to support them if they need more capital.
(Disclosure: I'm a shareholder in the company and the very modest investment I made about a decade ago has done very, very well.)
On the surface relocating manufacturing and design of products including commodities so regions are less dependent on 1 source is sound. But I am not sure if this will occur. Partly for political reasons, I doubt the politicians will rise their petty provincial stupidities, and partly because manglement has not figured out even now that being dependent on 1 source is basically stupid.
These initiatives tend to fail because they are using government funding to try to cover for the shortcomings of an economic system that's dominated by finance. Its possible that they might get somewhere because the EU has managed to build infrastructure but the result in the UK has always been a sorry mess. Its not the government that's the problem -- often their heart is in the right place -- but expecting an enterprise to prosper in an environment that requires significant long term capital investment with no certainty of making obscene profits is incompatible with the kind of mindset needed to develop a competitive semiconductor product mix. The most likely outcome is that the initiative will fall behind, be soundly criticized by politicians for being 'a waste of public money', be sold off to 'the industry' and dismantled like a butchered carcass for any prime bits worth keeping.
The UK seems to be worse off than the Continent in this regard. The Continent has managed to produce large scale enterprises like Airbus but I suspect that any attempt to make another large scale industry will by stymied by all of the 27 countries wanting 'their' piece of the action (Europe was a lot simpler back when it stopped at The Wall!).
The UK seems to be worse off than the Continent in this regard.
Because, In My Opinion, of UK "Common Sense" Management Culture that favours the rise of penny-pinching and corner-cutting through-and-through cunts into postions of great importance. "Sir" David Frost being a prime example of British Management Culture distilled into one pungent bag of disaster.
Nobody who has the choice wants to work, or do their best work, for any of those bastards, resulting in a large "cottage industry" in the UK of highly skilled and creative engineers and designers. But, alas, these cannot afford to drop 200 kGBP on Tory Party Donations so they will never get any "government" work!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
