UK's competition regulator fires red flare over Nvidia's $40bn Arm takeover deal Britain's Competition and Markets Authority (CMA) has thrown a spanner into Nvidia's buyout of Arm, saying the deal presented "significant competition concerns" in a report published today. The CMA's July report highlighted the danger of Nvidia "foreclosing" Arm's intellectual property and lessening competition for industry- …
Maybe someone has remembered Cadbury?
But seriously, look what the American government has done to Huawei in the name of whatever bullshit reason they can give as long as it does not come across as protectionism. If Nvidia get Arm they can fuck the world over if they want.
Which may be why given what Arm actually earns each year, Nvidia are buying it for an absolute fortune. I may well be wrong as my memory is not what it used to be but didn't Analysts expect Softbank to sell for about 18 billion?
Anyway, sod all that it's Friday. Happy boozing y'all
Sure, China isn't exactly squeaky clean on the protectionism front, but allowing American capitalism full reign to warp and fiddle with the free market, because Mr Chin has to shop in China's closed market, doesn't sit well with the rest of the planet.
Free thinking people the world over, are sick and fed up of America's never ending crusade to own it all, and then double the price, while halving the quality.
Kraft buying Cadbury, is a perfect example, might also add in Terry's of York, Why don't they just buy Santa's sleigh as well and properly fuck Christmas up for ever.
I guess from that view of things you aren't looking forward to the gradual introduction of vomit flavoured US chocolate? Only thing worse is "liquorice flavoured boiled sweets" from Norway (perhaps Sweden?) that were sold by Lidl - I thought they had a great "liqourice" flavour until I hit the " ammonia infused with mouldy haddock" flavoured centres :-(
If we can prevent the sale of a non-GB company (because we allowed the GB company to be sold to a foreign company as it obviously wasn't against the interests of GB) to another non-GB company as it's not in the interests of GB ... and at the same time prevent the horrendous potential of being overrun by foreign influenced, vomit flavoured so-called-chocolate fingers, that must be a good thing ... possibly.
It's spurred a lot fo chinese investment for exactly this reason. Odds are pretty good that the migration to it across chinese manufacturers (especially Huawei) is essentially unstoppable.
It's got to be noted that ARM has _already_ been restricted by the US government in what it can sell to China already. That's one of the reasons Softbank decided to put it on the block
At some point more sectors of the world are going to take the banking approach to USA interference - ie: "refusing to do business with americans". It's already happening with logistics where shipments are both being routed around the USA and away from carriage on US-owned companies, thanks to Covid-related shenanigans
You're talking about different things though. This intervention is from the CMA on grounds of risk to competition. Softbank weren't a player in the chip market, and so there was no competition reason to look at the deal.
What you're asking for is some sort of industrial policy, where the government controls who can own what company - or some sort of security policy to stop ownership from foreign nations we don't like. Japan and the US are allies - so even there you're not really talking about a security barrier, so much as a foreign ownership barrier.
Now that's fine, it's a perfectly rational policy choice. But it does have costs. It's much more risky investing into an economy, if the government might intervene when it comes time to try and get your money out, by selling an asset. We've been one of the top global destinations for inward investment for the last 20-30 years - killing that off will have costs, so it's a policy you need to look at carefully.
Murdoch became American to buy USA Media.
Marconi was forced out of what became RCA by the USA Government. RCA died either 1976 or 1986, I forget which, and Thomson bought the labels.
You often have to spend USA "aid" on USA Military gear and support.
> You often have to spend USA "aid" on USA Military gear and support.
Ditto with British foreign aid. The strings on it almost always mandate dealing with British companies (I've worked at the receiving end and quickly observed the primary beneficiaries of such charity were companies from the "donating" country)
>What you're asking for is some sort of industrial policy, where the government controls who can own what company
I think many of us would settle for statutory protections for employment, investment, buildings, borrowing levels and so on.
Right now there's nothing stopping a PE group or tech firm (foreign or domestic!) using cheap/dark money to take advantage of the weak pound to buy out a British company, load it up with debt, sell off its IP, sell off and rent back its premises and then clear off. We can only object on competition grounds and soon national security grounds. There's no room for maneuver. This has already happened to Asda and will also happen to Morrisons soon enough. Let's not even begin to talk about Vectura.
But as you say that would require some form of industrial strategy. From a government who watched Kabul burn from a beach, wondering what the fuss was about.
We GAVE hundreds of billions of dollars worth of military training and equipment to the Afhgan Army so that they could protect themselves from the Taliban. This was after giving them billions of dollars in military equipment and support with more than a little American blood in a half-hearted attempt to rid them of the Taliban in the first place. Given the history of other attempts to invade Afghanistan I wasn't keen on the idea of helping them in the first place. The fact that they were allowing terrorist that had attacked the US to use their country for training bases and the undeniably evil deeds of their government made me hope that we could help them. I blame the eventual outcome on Bush and his puppet master Cheney. They had the Taliban cornered with their backs against the admittedly porous Pakistan wall when they decided to run off to invade Iraq before the toughest part of the job was finished. And 20 YEARS later when it came time for their army to stand on its own two feet it dropped to its knees and gave up without firing more than a few shots. I don't remember hearing about any battles as each city fell, just that all the cities fell with breathtaking swiftness.
I have a lot of sympathy for the Afghan people, especially the women who must once again live under the boots of a brutally repressive government but it was not the US that betrayed them. It was their own army and government, with their "president" flying off to Russia (Russia?) with car loads of cash unwittingly(?) provided by the US.
It's no doubt more complicated than that, as it always is but on a side note I cannot remember the Norwegian king being much criticized for fleeing to Britain when the country was invaded or the French for moving the gold to Britain too.
For some reason we don't know, the army decided not to fight, perhaps they too understand that they have to sort it out among themselves.
The Vietnamese did it too eventually by themselves.
This post has been deleted by its author
Adding to my own comment.
There are those too with better inside information regarding Afghanistan than the majority of "us".
They will hardly agree on everything.
I think this is worth listening to also.
https://www.youtube.com/watch?v=eJ4Y0NMDn8s
The U.S. ignored corruption within the Afghan government. Did that lead to its fall.
I think the Afghan army was poorly organized and had exaggerated their ability. We can yell cowardice but as a grunt if you see clearly your unit is in shambles and you'll be overrun anyway, why stay to die for nothing?
I don't know if that's the reason why they crumbled but the point is, the afghan army was not one entity but lots of units all over the place with people making their own decisons for their own reasons, facing an enemy who'll slaughter anyone, including families for fun.
One province is mounting an organized planned resistance I wish them the best but the country will probably be in civil way forever.
The "old" Taliban were mostly Pashtun locals and had family ties to the areas they were in.
The "current" Taliban are a different animal - afghanis have been complaining for a while that they're mostly foreigners who entered the country to fight the USA (the same thing happened in Iraq)
You can't occupy a country if the locals don't want you there (The USA found this out in the Philippines in 1899-1935 as well as Vietnam), but in this case the USA has given the "resistance" an enemy to unify against. I can't see "The Taliban" remaining unified for long - they're already fragmenting into infighting factions
It's not that simple.
For a start, money was involved: https://www.washingtonpost.com/world/2021/08/15/afghanistan-military-collapse-taliban/
We GAVE hundreds of billions of dollars worth of military training and equipment to the Afhgan Army so that they could protect themselves from the Taliban. This was after giving them billions of dollars in military equipment and support with more than a little American blood in a half-hearted attempt to rid them of the Taliban in the first place.
And much of that hardware was sufficently sophisticated that it could only be maintained by 16,000 civilian contractors who the US also provided (which is how foreign "military aid" works - thinly veiled state support for domestic industry). Kind of sounds like a few vendors we could all mention!
When the USA abandoned Bagram in the middle of the night without so much as a "fare thee well" to the Afghan Commander, the ANA came out to find they had an airfield full of complex aircraft they couldn't maintain or operate beyond the next 36 hours. Of course there were local techs being trained up, but the Afghan education sector had been... neglected over the past 30 years. They don't have that native industrial base to support the equipment and were still reliant on foreign expertise to oversee & train the technicians and sign off work. How long can you keep a C-130 going without manufacturer support?
At which point all their shiny, fancy gear was barely able to keep up with a bunch of thugs on a Toyota with a heavy machine gun.
"if the government might intervene when it comes time to try and get your money out, by selling an asset."
Inadvertently you've put your finger on the problem. It's investors wanting to come in and make a big capital killing. If all investors think like that why should anyone want to buy the company to get their money out? The only ways to keep playing that game is the pyramid scheme - you borrow more money, load the company up with debt and sell on to someone who'll do the same and hope that you're not the one left holding it when the debt burden makes it collapse or, alternatively, strip the assets.
The only thing that makes it worth buying successfully is take ongoing income from trading profits - but in that case the sort of investment a government should encourage is one that creates the business to do just that in the first place. It may well be that that's what they actually do want to encourage. It generates employment - always something that helps get them re-elected. It contributes to GDP. It contributes to tax income. It's just that over the years governments of all colours have just let anything worthwhile be bought out, stripped and closed down.
What you're asking for is some sort of industrial policy, where the government controls who can own what company - or some sort of security policy to stop ownership from foreign nations we don't like. Japan and the US are allies - so even there you're not really talking about a security barrier, so much as a foreign ownership barrier.
It's certainly an area that needs very careful thought and implementation (beyond most politicians).
That being said it would be nice if UKGov had an industrial policy. Particularly with respect to manufacturing, but also IP and services, rather than winging it on a case-by-case basis and largely letting the markets do what they like (until it all comes crashing down per 2008 and "lessons are learnt"... until the next time).
Not the sort of 1970s picking-winners industrial policy, but just a bit of a plan to support key industries (beyond inventing military projects to ward off the administrators) and grow strategic sectors.
RISC-V vs ARM......
Two points.
1) You have confused an architecture with an implementation.
Most companies don’t buy an ARM *license for ISA* and roll their own. That’s mostly not where the value is. The value is, that ARM have lots of talented engineers, who have designed a viciously efficient and low-power implementation IP of that ISA.
In contrast, anybody can make a RISC-V autogenerated IP. Or even buy an optimised implementation from SI5. But the *power and silicon area* of that is simply not at all competitive with the best ARM cores. Not close.
2) Over the years there have been literally dozens of “really quite good” CPU companies and cores. I’ve worked for a couple. And most of them have died. They have died because their angel investors have all found out, the hard way, that having a great hardware idea is maybe one quarter of what is required to commercially successful. The other is the software toolset: debugger, compiler etc. Almost all of them have seen the compiler as an afterthought, at best leveraging gcc, and at worst releasing something lashed together by a couple of engineers during the hardware design as a test bench. This is simply never going to be good enough.
To make a *competitive* software tool-chain for RISC-V, *from where they are today*, would require a bare minimum of maybe 300 engineers working for the next 5 years. That’s maybe $300m of investment. And that really is a bare minimum, it could easily be double that unless the execution is perfect, and that still only gets them where ARM is today, not where they will be in 5 years time, which is the graveyard catch-22 of the would-be-disruptive venture capitalist. Where would SI5 get that kind of money?
To point out just one project here, who do you think is going to develop Rosetta tool chain for RISC-V? Apple? They don’t expect to pay for that, they expect their *IP supplier* to have done it. I can name, off the top of my head, dozens of projects that need doing for a real contender. And by “doing”, I don’t mean switching the compile-flag in LLVM and just accepting whatever dross comes out. I mean *actually doing the work to have an efficient bytecode*.
I agree with IP concerning RISC-V transistor layout has a way to go to be upto ARM, but...
I don't agree tool chains are an issue, gcc/LLVM improvements of the past 30 years can all be leveraged. The ecos systems concerning tool chains are easily ported and reused.
Indeed the same is true of the other kinds of transistor IP what exists inside a modern SoC implementation that too can all be leveraged.
Besides the real tool chain improvements for a CPU target are to leverage a particular implementation of IP see Intel/AMD and M0/coretex/arm64. See also recently intel no longer maintains their own compiler, they use LLVM now.
Many applications for RISC-V are in the area of MCUs and overlap Microchips products lines, SoCs and multifunctional controller chips. The applications need reliability over optimisation, robustness over extreme low power, low cost over performance, low density technologies that can be fabricated on older process nodes. All this stuff is largely already here.
While SiFive maybe working on a version to compete with ARM64 in the datacentre with virtualization et al. SiFives holy grail, not mine. What has taken intel/amd/arm 20 years to develop is not going to take RISC-V 20 years to catch up to where we are today.
Yes they (RISC-V/SiFive) may always seem to be behind but I don't think they need to get at the front to be wildly successful.
You have been able to get a GCC for ARM and debuggers etc (GDB) for years and years.
Nevertheless, ARM still sell a compiler and debugger toolchain. And companies buy it.
This is because it's optimised for ARM, and you get about 10% better performance (either speed or space) out of it.
For a commercial product, this equates to 10% faster, or less battery consumption, or a bit of both.
Here’s a list of the JTAG debuggers I can buy for next-day delivery. Point to the one that supports RISCV please.
https://uk.farnell.com/c/development-boards-evaluation-tools/debuggers-emulators-jtag-tools-accessories/debuggers-emulators-jtag-tools
I’m sure One Exists, but here’s a scenario - I’m a contractor, and my washing machine development out in Bangalore is stalled on the production line, some software problem. No prob, I’ll just ring them up and get them to stick a JTAG debugger on it. Oh, *they haven’t got one?!* WTAF. And it takes six weeks to get the Special One through Indian customs?! Yes, I know somebody should have thought of that, but they didn’t, did they. It’s the exact equivalent of why you never, ever, specify an unusual screw that needs the special screwdriver. [hint: not an invented scenario].
Gcc/LLVM. “can be leveraged”.
By *whom* on *which budget*?!
E.g. who is supporting FORTRAN on RISCV, or COBOL, or HiPE for Erlang?
Of course, it could be done. But it isn’t. There are literally hundreds of things like this that just don’t work out of the box. Many of them may be minority sports. But no *company* is going to buy a computer where you have to ask whether $SpecialThing is supported. We’ve had x86 working across the ecosystem for so long, we don’t realise just how much work it took.
ARM have spent the last decade catching up to that level of support, siphoning a river of cash from a profitable business to do so. What pot of money is RISCV using to re-do that?
You’ve completely misunderstood “Intel use LLVM now”. No. They don’t. They have contributed millions of developer-hours of effort to LLVM, such that LLVM now produces great results on Intel architecture. There are no optimisation fairies.
Secondly, you are arguing what is always argued, that RISCV can sit in the space of not-demanding applications, like MCUs etc. But then, there are literally *dozens* of CPUs in that space, all of which are actually a great fit for what they do. What exactly is wrong with an ATMEGA808, which has sold literally billions, costs cents, and does what it needs to do? Or one of the many Texas Instruments devices, or Analog Devices?
Given that I have the tool chain for all of these installed on my computer, and it worked on my previous project, and I have a skeleton for a functional application already working, because the new embedded project is 90% like the old one, and I have a parts bin of thousands of ADI devices that I would like to use up, and a working dev board on my desk, and they only cost 5p each. What possible incentive are you going to offer me to change horses?! Makes no sense at all.
There’s plenty of room in the embedded space, and it’s all *filled* with good solutions. The only people punting RISCV for this, are hobbyists who only know the big three (by comparison) big-iron CPUs.
"What exactly is wrong with an ATMEGA808, which has sold literally billions, costs cents, and does what it needs to do? Or one of the many Texas Instruments devices, or Analog Devices"
When push comes to shove, they're IP whose distribution is controlled by the US government - and it HAS been increasingly controlling that distribution to try and industrially disadvantage China
Risc-v's advantage is that the US government can't control it. The US already demonstrated they had the power to block ARM selling into China and that's caused widespread reevaluation of what is used in domestically produced/designed devices for international sale
In short, US government protectionist policy is driving rapid development and uptake of RISC-v in ways that the LongSoong chips could never get traction on
Actually, reading the following links for a RISC-V CPU, made by Micro Magic, *really carefully, both of them* may help in the understanding of the issue.
https://arstechnica.com/gadgets/2020/12/new-risc-v-cpu-claims-recordbreaking-performance-per-watt
https://www.prnewswire.com/news-releases/micro-magic-inc-delivers-ultra-low-power-64-bit-risc-v-core-301227471.html
So, at first sight, isn’t that amazing! It’s RISC-V, it hits 5GHz, and it’s up to 10x more power efficient. Actually, no this isn’t anything great about RISC-V at all, let’s dig in and see why.
First off, it’s only reaching about 25% of the IPC of the ARM. Now that’s not terrible in itself, because if it’s so much more power efficient, can’t you just have more cores? Their CEO suggests 25 cores.
Errr, no, not really. Once we run it full clock to get the performance up, it’s 5x more efficient rather than 10x. But still we need 4x more cores to get the same SoC performance (which is what the user demands). But that means 4x more L1 and L2 *caches* too (L3 can be shared). Coremark stresses CPU grunt, intended to run from small cache. The silicon area on a real die is maybe 75-90% cache, depending on whether you count L1 cache in the CPU core. But wait, it’s much worse than that. The RISCV has less OOO circuitry, so its going to need more cache than ARM per core too. So, actually your RISCV SoC is going to be 5-10x silicon area. Thats 5-10x price. Oh. Oh very bad.
So, RISCV needs too many cores, and doesn’t address the high performance market, but can be a great fit for the Raspberry pi lowend huh? Errr, well, no. Because you’ve just done the comparison against ARMs high performance core, with all the OOO bells and whistles that eat over 75% of the silicon area budget of the core. If you chose a low performance ARM Cortex, competing against the RISC5 at 2GHz say, the ARM wins again. Not by so much, but enough to mean that given a level playing field the ARM will win billions of sales, and the RISCV will be lucky to make millions.
So, here’s the thing. VCs are (no longer) stupid. If this is so useless, what’s going on, why are Micro Magic releasing RISCV (if not to fleece uneducated investors). Actually, there is a good reason, Micro Magic have indeed done something amazing! They have got 16nm finfet to run at 0.35V, which is basically at threshold, which saves tons of power. If you read their company, they are chip tools and services company, not CPU experts.
The low-voltage transistor design is their secret sauce, which they are trying to sell. That would work on any CPU including ARM or Intel, it’s got nothing to do with RISCV at all. I hope they develop this, and win big. So if you want to make a production prototype to prove the patent is solid and worth billions, yes indeed absolutely autogenerate some random RISCV you’ve pulled off the Internet. The last thing you want to do is license from ARM, because there will be IP implications.
Let's face it, we are at peak ARM and Softbank is right to offload it. NVidia will extract their costs while the ARM stuff slowly dies over the next 20 years. Companies come and companies go. It's only a matter of time before Android supports RISC-V (or something similar comes along) and then who's paying for ARM designs?
It's a shame that the UK can't set up new companies with new ideas anymore... Now the UK Government wants to run dinosaur tech companies like British Leyland.
"NVidia will extract their costs while the ARM stuff slowly dies over the next 20 years"
The very first thing Nvidia did when acquiring established linux distributions for network switch chipsets was to rip Broadcom support out of them in order to promote their own switching silicon
There are very real fears that if they acquire ARM, they'll shut Broadcom out via licensing games
Nvidia are the current holders of the "buy it, then extinguish it" torch of anticompetitive behaviour
So Intel makes GPUs (the upcoming "Arc", the "Server GPU", as well as the built-in GPU-on-Chip stuff (IRISxe or whatever).
AMD makes GPUs (Radeon).
NVidia is unquestionably the dominant force for GPGPUs and general PC GPU technology.
ARM, though, has a very small footprint in places where NVidia is strong.
So the argument seems to be that NVidia could use ARM to displace.... INTEL?? If that were so, surely that's a GOOD thing?
(Yeah, sure, NVidia has their Tegra thingies, and hypothetically if NVidia owned ARM then instead of just using ARM inside the Tegra chips tied to NVidia cores they could.... use ARM better? Maybe NVidia-ARM would strong-arm Apple and Samsung into using CUDA instead of Mali? Oh, the horror!)
This "decision" seems to be an effort to rationalize an end state that has the UK worried, but I suspect it fundamentally has more to do with moving ARM from Cambridge to California...
China goes for riscv. Their numerous universities work on riscv designs. Arm is more desirable than Intel because they can do production themselves. One thing holding china back is wafer printing https://www.wsj.com/articles/china-wants-a-chip-machine-from-the-dutch-the-u-s-said-no-11626514513
The top 3 supercomputer in the world the Chinese Sunway uses riscv, the cpumesh is quite impressive assuming it is 28nm from 2016. So when they shrink this to 3nm?
This can all be quietly sorted out with a bung to party coffers. .... Anonymous Coward
Only in the world of puerile dreams, AC. Everywhere else does everyone compete against those failsafe sources providing saints and sinners and angels and daemons their special forces and almighty services which cannot be bought although you can pay handsomely for them to supply one and/or all unparalleled benefits for the overwhelming leverage that accompanies a totally unexpected inequitable advantage.
Strangely however...... although a bung to party coffers in an age of stealthy exploitable 0day vulnerabilities is never going to sort anything out quietly especially whenever it guarantees the exact opposite would more likely be true with such a bung to party coffers able to create an absolute stink to destroy such as is surely supportive of corruption ..... is it surprisingly worthy of enthusiastic support.
And that takes one into the world of quantum, where a this is a that and together something else altogether quite different and wonderful and/or deadly and Great Game Changing.
I'm expecting a million down votes for saying this but as part of the European Union we ceded control of competition policy. For many years the CMA had little sway over M&A over (I think) £100m. Whether other nations would have cared so much about ARM is unclear but more clear would have been our difficulty to exercise national concern.
While it seemed to be true that not every country adhered to the rules, that doesn't make a good counter argument
NVIDIA is fast becoming the new Intel and really needs breaking up. To make matters worse, they also funded a pork barrel supercomputer project in the UK near Cambridge so they could win UK Government plaudits. As far as I am aware no useful science has been done on that service, just some AI Graphics. Real scientists would do better working with the University. A real pro-British Politician would do better blocking this deal and help ARM float on the LSE.
They make what? Graphics cards, computing/machine learning cards, roughly based on the same cores and some mildly successful ARM SoC's with some their of Gfx/ml IP - mostly used by Nintendo and in cars. What do you want to break up? I guess they could outsource the HR dept.
One of the biggest sells on Brexit was the new trade deal with the US. They may not care in the US, or they may care that a Japanese company was allowed to buy ARM, but one of their larger tech companies is going to be blocked from buying it from the Japanese company...
Trade deals usually favour the stronger party (being in the EU used to help with that, but that's history) and the UK being one of the top ten economies will still do quiet well. However, when up against someone larger, e.g. the US, an already weak position is not going to be helped by this. If one looks at the mood in the US, Trump's America First voters number 70 million, and the Afghan surprise to add to that; don't expect a happy bunny in the US administration if this is blocked.
Maybe the USA might take us seriously if we did what’s in our interests not theirs!
TSR2…
Harrier for F35…
Cobham…
Not that it’s going to happen, but telling the US to go forth and multiply, as per Hugh grant in love actually once in a while wouldn’t necessarily be bad.
Oddly enough I have a 1/48 TSR-2 parked on a shelf above my desk! That was largely expediency, a unique British trait. If we can buy cheaper, why do it ourselves? Only country to give up satellite launch capability etc. However at times it can literally be a lifesaver, think Nimrod and the amount of money poured into that, and it took the deaths of an entire crew to finally give up. Now that we're out of the EU, telling the US to something itself is now a lot harder than it was previously.
I'll round off with a quote:
“An increase in inbound merger and acquisition activity was one of the obvious consequences of Brexit and weakened sterling, but few expected it to manifest itself so quickly or at so large a scale.”, Dan Ridsdale, analyst at Edison Investment Research.
Oddly as remainer, I made a good chunk of cash on the ARM takeover!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
