Google, Microsoft, Qualcomm Protest Nvidia's Arm Acquisition (bloomberg.com) 47
Some of the world's largest technology companies are complaining to U.S. antitrust regulators about Nvidia's acquisition of Arm because the deal will harm competition in an area of the industry that is vital to their businesses. Alphabet's Google, Microsoft and Qualcomm are among companies worried about the $40 billion deal and are asking antitrust officials to intervene, Bloomberg News reported Friday, following up on CNBC's report from earlier today that talked only about Qualcomm's efforts. At least one of the companies wants the deal killed, Bloomberg added. From the report: The acquisition would give Nvidia control over a critical supplier that licenses essential chip technology to the likes of Apple, Intel, Samsung Electronics, Amazon.com and China's Huawei Technologies. U.K.-based Arm is known as the Switzerland of the industry because it licenses chip designs and related software code to all comers, rather than competing against semiconductor companies. The concern is that if Nvidia owns Arm, it could limit rivals' access to the technology or raise the cost of access.
It's bad for everyone (Score:5, Interesting)
This buyout is bad for everyone except Nvidia. The only bright side is that if it happens, it could push smartphones onto RISC-V CPUs in the future - although I don't think there are any with similar power usage and performance to ARM CPUs right now.
Re: It's bad for everyone (Score:2)
And that's barely a feasible option. There is a big cost to having a fragmented ecosystem of incompatible chips. Probably if the merger goes through there will be several companies doing high performance custom ARM core designs like Apple. They will have to pay a license cost to Nvidia but will sue if Nvidia raises it too much.
Re: (Score:2)
It would actually give FLOSS OSes and apps an advantage since it would be easy to recompile apps for the new architecture. Closed-source and multi-arch don't mix, but it's not a huge problem if the source is available.
Some app categories are hard to make Free (Score:2)
I'm curious how things like video games* and subscription streaming video players could be developed for distribution as free software. The business models of these segments of the industry have historically relied on hoarding source code. In the alternative, I'm curious how a smartphone platform can gain traction should it decide to just go without games and Netflix.
* Excluding computer adaptations of decades-old tabletop games such as Chess and Go
Re: (Score:2)
Well, today many games on phones are already being distributed as java which as any nerd knows is a higher level representation that gets JIT compiled to machine code or interpreted, depending. Java binaries contain far more information about the original source than C/C++ bins do, but this doesn't matter much.
Mainly this is because game software has always been fairly easy to pirate and very little of the source is cutting edge or innovative. In fact, the best currently available game engine (UE4) is ope
Re: (Score:2)
There is a big cost to having a fragmented ecosystem of incompatible chips.
Yes, we should have one manufacturer. It can be state run. Also, there's a cost to producing different -types- of chips. We should only have one microprocessor for all. Call it the "people's processor" , PP for short.
Excellent plan comrade.
Re: It's bad for everyone (Score:2)
I know you are trolling but I am not saying there will be any government action here. I am saying qualcomm and ti and mediatek are probably not going to go to risc.
Re: (Score:2)
ARM isn't just CPUs, it is an entire suite of components that work with their CPUs and each other. This deal will screw small companies who rely upon that suite.
Re: (Score:2)
They'll be pushed to use GPU cores owned by NVIDIA/ARM instead of different GPU cores owned by ARM ... such a huge difference.
Re: (Score:2)
So companies using ARM® CoreLinkTM NIC-400 Network Interconnect are now going to be pushed to use GPU cores? I had no idea network interconnects were so graphic core intensive, thank you for that information.
Re: (Score:2)
Those licensees don't compete with NVIDIA's existing business. It makes ARM money ... NVIDIA also wants to make money, they are not going to stop.
MALI is the only real thing ARM does which NVIDIA will want to suppress. ARM pushing one in house GPU solution or NVIDIA/ARM pushing another, meh.
Re: (Score:2)
Getting rid of MALI would be great. But it's not like nVidia has any GPUs worth shit...
Re: It's bad for everyone (Score:2)
Re: (Score:2)
Re: (Score:2)
Well, Linux works fine on my desktop (26 years and counting), and a proper RISCV machine (not a mere board) is coming my way. (Was supposed to ship Jan 1, goes Mar 12, aargh).
Re: (Score:2)
If Neoverse bleeds to death because the increased investment rate inspired by Softbank disappears and ARM retrenches on mobile, that could also be bad for everyone.
Re: (Score:2)
Wait, now Google, Microsoft, and Qualcomm are trying to stop the sale? A few days ago, according to this site, it was the International Communist Conspiracy doing it. Does this mean Google, Microsoft and Qualcomm are controlled by China?
The above is snark but it's a serious comment, why is it China one day and Not China a few days later?
Re: (Score:2)
Market consolidation for me, but not for thee.
The UK shouldn't have allowed this deal to happen! (Score:3)
A bit late to complain... (Score:2)
...as ARM has been owned by the Japanese company SoftBank since 2016.
Re: doesn't matter (Score:2)
Doesn't a purchase by nvidia put it under the jurisdiction of the US government?
That sounds like a massive difference to me.
Re:The UK shouldn't have allowed this deal to happ (Score:4, Informative)
Quite simply, this was one of the brightest of bright spots in the UK tech world. They should have blocked any deal for a foreign entity to buy it.
I don't know the full history of ARM, but do note that the party selling it to nVidia is Softbank. Of Japan. So it's at least one acquisition too late to be beating that drum.
Re:The UK shouldn't have allowed this deal to happ (Score:5, Informative)
One of the Acorn-created expansion units for the BBC Micro was a "Second Processor" unit - a faster 65C12 CPU and 64kb of slightly faster RAM, converting the host BBC into a raw input-output device.
Acorn developed some chip-design software for the "2nd Processor BBC" and one of their engineers came up with a design for a RISC-based, 32-bit CPU. All of the testing of the "ARM processor" the Acorn Risk Machine was basically done in software simulations on a BBC Micro with a 65C12 second processor.
The first full Acorn Risk Machine system, the Archimedes, produced a number of initial versions for which the A440 was the most powerful. It featured 4Mb RAM, 512kb ROM, an internal 40Mb MFM formatted hard drive and a 3.5" diskette. The entire machine comprised 4 main chips, some RAM and some very basic glue logic. The 4 chips were the ARM CPU, IOC (Input-Output Controller), MEMC (the Memory Controller) and VIDC (the Video Display Controller).
The first generation ARM had a 32-bit data register, a 26-bit address bus and an absolutely delightful instruction set with 44 basic instructions. However, each of these 44 could be dynamically extended because they could be combined with condition flags from registers that picked up signals from hardware. So you could have "ADD" to add registers together, or you could have "ADDCC" to mean "add these two registers only if the Carry flag is clear [0].
This basic conditional flags on the front of the core instructions helped usher in what would become the vastly more sophisticated pipelining and conditional execution/speculative execution that we see in today's latest processors.
Acorn Computers failed to keep pace with the sheer onslaught of the Wintel platform and was ultimately lost to history, but the ARM CPU design and chip family lived on.
The thing that helped this was the insanely simply instruction set - IIRC, that first ARM CPU consisted of all of 44,000 transistors. [The very first successfully manufactured part that came off the fab line worked, first attempt. Lots of good testing]. Despite the apparent simplicity, it was fast - it ran rings round the contemporary competition. Since those early days, ARM has evolved, expanding that knowledge of chip design by basically developing an approach for modular, multi-core CPUs that are incredibly power efficient.
And if you're getting the impression that I kinda liked my old Archimedes A440... well, you'd be right.
Intel and AMD are much faster and more powerful, I've no doubt. But there is something that is just so elegant about the ARM design.
Fugaku (Score:3)
No need to capitulate here. When speed, cost-per-core, and energy consumption are factored together, ARM actually trounces X86. This is demonstrated in the recent launch of the Fugaku supercomputer that occupies the top slot on the Top500 list [top500.org]. That beast has 7.6 million ARM cores and is three times faster than the next competitor. And unlike the GPU-based supercomputers on the list, it's a general purpose CPU which means it can crunch any proble
Re: Fugaku (Score:1)
repost? (Score:1)
Re: (Score:2)
Same story, different source - the other article referenced https://www.cnbc.com/2021/02/12/qualcomm-objects-to-nvidias-40-billion-arm-acquisition.html [cnbc.com]. The two articles were 2 hours 46 minutes apart, this must be close to a Slashdot record.
More like follow-up (Score:2)
This is a follow-up about Google and Microsoft joining Qualcomm's protest.
Re: (Score:2)
Good luck getting device manufacturers to agree on one set of extensions to RISC-V to form a viable ABI target. Otherwise, it's like ARM a few generations ago, where there were a couple incompatible floating-point instruction sets.
RISC-V (Score:2)
One thing for certain, Nvidia buying ARM will validate motives for the RISC-V permissive licensing model.
Risks, which are quantifiable liabilities, are accepted as a normal part of business whereas uncertainties, which are unquantifiable potential liabilities, are avoided like the plague. That puts a lot of pressure on large corporations with a lot at stake to avoid proprietary licenses on critical dependencies, because you can never know when the license holder, or new purchaser of the license, could squ
Re: RISC-V (Score:2)
Similar claims could be made of technology from companies that fall under the jurisdiction of repressive governments, as was demonstrated by the Trump administration and their "entity list".
Re: (Score:2)
RISC-V oversight and foundation has been moved to Switzerland specifically to make it neutral.
Wikipedia - RISC-V International [wikipedia.org]
Only we should have lock ins? (Score:1)
How are they rivals? (Score:2)
NVIDIA's embedded chip market share is tiny and if they have to promise to license them as IP instead of selling them directly they probably will, no skin off their back.
Re: (Score:2)
NVIDIA's embedded chip market share is tiny
And I'm sure they are happy with that, and are just acquiring ARM to serve as a benevolent benefactor of the IP.
Re: (Score:2)
There's profit in selling it as IP macros too, with ARM perhaps more than selling it as ICs.
Monopolies (Score:5, Funny)
I absolutely love watching a bunch of monopolies complain about a potential monopoly that puts their monopolies at risk. billionaires bitching that other billionaires may become more billionaire than they already are.
Re: (Score:1)
It shouldn't matter who's bitching. What should matter is if their reasons are valid or not. Nvidia owning ARM is a bad thing, regardless of who is saying so.
Re: (Score:2)
Re: Monopolies (Score:2)
It matters because it is more fun to watch...
Re: Monopolies (Score:2)
The bullies of tech... (Score:1)
The bullies of tech....now they don't want to be bullied, and don't want another big acquisition they didn't think to do. Me without my hankie, wait there it is....
Industry foundation (Score:1)