NVIDIA buys ARM: consequences for the industry

ARM, NVIDIA: grave

The news had already been around for a while. Rumors of a possible purchase of ARM by NVIDIA they were getting stronger. Despite that, there was still some hope, such as that the acquisition would not be accepted, or that Boris Johnson would veto the purchase to save the Cambridge headquarters in the United Kingdom. But none of that has happened and the worst expectations have been met. Nor is it that there was much hope ...

Many may think that the movement is a positive thing, and the truth is that it depends for whom. NVIDIA has certainly not paid 40.000 million for pleasure. The move will bring him great benefits and a much more dominant position in the sector, but of course there will be many victims along the way, and the main one could even be the ARM itself.

Introduction to ARM

ARM logo

Acorn Computer It was a company founded by Hermann Hauser and Chris Curry, and with a project led by Sophie Wilson and Steve Furber, among others. This would begin to develop the ARM architecture in 1983, launching its first product based on it in 1987. The initial objective was to develop an advanced processor with an architecture similar to the MOS 6502, RISC type. This way he could empower his line of personal computers to replace the 6502 chips on which they were based then, and whose developers were comfortable.

In the beginning, Acorn RISC Machine (later Advanced RISC Machine) had practically no interest beyond these proprietary products. But with the advent of mobile devices, their good relationship performance-energy efficiency, put them on everyone's goal. They went from being almost a rarity to being in a multitude of devices, from modems, routers, TVs, as controllers of a multitude of devices, to mobile devices.

In recent years, companies like Cavium (now owned by Marvel) with its ThunderX, Amazon with its Graviton, Fujitsu with its A64FXHe himself EPI project, etc., were interested in ARM beyond mobile electronics, to implement them also in the field of HPC, As the data centers. And not only that, some were also creating high-performance chips to start powering some PCs, such as certain Chromebooks, the Cupertino company itself with its Apple silicon once they have dumped Intel, etc.

In short, "overnight," ARM went from being little known to practically Everywhere. Something that is also reminiscent of the progression of the Linux kernel itself ...

In short, Arm's new business was transformed, becoming one of the most sought-after companies. The Japanese Giant Softbank made a disastrous move for Europe, buying the company for 28.950 million euros, taking away one of the most important technological properties that remained in the Old Continent (which is in clear technological dependence on the US and China). But the Japanese had its property for a short time, since the purchase took place in 2016 and in 2020 it has been sold ...

And not precisely because the business was not profitable, since the way in which Arm has to contribute its technology to others is quite productive. And he does it with two different models:

  • Allows use your ISA ARM, that is to say, the repertoire of instructions that they have designed. Anyone who wants to use it can, as Fujitsu has done for its A64FX chips, which are a microarchitecture designed from scratch using this set of instructions. Apple Silicon is also another of these cases, using an ISA ARM, but with a design made by Apple itself for its A-Series chips.
  • Another possibility that it offers is the licensing of its IP cores already designed. That is, to provide a ready-made microarchitecture so that other designers can integrate them into their own designs. This is the case of the EPI project itself (ARM CPU + RISC-V accelerators), or of the majority of SoCs for mobile devices, such as Qualcomm Snapdragon, Samsung Exynos, Mediatek Helio, HiSilicon Kirin, etc., which integrate one or more of the Cortex-A, Cortex-M cores,… In this case, the costly process of designing the microarchitecture is saved.
Do not confuse the ISA, which is the definition of a series of instructions that can be executed, the types of data that can be handled, the format, ... with the microarchitecture, which is nothing more than the physical implementation of the design capable of executing said instructions defined in the ISA. The same ISA can be implemented in many ways, that is, there can be many microarchitectures, but the same microarchitecture cannot be compatible with several ISAs, at least natively without emulators or similar tricks.

In both cases, Arm is paid for obtaining said benefit ... A benefit that has not been the NVIDIA's main booster to buy the Arm division from SoftBank, since Graphzilla's interests go beyond that, and are focused on obtaining a new advantage and dominance in certain sectors as I will now detail. By the way, as you already know, the purchase will close for 40.000 million dollars, about 33.770 million euros.

Also, that model is the basis of Arm's success. If it were removed, the success could disappear and go from a triumphant Arm to a mere tool for the benefit of NVIDIA. And I want to make it clear that this is not a monologue against NVIDIA, but it must be recognized that it involves very serious risks for everyone. Never before has there been so much concern about an acquisition of this caliber.

Who will be affected by NVIDIA's purchase of ARM?

ARM chip

Some Authorized voices and industry analysts of technology, and voices quite close to Arm, assure that this agreement could spell the end of Arm itself, at least as you know it now. Becoming one more product of NVIDIA stuff to monopolize some sectors in which now NVIDIA cannot dominate in front of other big companies.

Although many assure that The models of IP cores or the use of the ISA would remain intact after the agreement, not all are equally safe. Therefore, the fact that they stopped providing any of them would mean a great loss and setback for companies such as Samsung, Qualcomm, Mediatek, and a long etc. that now depend on one of those models.

By the way, I am always citing companies like Samsung, Qualcomm, Mediatek, HiSilicon, Apple, etc., but they are not the only ones. Others like Intel, AMD (for their Security Processors), Rockchip, Marvel, Renesas, STMicroelectronics, NXP, Amazon, Fujitsu, Broadcom (for, among other things, the chips of the Raspberry Pi) and many more. All of them are now under significant uncertainty, and many of them are direct competitors of NVIDIA in some sectors ...

What's more, now belonging to an American company, Trump's veto in his trade wars against China or Europe, he could also prevent some from using Arm's technology, which would be truly catastrophic. And is that designing a competitive microarchitecture from scratch is not a matter of days or months, it takes a lot of time and money, so it would put many companies at a clear disadvantage.

The danger is not so much in NVIDIA itself, as in the laws of the United States that will now control what is done with Arm. In fact, Hermann Hauser, Arm's co-founder, wrote a letter to UK Prime Minister Boris Johnson asking him to prevent the purchase of the company. Something you can put 0 hope on considering that it is about Boris and the relations that exist between the United Kingdom and the United States. Hermann himself assures that the technological and commercial war between the United States and China will leave collateral damage. that will affect the United Kingdom. It even claims that it puts Arm's current business models at risk, which are something of a "Swiss" model of the industry with more than 500 licensees, many of them competitors of NVIDIA itself. All of them could lose that current favor treatment.

Also, voices like Ryan Smith, of AnandTech, has expressed that the agreement for the acquisition was the easy part of the deal. The difficult thing now will be to convince all those consumers who now depend on Arm to stay.

One of the most critical voices has been a retired ex-engineer from the industry, as is Chia Kok Hua. He has been alarming about this possible purchase for some time before it occurred, and he claims to have first-source information about the agreement, assuring that it is not good. He goes further and assures that, even if NVIDIA maintains the business as it is now, its competitors would not have much to do because of the advantage that NVIDIA would now have.

Why do you say that? Well, simple, and now Arm was only a designer who was not dedicated to producing or selling his own designs, but simply developing for others. Therefore, it was not a threat, but simply a source of technology. Instead, NVIDIA will not only be a source, but also a competitor, and will do its best to profit from it, do not hesitate. He will use his status to gain an advantage over the rest.

For example, you could introduce changes in the ISA or the ecosystem for your own benefit, which would not be suitable for other designers.

NVIDIA the great beneficiary

NVIDIA Logo

El great beneficiary of this movement is NVIDIA itself. Arm's move has not been the only one that puts her in a much more advantageous situation:

  1. NVIDIA acquired Melanox for 6.900 billion dollars last year. Therefore, it keeps the InfiniBand and Ethernet technologies that this company had in its possession. In other words, key technologies for high-speed networks used in the HPC sector. Now that technology is under the designation NVIDIA Networking.
  2. NVIDIA acquires Arm for 40.000 million dollars. Yes, almost $ 47.000 billion in spending, but now you are in a position that is going to bring you much more than that.

What do I mean by this? Well, simple, and NVIDIA is now in the best position to dominate in the HPC sector, and even put very powerful companies at risk as I detail in the next section. Reason? No other has comprehensive solutions, NVIDIA now does: ARM CPUs + GPUs + Networks. Who can match that?

x86 in danger

NVIDIA supercomputer

As I mentioned in the previous section, even x86 could be in serious jeopardy after this move from NVIDIA. At least in the HPC and we will see if also in other sectors, and I am not saying so. There is also some concern among certain workers in companies such as Intel and AMD, which could be other collateral victims of this agreement between Graphzilla and Arm.

In the data center industry, Arm is becoming increasingly important, and now with all these acquisitions, NVIDIA will be able to position itself as an undisputed leader and displace the Intel Xeon and AMD EPYC chips, which until now needed together with its GPUs, but now no longer.

Remember that AMD is a somewhat weaker company than Intel, and could be one of the main affected. And after this rebound with his Zen, a new setback could be taken with this movement of his direct competitor in the field of graphics. Intel is the giant, the Chipzilla, but a very, very weakened Chipzilla and in a situation that is not the best, so a slight breeze could shake its leadership ...

Intel has a strong market for CPUs, but is still weak when it comes to GPU solutions, despite Intel Xe. AMD is the opposite, it is relatively strong in GPUs, but its CPU market share is not as strong as Intel's, even though Zen has managed to put Intel on the ropes. Instead, NVIDIA now has all the strengths after the purchase ...

I insist, you have to take your hat off to the NVIDIA movement, which will win a lot, but could bring serious problems for the rest. Therefore, even if it is a masterful and strategic movement, it is not hopeful at all. In fact, monopolization and these abysmal advantages always end up hurting the users themselves… Jensen Huang, CEO of NVIDIA, has been clever, but it is a move that some like Kevin Krewell describe as «very reckless move«.

And by the way, Apple, which has gotten rid of Intel and has started its own path with Apple silicon based on the ISA ARM, could also be seriously affected. Some analysts say they have two options, or try to stop NVIDIA, or take an alternative path. The first I do not see, since Apple does not compete directly against NVIDIA in the computer sector, and allocating resources for a fight of this caliber would not be beneficial. But the second alternative is not a cheap and short-term solution either ...

And Raspberry Pi and Arduino?

Raspberry PiArduino

It is also rumored that the SBC Raspberry Pi It is in danger, because it uses Broadcom ARM chips. But not much has been said about the development board Arduino, the other free platform that also has some ARM-based boards, and not just Atmel Atmega chips.

Depending on the license agreements that NVIDIA maintains, they could be affected to a greater or lesser extent. It's still too early to decide, but it wouldn't hurt to ditch ARM and look to RISC-V, which is an open ISA. In fact, there are already some very interesting RISC-V development boards ...

Currently, Broadcom is the one that makes the SoC for the Raspberry Pi, while Atmel makes it for the Arduino. Depending on the I try that they have these with NVIDIA will depend on the future of these boards.

Eye! The same can be said of all those compatible plates or similar that exist in the market, such as ODROID, Orange Pi, Banana Pi, UDOO, and boards such as Beagle, Tenssy, etc.

ARM microcontrollers

Cortex M, MCU, microcontroller

Another of the great forgotten analyzes that I have seen is the Cortex-M, Arm's series of MCUs or microcontrollers. These chips are intended for other purposes, such as embedded or embedded devices, industrial machinery, vehicles, IoT, everyday consumer devices, etc.

This line will be profitable for NVIDIA? Depending on whether it is in NVIDIA's interest or not, it could be somewhat relegated, causing many who now rely on it to lose these IP designs. And that would go beyond the very sector of computers and supercomputers.

The same is true for ARM Cortex-R, another series of RISC CPUs based on ARM and, in this case, optimized for secure and critical applications, as well as for Real-Time (real time). Another key piece for certain industrial and other applications.

Of course, although it is not so popular, it is a very juicy sector. And again we find another question. And, although NVIDIA maintains the development of these MCUs, it could get a gran ventaja of this in key sectors for the future such as vehicles, IoT, etc. Great benefit of one, detriment of many ...

Collateral affected: RISC-V

RISC-V logo

NVIDIA is not the only beneficiary of this move in the purchase of Arm. There is another beneficiary, but almost collaterally. Without looking for it, Isa Risc-V could be the big winner, as many of those current disgruntled Arm customers could end up calling in RISC-V, which would attract more investment, development, and empowerment of the ecosystem.

By the way, even though NVIDIA is one of the partners RISC-V Foundation, don't think he did it with them in mind. In fact, if RISC-V begins to gain followers after the movement, it would also become an enemy for NVIDIA itself. So we'll see what happens to your current contribution ...

Mike Demler, one of the senior analysts at The Linley Group, has also assured that «key people could defect. More Arm customers could look to RISC-V«, Referring to the agreement between NVIDIA and Arm.

Maybe others like MIPS and OpenPOWER They could also benefit from this deal, as these ISAs could charge interest if NVIDIA decides to do weird tricks with ARM. We will see…

Conclusion 

Ultimately, the move is a clear step forward for NVIDIA, but a serious throwback for all others. And even if they kept their current Arm customers and did not affect the current licensing models, NVIDIA itself will gain an advantage in the market, and that could have its consequences for customers and users as well, as competition will suffer.

Much certainty that NVIDIA will win, but a lot of uncertainty in everything else ... Time will tell.

Now the famous phrase that Linus Torvalds pronounced a long time ago referring to NVIDIA ... maybe now it makes a little more sense. And now it will be time to look towards RISC-V and wish him all the luck for the good of all ...


4 comments, leave yours

Leave a Comment

Your email address will not be published. Required fields are marked with *

*

*

  1. Responsible for the data: AB Internet Networks 2008 SL
  2. Purpose of the data: Control SPAM, comment management.
  3. Legitimation: Your consent
  4. Communication of the data: The data will not be communicated to third parties except by legal obligation.
  5. Data storage: Database hosted by Occentus Networks (EU)
  6. Rights: At any time you can limit, recover and delete your information.

  1.   qtrit said

    Intel had already been threatened for several years, seeing how the train was coming on, between the kick of apple (well deserved) and that they have been selling "Refritos" for almost 6 years, what Intel has on the table is 100% his fault .

    Honestly, I think Intel does something very transgressive for the next 2 or 4 years maximum or they are going to have a historical business segmentation problem.

  2.   FAMMMG said

    This will lead to dispersion in architectures.
    It will be the same as google and android, at first everyone happy and then everyone for their interest but with the same ARM dependency.

  3.   Miguel said

    YA la Jodio, as always the insane and destroying Monopoly of healthy competition and especially the consumer ends up affected, the same one who ends up paying the bill for these infamous monopoly practices

  4.   Carlos Sapa said

    As a collector and fan of the defunct 3dfx, bought by nvidia, and with the background of nvidia, if the purchase is approved it is the end of ARM as we know it, there will be no more licenses, just that I just bought an amlogic s922x to experiment with Linux after having returned from my Mandrake days