Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?

[Sal Ammoniac]

https://arstechnica.com/gadgets/2020/09/nvidia-reportedly-to-acquire-arm-holdings-from-softbank-for-40-billion/

[Kjelt]

It all depends what they are going to do.
If they are like the previous buyer and just let them do their stuff I doubt anything will change.
If they want a big ROI for their 40 billion some things might become more expensive.

[Sal Ammoniac]

If they want a big ROI for their 40 billion some things might become more expensive.

With RISC-V breathing down their neck, I don't think they can afford to increase their licensing fees much, if at all.

[Scrts]

If they want a big ROI for their 40 billion some things might become more expensive.

With RISC-V breathing down their neck, I don't think they can afford to increase their licensing fees much, if at all.

On general purpose small microcontrollers - maybe. However all heavy-lifting processors, safety critical processors and microcontrollers, DSP stuff are way more advanced.

[Muxr]

This is Nvidia continuing their "Embrace, extend, and extinguish". Time to avoid ARM MCUs for future projects imo. The only way this ($41B deal) is profitable to Nvidia is if they squeeze the market. ARM makes about $300M annually. At that rate it would take over 130 years to pay for itself.

[dmendesf]

Think they will make ultra high end cores just for them for supercomputer level stuff, more expensive cortex-A for everybody else (maybe with some twist like calling them "hands" to try not to give the ISA for free to Apple and other holders of perpetual licenses) and the microcontrollers will just die as RISC-V ate the lunch.

[james_s]

Pretty hard to avoid ARM cores these days, they are made by so many companies now, it's clear ARM is the future of microcontrollers, at least the foreseeable future. If NVidia killed off the company they would just be made by dozens of different companies in China.

[Sal Ammoniac]

If NVidia killed off the company

That's not possible... It's even in the US Constitution: "The right of the people to keep and bear ARMs shall not be infringed."

[Bassman59]

Pretty hard to avoid ARM cores these days, they are made by so many companies now, it's clear ARM is the future of microcontrollers, at least the foreseeable future. If NVidia killed off the company they would just be made by dozens of different companies in China.

The former is true -- between ST, TI, SiLabs, NXP, Microchip and others, ARM is everywhere and I can't imagine that those companies would stand for having their product lines destroyed by NVidia.

As for the latter, killing off ARM, I don't even see why NVidia would move to eliminate what is basically an easy revenue stream. Sure, NVidia could introduce their own line of ARM microcontrollers and microprocessors to compete with their licensees, but really the competition is all about features other than the processor core.

One has to wonder if any of the vendors I mention have background projects based on alternative architectures, whether RISC-V or something else. Microchip still has MIPS products. (I don't know enough about MIPS to guess why that architecture fell from favor.)

[brucehoult]

If they want a big ROI for their 40 billion some things might become more expensive.

With RISC-V breathing down their neck, I don't think they can afford to increase their licensing fees much, if at all.

On general purpose small microcontrollers - maybe. However all heavy-lifting processors, safety critical processors and microcontrollers, DSP stuff are way more advanced.

I won't name companies, but RISC-V is heading into cars and airliner avionics right now.

Heavy lifting ... it was announced today that a RISC-V PC with A53-class CPU will be demonstrated at the Linley Processor Conference in October.

https://www.businesswire.com/news/home/20200914005108/en/SiFive-To-Introduce-New-RISC-V-Processor-Architecture-and-RISC-V-PC-at-Linley-Fall-Virtual-Processor-Conference

If they've hit their MHz targets then that should come in at significantly better performance than a Raspberry Pi 3 or 3+, but below the Pi 4. So calling it a "PC" is maybe a stretch. In conventional PC terms it's probably around the last Pentium 3 or PowerPC G4 machines -- but quad core and 64 bit. So more like the first Kentsfield Core2 Quad's in early 2007 (and similar MHz) but maybe 1.5x to 2x slower microarchitecture.

That U74 core was announced just under two years ago, and SiFive's A72-class U84 core was announced just under one year ago, so you can probably expect samples and demonstration of an A72-class SoC in about a year from now.

Alibaba has announced an SoC with A73-class cores (Xuantie-910) and say they already have it in production use in their datacenter. It would be very interesting if they put those up for retail sale.

Of course RISC-V is behind ARM and coming from a standing start, but the gap is narrowing.

[westfw]

I worry what this will mean for the Cortex-M "embedded" ARMS (Cortex-R, too, I guess.)
NVidia can be predatory or not with the high-end phone/tablet/desktop AI/GPU/64bit IP, and the industry can complain, take legal action, or follow alternatives.  But I don't see NVidia as being very interested in the low-end, and it could just get neglected into irrelevance :-(  (and ARM has been relatively active.  M3, M0, M0+, M4, M7, M23, M33...)

With RISC-V breathing down their neck

Remember that "RISC-V" is just an ISA, not an (essentially complete) core design that a (relatively clueless) semiconductor company can buy and plunk on silicon "easy peasy" (with customers having a high degree of confidence that they'll work and be compatible.)  The equivalents of ARM as an IP provider are the likes of SiFive (um... That might be all?  the other RiscV members seem to be mostly silicon vendors?  Open Source?)  And they're comparatively tiny.

[brucehoult]

With RISC-V breathing down their neck

Remember that "RISC-V" is just an ISA, not an (essentially complete) core design that a (relatively clueless) semiconductor company can buy and plunk on silicon "easy peasy" (with customers having a high degree of confidence that they'll work and be compatible.)  The equivalents of ARM as an IP provider are the likes of SiFive (um... That might be all?  the other RiscV members seem to be mostly silicon vendors?  Open Source?)  And they're comparatively tiny.

Andes license cores. They do more RISC-V deals than SiFive does (at least in 2019, I don't know figures for this year) because they're a long established IP vendor who had their own ISA (supported by the Linux kernel, gcc, llvm) and a large existing customer base. They decided severa years ago to switch wholesale to RISC-V and plonked RISC-V cores into their existing ecosystem. They've said in 2019 their RISC-V sales overtook their legacy NDS32 sales.

In Europe the PULP project RI5CY and Zero-RISCY cores seem to be going into a significant number of designs (and NXP is supporting that).

There are a bunch of other RISC-V core IP vendors such as CloudBEAR, Codasip, IQonIC, Nuclei, Syntacore. I don't know what kind of sales they have.

[ejeffrey]

Pretty hard to avoid ARM cores these days, they are made by so many companies now, it's clear ARM is the future of microcontrollers, at least the foreseeable future. If NVidia killed off the company they would just be made by dozens of different companies in China.

The former is true -- between ST, TI, SiLabs, NXP, Microchip and others, ARM is everywhere and I can't imagine that those companies would stand for having their product lines destroyed by NVidia.

As for the latter, killing off ARM, I don't even see why NVidia would move to eliminate what is basically an easy revenue stream. Sure, NVidia could introduce their own line of ARM microcontrollers and microprocessors to compete with their licensees, but really the competition is all about features other than the processor core.

It makes no logical sense to kill of ARM but that doesn't mean that internal politics and shareholder pressure won't make them do it anyway.  It would by implementing increasingly onerous licensing requirements or dicking around with qualcomm style patent extortion that would be annoying and expensive enough that customers are willing to sink some money into developing the RISC V ecosystem. 

This isn't a question of something happening overnight.  Nvidia has claimed it is business as usual.  The question is what do they do over the next 5-8 years. 

[boz]

When I see a $1 RISC-V in LQFP that I can buy from digikey.. Sign me up, until then ARM all the way

[David Hess]

My guess is that future licenseable ARM cores will be crippled in favor of the processors NVIDIA sells.  This will not affect companies that have an architecture licenses and design their own cores like Apple, at least initially, but it will sure affect all of Apple's competitors including Samsung who not long ago disbanded their development team.

ARM documentation previously made publically available by ARM will be placed behind NDAs, or worse.

And if Oracle wins their copyright lawsuit over APIs, NVIDIA will release hell on developers.

[ejeffrey]

When I see a $1 RISC-V in LQFP that I can buy from digikey.. Sign me up, until then ARM all the way

That's not the way to look at it.  Its "When the people making $1 microcontrollers decide to start using RISC-V".  Right now, the ARM ecosystem is in pretty good shape and the licensing costs are low enough there isn't a lot of incentive for chip designers to rock the boat.  But if NVidia decides to extract more cash out of customers to justify the acquisition they are going to be initially successful but create a lot of incentive for an alternative.  I have no idea what NVidia is going to do, but "lets squeeze our captive customers for a lot of money this quarter, long term viability be damned" is definitely the kind of failure mode that a lot of businesses demonstrate so it isn't out of the question.

[JPortici]

I won't name companies, but RISC-V is heading into cars and airliner avionics right now.

Infotainment? ADAS?
I really don't see RISC-V going anywhere near the engines right now as i don't know of any implementation from the big guys. And in any case ARM had just begun entering that space, TriCore and Power Core still going strong..
Also i don't see it being used for dumber actuators, still probably Power Core if safety critical like fuel pump controllers and 8bit ST/PIC if really dumb like door or seat controller.
so if i had to guess, probably some coprocessor for ADAS applications..

I too fear of what's going to happen to Cortex R. Cortex M, i don't care that much

[ali_asadzadeh]

So can we hope for more AI and Graphics into ARM Cortex M profile soon!

[brucehoult]

When I see a $1 RISC-V in LQFP that I can buy from digikey.. Sign me up, until then ARM all the way

Haven't hit $1 yet, but there have been $5 chips for about three years already:

https://www.mouser.com/ProductDetail/Crowd-Supply/cs-hifiveb-02?qs=Zz7%252BYVVL6bGRLEumsP1bRg%3D%3D

(it's not clear from that page, but it's a 5-pack)

[iMo]

..
One has to wonder if any of the vendors I mention have background projects based on alternative architectures, whether RISC-V or something else. Microchip still has MIPS products. (I don't know enough about MIPS to guess why that architecture fell from favor.)

MIPS has been sold 3x times in last 7y, afaik. Some of the buyers reported losses, afaik..

[brucehoult]

Or $9.50 for dual core 64 bit 800 MHz with 8 MB onboard RAM and "AI accelerators" (neural network, AES, SHA256, FFT). Includes an MMU (though nonstandard) so you can technically squeeze some kind of very minimal Linus on it)

https://www.mouser.com/ProductDetail/Seeed-Studio/114991695?qs=u16ybLDytRZ0%252BIE52cskuA%3D%3D

[brucehoult]

Or $1.70 from AliExpress for a 108 MHz STM32F103 clone with RISC-V core from a Chinese company that has been making ARM based STM32F103 clones for a few years.

https://www.aliexpress.com/item/4000843847611.html

[Jeroen3]

I don't know yet. People where screaming like their house was on fire when GitHub was bought by Microsoft. Yet, Microsoft hasn't renamed it to Azure DevOps yet.

This is kind of the same I think, but at a much larger scale.
And I don't know if it happened in history before that someone acquired the body behind the industry standard.

It would be the same as Daimler buying Bosch gmbh.

I am not worried, one bad move from Nvidia and a lot of vendors will diversify.

[David Hess]

And I don't know if it happened in history before that someone acquired the body behind the industry standard.

I see the situation more like when 3dfx bought STB, which put them in competition with their customers.  Do you think NVIDIA will resist manipulating the ARM standards to the detriment of their competitors, who are also their customers through ARM?

[Jeroen3]

I think what will happen is that Nvidia will optimize future ARM to fit their proprietary architecture to work better with ARM then the competitor.
Perhaps even bundle discounts.
This way competitors are legally not harmed in any way, but in subtle ways their developments are disturbed.

Like a rediculous stupid change sold as "revolutionary", but since you're the first to know and have adapted all your products, the competitors have delays of years.

A bit like Apple saying (we'll know tomorrow), as of next year no more x86 macbooks will be sold!
Companies like Adobe will have a problem.