Nvidia Reportedly Could Be Pursuing ARM In Disruptive Acquisition Move (hothardware.com) 89
MojoKid writes: Word across a number of business and tech press publications tonight is that NVIDIA is reportedly pursuing a possible acquisition of Arm, the chip IP juggernaut that currently powers virtually every smartphone on the planet (including iPhones), to a myriad of devices in the IoT and embedded spaces, as well as supercomputing and in the datacenter. NVIDIA has risen in the ranks over the past few years to become a force in the chip industry, and more recently has even been trading places with Intel as the most valuable chipmaker in the United States, with a current market cap of $256 billion. NVIDIA has found major success in consumer and pro graphics, the data center, artificial intelligence/machine learning and automotive sectors in recent years, meanwhile CEO Jensen Huang has expressed a desire to further branch out into the growing Internet of Things (IoT) market, where Arm chip designs flourish. However, Arm's current parent company, SoftBank, is looking for a hefty return on its investment and Arm reportedly could be valued at around $44 billion, if it were to go public. A deal with NVIDIA, however, would short-circuit those IPO plans and potentially send shockwaves in the semiconductor market.
Re:SPARC? (Score:4, Interesting)
How exactly are they more enterprise ready than ARM?
AWS runs ARM instances now.
NVIDIA already uses ARM for their Tegra platform.
ARM is basically considered the #2 development platform right behind x86. Every major OS supports it, and client software followed quickly.
Re: (Score:2)
POWER just like SPARC seems to have designed for big iron enterprise DB applications. But these days, energy use and cooling are becoming very important even in data centers. ARM is great for that. With proper cooling and power that are available in a server or desktop enclosure, ARM chips will run circles around market leader chips.
Re: (Score:2)
Re: SPARC? (Score:2)
That worked a bit better a few years ago.
Today, AMD is running cool and Intel is making space heaters--and will apparently continue to for a couple more years.
Re:SPARC? (Score:5, Interesting)
AArch64 has been designed with hindsight. It isn't a 64-bit extension of ARM. It lacks the "free" barrel shift and predicates, it has twice as many general-purpose registers, the shift/rotate/mask instructions are completely different, the condition code handling is different. In fact, AArch64 is more like PowerPC's weird cousin. It's easier to make an efficient, modern implementation of AArch64 than it is for SPARC, PowerPC, MIPS, etc. because it's a cleaner, more modern architecture. PowerPC, SPARC and MIPS are decades old at this point, and weren't designed with deeply pipelined out-of-order multi-issue implementations in mind, there are trade-offs in the instruction encodings that turned out less than ideal, etc. Sometimes it's better to start over. The real competition for AArch64 should be RISC-V - it's designed with the same things in mind, but it's turning out to be a bit of a non-event.
Re: (Score:2)
The real competition for AArch64 should be RISC-V - it's designed with the same things in mind, but it's turning out to be a bit of a non-event.
I think some RTL being released isn't really the whole "event". ARM is popular in part because of its massive support structure and toolchains etc. It's also tried and proven. It will take some time for RISC-V to achieve the same level of support, but it's going to be a huge competitor to ARM's M lineup if/when it gets there. This purchase would be risky for Nvidia.
Re:SPARC? (Score:5, Insightful)
The fastest chip in the world has never been ARM
The world no longer cares about instructions per second.
What matters today is instructions per watt.
Re: (Score:2)
The world no longer cares about instructions per second.
What matters today is instructions per watt.
The world doesn't care for your absolutes. Instructions per second and instructions per watt are two different metrics used in different scenarios for different purposes. Just because you want extra battery life for your phone doesn't mean I don't want a blazingly fast workstation for AI data crunching. Your "world" and another persons "world" are ... errr... different worlds.
Re: (Score:2)
Power consumption matters even on the desktop because overall power consumption has increased* and energy prices have also increased. The power bill is an important part of TCO.
* Per-MIPS power consumption has decreased somewhat steadily over time, but system power consumption has gone up from its low point.
Re: SPARC? (Score:1)
Power consumption matters even on the desktop because overall power consumption has increased
Peak power consumption has increased; average consumption has likely held steady - or even dropped, thanks to better power management.
Re: (Score:2)
What matters today is instructions per watt.
More properly, instructions per second per watt, or instructions per joule.
Re: (Score:1)
They're buying the established ARM user base as well is what you're missing. It's hard to put a price tag on that.
Re: (Score:3)
SPARC IMHO has been dying a slow death for decades, except in big iron enterprise DB applications. This is why Sun Microsystems's stock price never recovered after the dot com crash and the company was eventually absorbed by Oracle. This RISC architecture has never been designed for low power or mobile apps originally and various SPARC implementations have always been trailing Intel x86/x86_64 in raw performance. Some argue this is because the SPARC architecture is overly complicated to reap the benefits of
Re: (Score:3)
SPARC died (whether it knew it or not) when Sun missed an entire generation because it was too slow to bother releasing it.
But really, it was obvious that it would be over for Sun a long time before that. Everyone else's SPARC processors were faster than theirs. We used Ross HyperSPARC at one of my employers, they absolutely spanked Sun's processors at the time. Then UltraSPARC came out and it was fairly speedy, but it was either grossly expensive or just in basically a PC with a SPARC in it, and still too
Re: (Score:2)
Ya, POWER is not designed for low power or small systems. One of POWER's (PowerPC in Apple was a different animal) key strengths is that it is designed for high volume enterprise workloads. Large cache's to keep the cores fed with data, and also very good multi-chip performance (due to designs of inter-chip communication).
Re: (Score:2)
Remember that Apple dumped PowerPC when it became clear that this is not a good architecture for laptops, much less mobile devices.
Apple dumped PowerPC because neither Motorola now IBM could cope with their demand for laptop chips. That is all.
Re: (Score:2)
Microsoft (Score:1)
Re: Microsoft (Score:2)
Wintel still being alive somewhere? Dunno.
Re: So disruptive! (Score:4, Funny)
Re: So disruptive! (Score:3, Funny)
But what about the chunnels of vertical scrum-AI integration?
Re: (Score:2)
Just talks so far (Score:2)
At this time is this all just talk and nothing has happened yet, with SoftBank even saying they might not sell at all when the price isn't right. Apple also must have shown interest, and also owned Arm once in the past, but currently must have bailed out of the talks. Even Intel has been mentioned as one of the possible buyers. SoftBank only ever bought Arm for profit. Who knows how this will end. I just hope it doesn't end in a disaster deal.
Re: (Score:3)
Softbank have already made noises about not getting enough money from ARM because they paid way too much for it, which is the start of the beginning of the end.
Long term a bunch of vultures will be picking over the bones of Softbank, but medium term Nvidia might be able to get ARM for a lot less than Softbank p
Re: (Score:2)
Then maybe Apple waits, and grabs nVidi-ARM after they consolidate. They don't have their own in-house GPU department yet...
Re: (Score:2)
Then maybe Apple waits, and grabs nVidi-ARM after they consolidate. They don't have their own in-house GPU department yet...
Um, wrong.
Apple has been designing their own GPUs for a few years now:
https://appleinsider.com/artic... [appleinsider.com]
Re: (Score:2)
They'd better do something to make up for the fiasco of WeWork.
Re: (Score:2)
AMD's stock is up through the roof. This would be the perfect time for them to do a stock swap. In one swoop, that would put them much closer to the size of Intel (at least in terms of deployed devices).
proprietary ? (Score:2)
Re: proprietary ? (Score:2)
In that case, everybody involved in ARM will chip in and buy Nvidia together, eviscerate their CEO, take out his appendix, throw away the CEO, and appoint the useless flap of colon as the new Nvidia CEO.
Re: (Score:2)
"There is another theory which states that this has already happened.” -- Douglas Adams
Re: (Score:2)
In that case, everybody involved in ARM will chip in and buy Nvidia together, eviscerate their CEO, take out his appendix, throw away the CEO, and appoint the useless flap of colon as the new Nvidia CEO.
When you dream man, you DO dream BIG! I hope this particular dream comes true.
Also, I'm guessing that you're a Philip K. Dick fan? Nothing specific in mind - there's just something about your whole vibe that reminds me of him.
Re: (Score:3)
I suspect that the long term effect would be to promote RISC-V.
Re: (Score:2)
nVidia could buy SiFive for a fraction of the cost of ARM and get ahead of the competition!
Re:proprietary ? (Score:4, Insightful)
So? SiFive isn't the only company working on RISC-V cores.
RISC-V is an open-source ISA. Seeing a commonly used core get locked down would show the benefit of an open-source solution.
Re: (Score:2)
NVDA has been using RISC-V [riscv.org] in their products [nvdla.org] longer than most other companies. So what? Maybe they should buy themselves?
For embedded software engineers and ASIC hardware engineers, RISC-V is a great deal. It's flexible enough to adapt to specific needs, but you get to leverage a common set of tools like compilers and verification tests. On the low-end it competes heavily with ARM Cortex-M0, on the high-end it isn't quite there yet. It theoretically can do enough DSP-like work to replace an embedded DSP or
Re: (Score:2)
That seems too RISC-E
Re: (Score:2)
What happens to Mali?
nVidia are a GPU company, first and foremost. I doubt nVidia are that keen on maintaining a graphics IP on behalf of their competitors.
Re: (Score:2)
ARMs are in everything - to lock it down would be commerically suicidal (for the future Arm division of nvidia, if they bought it). Not to say you couldn't just burn a load of lucrative contracts, but it's unlikely nvidia could buy Arm with petty cash they have laying about - more likely they'd have investors to pay off, so will need the income from existing Arm contracts.
A more likely move, IMHO, is some embrace and extend. That is, maybe some features that promote the use of nvidia chips alongside an arm
Re: (Score:2)
Doing so will destroy the value of ARM. Semicos are currently tied to ARM, not because they particularly want to but none want to be the first to try to take the hit on making a different platform viable. If a new ARM howner forces the issue by denying licensing then it will take less than 3 years before you get a huge surge of RISCV products with enough capital behind it to ensure a successful transition. Semicos will be vey happy to avoid ARM royalties.
Remember what "most valuable" means (Score:4, Insightful)
It does not mean real money. It means "current favorite toy of the gamblers".
Whose goal is to /extract/ as much real money as possible.
It says nothing about whether they will do that with or against Nvidia.
Re: (Score:2)
https://en.wikipedia.org/wiki/... [wikipedia.org]
I mean, they already do ARM cores... I guess buying ARM would save them licensing fees?
Re: (Score:2)
Selling to Nvidia could be very bad for ARM. For a start most ARM chips are made in China and used in Chinese products, so coming under US sanctions against China would be a disaster for them.
That could be fun (Score:3)
As Apple hates Nvidia so much they put inferior graphic chips in their hardware just because the aren’t Nvidia.
Re: (Score:2)
Or Apple just does a hostile takeover of nVidia and kills two birds with one stone.
Re: (Score:3)
Ahh yes, I look forward to my future GPU: All white with a recessed low profile fan in a magnesium shroud, that thermally throttles every time you load up Photoshop, has a single Thunderbolt port on the back, comes with a dongle pack that weighs more than the GPU to allow you to use "legacy" DisplayPort and HDMI connections, and comes with a proprietary PCI-e rear bracket which only supports cases with glass panels and vertical GPU orientations, because otherwise how will anyone see the Apple logo.
Apple and RiscV (and Ars Technica) (Score:1)
Ars Technica has an article about this also. They specifically mention Apple not being particularly interested and that, for many licensees of the technology, it would be a regulation nightmare to buy Arm. Nvidia is a licensee also, but apparently not so much in direct competition with other licensees, if I understood the article right.
I can't help but think about RiscV though. I wonder if the current owner, Softbank Group, a big Japanese holding company, sees Arm's value as peaking right now, with Risc
Re: (Score:2)
I wondered about Softbank's timing too and it did seem to follow Apple's announcement (confirmation) that they would be transitioning MacOS to ARM.
Yes, I understand that there are regulation hurdles so Apple may not be in a position to acquire or even interested in doing so.
However Apple's big reveal has put the spot light on ARM being a viable desktop/laptop alternative to the x86 monopoly.
Nvidia isn't a bad match from the perspective of Apple and other ARM licensees as they sell chips not devices.
Qualcomm
Re: (Score:2)
Re: (Score:2)
Could boost RISC-V (Score:3)
Nvidia doesn't know how to play nice with others and if they acquired ARM then it could easily be the promotion of RISC-V as an attractive alternative to ARM for low-end cores. ARM has some great stuff but they are hardly a unicorn in the CPU market.
Re: (Score:2)
Re: (Score:2)
I just do not see why RISC-V would be considered a real competitor in this market at all?
Because you don't have to pay licensing fees.
Re: (Score:2)
Re: (Score:2)
Not paying the licensing for an ISA that has no commercially available cores isn't really a selling point.
You're wrong about that. https://www.anandtech.com/show... [anandtech.com]
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Ok, you are conflating cores and ICs. Cores are what ARM provides to chip designers. ARM doesn't make any ICs themselves, thus they are "fabless". Chip designers license various cores from ARM and then add the peripherals to the chip. The chip designers then have their designs fabricated into ICs which they can either use or sell.
Using a RISC-V core would mean chip designers no longer have to pay ARM a licensing fee for using an ARM core in their design. As a result they can make a greater profit per c
Re: (Score:2)
I was pointing out that the lack of commercial licensing wasn't stopping them from using RISC-V.
there are plenty of free cores: https://github.com/riscv/riscv... [github.com]
Re: (Score:2)
Re: (Score:2)
Most companies that use ARM are not browsing ARM core designs and then getting them produced,
That's fine. I'm talking about the ones that design and sell them.
As I wrote in my other posts:
Using a RISC-V core would mean chip designers no longer have to pay ARM a licensing fee for using an ARM core in their design. As a result they can make a greater profit per chip or sell more chips by offering them at a lower price.
This is important because it will effect the companies that are looking to design products that need SoCs.
I also gave you a link to RISC-V SoCs that you can buy currently: https://gith [github.com]
Re: (Score:2)
If you are interested in SoC ICs that have RISC-V cores, here's a list: https://github.com/riscv/riscv... [github.com]
Re: (Score:2)
Not paying the licensing for an ISA that has no commercially available cores isn't really a selling point.
so... these cores don't exist? https://github.com/riscv/riscv... [github.com]
Re: (Score:2)
Re: (Score:2)
As far as I know, nobody even produces a RISC-V core for sale currently?
prepare to be dazzled because they are free: https://github.com/riscv/riscv... [github.com]
Re: (Score:2)
You underestimate the amount of investment everyone has in ARM. A lot of companies have put a lot of effort into optimizing their ARM based designs and switching now would be expensive, and that's before you even consider the software side of things.
Nvidia would really have to screw them hard to make it worthwhile.
Re: (Score:2)
Re: (Score:2)
Well... when Microsoft bought Mojang, I figured that was the end of the line for Bukkit and other such extensible server implementations. But they've changed absolutely nothing about that, and the only obviously Microsoftian elements are related to the way they port it to everything and use it to garner goodwill as a popular "face". They've left the golden goose, which is the Java version, to continue pretty much on the trajectory it had before the takeover. So it is possible. It just takes a company that h
Re: (Score:1)
Find me this sequence of words in an actual book. I'll wait...
Sadly, it will poison the well... (Score:1)
Nvidia loves lock-in tech.
They will find a way to screw everyone, this would be a catastrophe to ARM...but perhaps a warning to the market and invest in a more open cpu, like Risc-V or POWER.
This is the worst news (Score:2)
If this were to go through, it would bury the microprocessor market. Arm is in so many chips. Eventually many industries would move to RISC-V
Re: (Score:2)
Re: (Score:1)
Wow... (Score:2)
And here I was for years thinking nVidia would wrangle an x86-64 license from AMD.
I'm so wrong.
Not a bad idea. They used to make a great SoC (Score:2)
Not a bad idea. They used to make a great x86-compatible SoC. I own one of the netbooks that uses it and, for the period, its processor and graphics performance were superior to competing chipsets.