Nvidia CPU's? Nvidia is looking at a takeover bid for ARM

[Rroff]

ARM generally hasn't needed to hit as high speeds to compete against x86 - although that doesn't hold true across the board - hence why many implementations of RISC systems have dedicated co-processors or add-in accelerators for certain tasks like media.

 

[HACO]

If x86 was "designed with top speeds in mind" why are there loads of useless instructions like Loop or Rotate and costly prefixes consuming instruction bytes and increasing decoding complexity? Getting rid of them would make the front end faster and make programs smaller.

It’s because it’s not designed with top speed in mind. It’s designed with top compatibility in mind. x86 is all about compatibility above all else. Speed and efficiency are both secondary.

ARM generally hasn't needed to hit as high speeds to compete against x86 - although that doesn't hold true across the board - hence why many implementations of RISC systems have dedicated co-processors or add-in accelerators for certain tasks like media.

x86 processors have the same co-processors and accelerators (some of which are actually ARM, like AMD’s PSP). But then again, they’re also RISC like every modern processor.

 

[HACO]

Looks like Samsung is interested to buy a stake in ARM:

https://www.koreatimes.co.kr/www/tech/2020/08/133_293819.html

 

[drunkenmaster]

If x86 was "designed with top speeds in mind" why are there loads of useless instructions like Loop or Rotate and costly prefixes consuming instruction bytes and increasing decoding complexity? Getting rid of them would make the front end faster and make programs smaller.

Again I didn't say x86 the architecture, though old instruction sets within x86 mean of course there is slower versions of crap in it. What I said was the general chips being made by Intel and AMD, the only really relevant x86 people, have been making high speed optimised chips for decades.

 

[drunkenmaster]

We're obviously both speculating on intentions of Nvidia, you saying they'll just keep ARM as-is to diversify and I'll say they want to do something big with it. The rest is really off-topic (would be happy to have that discussion though in the appropriate topic).

I mean that isn't what I said at all. WHat I said was they currently have access to all of ARMs IP because that's how ARM operates. You don't need to spend 55billion when you can buy a license to the full IP they operate for <10million + tiny royalties per chip. In fact it's rather crazy to say that 55bililon so they can create a custom high speed desktop ARM chip is the reason they got it because they can literally achieve the same by buying the licence and putting together a team of engineers to design a custom core for a teeny tiny fraction of that pricing.

What I also said was I can't even tell what their entire goal is but that a large part of it, or most of it would be to diversify away from only selling graphics cards. Which doesn't preclude making higher and higher speed versions of ARM, I just said I don't think that is the primary goal nor would it be achieved successful in a short period of time like less than 3 years.

 

[HACO]

I mean that isn't what I said at all. WHat I said was they currently have access to all of ARMs IP because that's how ARM operates. You don't need to spend 55billion when you can buy a license to the full IP they operate for <10million + tiny royalties per chip. In fact it's rather crazy to say that 55bililon so they can create a custom high speed desktop ARM chip is the reason they got it because they can literally achieve the same by buying the licence and putting together a team of engineers to design a custom core for a teeny tiny fraction of that pricing.

What I also said was I can't even tell what their entire goal is but that a large part of it, or most of it would be to diversify away from only selling graphics cards. Which doesn't preclude making higher and higher speed versions of ARM, I just said I don't think that is the primary goal nor would it be achieved successful in a short period of time like less than 3 years.

The assumption that “buying a license is all they need” is really our disagreement here.

Nvidia doesn’t have the skills to compete with Intel and AMD in the CPU space at this moment.
ARM, at this moment, doesn’t have anything Nvidia can license to compete with Intel and AMD. They however know how to make CPUs a lot better than Nvidia does.
ARM’s priorities, for the time being, are not to go all-in against Intel and AMD.

If Nvidia wants to be in the CPU space (otherwise why would they be interested in ARM), a merger/acquisition with ARM makes sense. ARM has the skills and talent to compete with Intel and AMD if their priorities were different. Nvidia can set those priorities after that acquisition.

Your whole “they can license it and it will be identical” point which you keep repeating is irrelevant. Nvidia can’t license a product that ARM doesn’t make. They can however instruct ARM to make it if they own it. They can also do what Apple did, buying smaller promising chip designers and attracting talent from everywhere to build up their own team, it will be more expensive and will take longer, but works. Apple had probably invested more than $55 billion in their chip design though.

And they won’t spend $55 billion out of their balance sheet for the acquisition themselves, all deals of this magnitude will be mergers. This merger makes no sense if it ends up diluting Nvidia by $55 billion to just earn $1 billion in profit a year (ARM’s profits last year). It must be strategic and product-oriented otherwise Nvidia’s board will be crazy to accept it.

 

[HACO]

Looks like Samsung is interested to buy a stake in ARM:

https://www.koreatimes.co.kr/www/tech/2020/08/133_293819.html

Now denied by Samsung:

https://finance.yahoo.com/news/samsung-denies-interest-arm-holdings-023937318.html

 

[CAT-THE-FIFTH]

One of the ARM co-founders is saddened by what happened:
https://www.bbc.co.uk/news/technology-53637463

The co-founder of the company described as the jewel in the crown of British tech has said it would be disastrous for it to be sold to a US computing firm that is reportedly negotiating a takeover.

It's the second time in four years that the future of Cambridge-based chip-designer ARM Holdings has been uncertain.

In 2016, Softbank ended up buying it. But the Japanese firm is now reportedly in advanced talks to sell it to Nvidia.

Hermann Hauser told the BBC he thought the UK government should intervene.

The tech entrepreneur - who spun off ARM from Acorn Computers in 1990 - says ministers should help make it an independent UK business again.
'Unsuitable owner'

ARM creates computer chip designs that others then customise to their own ends. It also develops instruction sets, which define how software controls processors.

Just about every modern mobile phone and smart home gadget is powered by a chip that relies on one or both of these innovations.

When Softbank bought the firm for £24bn soon after the referendum to leave the EU in 2016, it was hailed by the government as a vote of confidence in a post-Brexit Britain.

But Dr Hauser said at the time it was a sad day for him and for UK technology.

He explained that ARM's business model - in which all the big chip-makers license its technologies - made Nvidia an unsuitable owner.

"It's one of the fundamental assumptions of the ARM business model that it can sell to everybody," he explained.

"The one saving grace about Softbank was that it wasn't a chip company, and retained ARM's neutrality.

"If it becomes part of Nvidia, most of the licensees are competitors of Nvidia, and will of course then look for an alternative to ARM."

While Dr Hauser voted against the Softbank deal in 2016, he says the Japanese firm kept its promises to retain Cambridge as the main focus of ARM's research and to boost employment there.

He has little faith in that remaining the case if Nvidia takes over.

Even the ARM co-founder said Nvidia would be a terrible owner,so yes a consortium would be a better choice. As he mentioned ARM has to stay neutral and be easily licensable,other as he said companies will look at other alternatives.

 

[Dalandius]

But it will eventually make ARM more niche. Apple has a ton of sway in the US,so not sure whether Nvidia can become more cocky with them. They have nearly $250 billion in free cash. Samsung could design its own cores,and is using RDNA on its latest SOCs. I also think about the Chinese companies such as HiSilicon,Rockchip and MediaTek,will also move to other uarchs,because even if ARM is kept as a "UK based subsidiary",Nvidia will can be lent on during any disputes.Qualcomm are the ones in the biggest bind here,but again have a lot of important cellular patents,etc.

Edit!!

Honestly,the UK government should have kept ARM independent instead of being swallowed up by another large company. In the end,I can see it slowly moving ebbing away,if Nvidia and others start mucking around with the existing model.

True except that cash is being held off shore rather than being in the USA.

 

[CAT-THE-FIFTH]

True except that cash is being held off shore rather than being in the USA.

ARM was created as a joint venture between Acorn Computers,Apple and VLSI back in 1990. So the thing is Apple,like some of the other companies can move over to other uarchs,ie,something such as RISC V might end up getting more prominence. ARM needs to stay neutral IMHO,and its neutrality is why it was widely licensed.

 

[HACO]

ARM was created as a joint venture between Acorn Computers,Apple and VLSI back in 1990. So the thing is Apple,like some of the other companies can move over to other uarchs,ie,something such as RISC V might end up getting more prominence. ARM needs to stay neutral IMHO,and its neutrality is why it was widely licensed.

Also adding that migrating away form ARM is a lot easier than migrating away from x86. If Apple or Fujitsu for example wanted to move away from ARM, they can continue to support hardware decoding of ARM ISA in their processors just as they do now, they would just add to it decoding of another ISA as well until they’re happy to drop the legacy bits a few years later, i.e. no software emulation necessary.

 

[LoadsaMoney]

TSMC and Foxconn jump in.

https://wccftech.com/arm-becomes-th...g-war-as-tsmc-and-foxconn-jump-into-the-fray/

 

[Orcvader]

TSMC and Foxconn jump in.

https://wccftech.com/arm-becomes-th...g-war-as-tsmc-and-foxconn-jump-into-the-fray/

I hope this is true, I can't see ARM being as open as it currently is with Nvidia or any other major fabless designer. TSMC is in a very good position at the moment so they're probably the best possible buyer to win against Nvidia's bid.

 

[HACO]

I hope this is true, I can't see ARM being as open as it currently is with Nvidia or any other major fabless designer. TSMC is in a very good position at the moment so they're probably the best possible buyer to win against Nvidia's bid.

Complete TSMC ownership of ARM is dangerous, because it turns TSMC into a new Intel. They can either ensure ARM's best and latest designs only work with TSMC, and/or they can keep their latest fab technology for their own ARM processors, leaving competitors (AMD, Apple, Nvidia, etc) to older fabrication.

Any single big tech corporation owning ARM is pretty dangerous. Nobody will pay $55 billions for a $1 billion a year profit. Anyone who buys it will want to leverage their position/tech to advantage themselves in a market in a way that they currently can't by just licensing the technology.

The only good outcomes for ARM are either a consortium ownership from major tech companies (which Nvidia, TSMC and others can be members), or becoming British again using a new IPO.

 

[Orcvader]

Complete TSMC ownership of ARM is dangerous, because it turns TSMC into a new Intel. They can either ensure ARM's best and latest designs only work with TSMC, and/or they can keep their latest fab technology for their own ARM processors, leaving competitors (AMD, Apple, Nvidia, etc) to older fabrication.

Any single big tech corporation owning ARM is pretty dangerous. Nobody will pay $55 billions for a $1 billion a year profit. Anyone who buys it will want to leverage their position/tech to advantage themselves in a market in a way that they currently can't by just licensing the technology.

The only good outcomes for ARM are either a consortium ownership from major tech companies (which Nvidia, TSMC and others can be members), or becoming British again using a new IPO.

Fair point, did not think about that, TSMC becoming the ARM version of Intel with x86 would actually be very dangerous.

Consortium does seem like the next best thing but seems unlikely if Softbank is just asking tech companies to buy the entire thing. Going independent would be very ideal but sadly seems to have the lowest chance of happening.

 

[Rroff]

TSMC interest could be a bit of push and shove with nVidia over them trying to strong arm them. TSMC I don't think has any interest in being a design company as well - their focus is enabling fabless design houses - as long as people are coming to them for production it is actually better for them if ARM isn't owned by themselves or a company that might seek to take a strangle hold of ARM.

Foxconn is a more curious one.

 

[HACO]

TSMC interest could be a bit of push and shove with nVidia over them trying to strong arm them. TSMC I don't think has any interest in being a design company as well - their focus is enabling fabless design houses - as long as people are coming to them for production it is actually better for them if ARM isn't owned by themselves or a company that might seek to take a strangle hold of ARM.

Foxconn is a more curious one.

TSMC denied any interest too.

 

[HACO]

EXCLUSIVE: US chipmaker Nvidia closing in on deal to buy Arm by the end of the summer

https://www.standard.co.uk/business/nividia-buy-chipmaker-arm-a4524761.html


Looks like it's happening. Nvidia looks more and more likely to buy ARM now. Time to push development of RISC-V, but that's at least half a decade or more away.

 

[jigger]

EXCLUSIVE: US chipmaker Nvidia closing in on deal to buy Arm by the end of the summer

https://www.standard.co.uk/business/nividia-buy-chipmaker-arm-a4524761.html


Looks like it's happening. Nvidia looks more and more likely to buy ARM now. Time to push development of RISC-V, but that's at least half a decade or more away.

Probably about the same amount time it will take to finalise the sale of ARM and rise the funding.

It’s a rather RISCy deal in the current environment.

 

[HACO]

Probably about the same amount time it will take to finalise the sale of ARM and rise the funding.

It’s a rather RISCy deal in the current environment.

It's a major acquisition/merger for Nvidia (by far the largest in their history), so yeah, pretty RISCy.