Nvidia abandons deal to buy ARM *spawn*

The FTC sues quite a bit for competitive reasons, so if there is any good news it is that the FTC will need to prove their case in court before a "neutral" judge.
Not sure why their anti-competition/monopoly suits against Facebook keep getting thrown out of court.

 

I sorta thought it would never happen and now am conflicted. Amd has graphics and CPU, Intel is attempting gpu. Apple has hardware software and chips now. It seems only fair to let Nvidia get stupidly big as well, but they have to stop the madness of endless mergers and so forth at some point. "Welcome to Costco I love you"

 

This isn't really comparable. NV are free to use the ARM architecture however they want to as long as they have a license. Just like Apple. They are also able to modify and extend the capabilities of whatever they license, however they wish to. Again, just like Apple. Nothing prevents NV from making an Apple M1 style chip using ARM + NV IP if they wanted to.

This isn't like AMD and Intel where x86 is basically locked to those 2 companies and good luck getting them (mostly Intel) to allow you to use x86 however you wish.

Regards,
SB

 

But that's kind of a double standard, isn't it? nvidia has to license a cpu architecture or isa because they make gpus, but intel and amd can make gpus and tightly control x86 (if it's not just intel doing that exclusively). Not that I think Nvidia owning arm would be good.

If anything it seems like intel should be forced to break off part of it's business into an arm-like license holder, seperate from the actual intel cpu business. That would make things "fair", so to speak. You have x86 as a business that licenses designs and the isa, and Intel that designs, manufactures and sells cpus and gpus to the market.

 

AMD is effectively licensing x86 from Intel so I fail to see the difference in that aspect. Sure Intel doesn't have to license x86, however, there are things they do have to license, like x64 (64 bit extensions to x86) from AMD.

Regards,
SB

 

Nothing prevents NV from doing that as well. So this obviously wouldn't hurt NV in any way shape or form.

And while Apple, Qualcomm and Samsung could do that, they also benefit from not having to spend the R&D required to improve the core capabilities like ARM does. It's cheaper for them to modify and extend the architechture than to do the R&D necessary to keep up with the likes of Intel. Well, Apple might be willing to do that, but unlikely that Qualcomm or Samsung would be remotely as successful compared to ARM in improving the overall core arch.

Regards,
SB

 

So then why can't Nvidia own arm and continue to license the isa or arcthitecture to a select few?

 

Qualcomm seems to be gambling on the team they got from Nuvia. They could very well think ARM being forced into an IPO and needing to do some serious cost cutting as a result, would hurt Mediatek and Samsung far more than them.

 

It probably has to do with whether relevant regulatory bodies believe that NV would withhold IP based off of the ARM purchase from anyone that NV views as a competitor.

ARM has traditionally licensed to anyone and everyone equally. That includes companies that could be viewed as competitors such as Intel and AMD.

Considering NV's recent history of becoming more closed versus more open as they were when they were starting out, that likely raises some red flags at regulatory commissions.

This is only speculation, but here's what I think.

• NV gains almost nothing from the purchase of ARM except perhaps more engineers.
  • Anything they can do by owning ARM they can do without owning ARM.
• So it comes down to NV wanting control over ARM IP.
  • Regulators then have to consider whether NV will continue to license ARM to anyone who wants a license.
    • This would include any customizations or extensions to the ARM architecture.
      • Basically would ALL ARM related IP continue to be freely available to license or would NV decide to make some ARM related IP proprietary to them and them only in order to gain a competitive advantage?
      • This is likely why Apple wasn't interested in making an offer for ARM.
    • This would also involve looking at past NV practices and how they have handled their in house IP.
      • Have they been open with licensing their IP to other companies?
• This is important because so many large corporations rely upon the ability to license technology from ARM which puts them at an even footing competitively with other licensors of ARM IP.
  • Even footing WRT the licensed IP and not WRT to any modifications or extensions they do after they have licensed the IP.
  • Since NV unlike current owner Softbank would want a competitive advantage from the this aquisition, can it be trusted to license ALL ARM related IP to anyone who wishes to license it?
    • Basically would NV license any improvements to the ARM architecture like ARM/Softbank has done?
    • Or would they keep any improvements they made to the ARM arch for themselves and only continue to license ARM IP prior to their acquisition of ARM?
      • This is perfectly fine if you are licensing the IP from ARM.
      • However, regulators would likely see this as being non-compatible with ARMs position in the market.

So, why is it OK for Intel to not license their tech to everyone? Because they've never licensed their tech to everyone and thus entire industries aren't reliant on licensing their CPU tech.

Basically think of it like this. If Intel were up for sale, would it greatly impact other companies if the purchasor of Intel decided not to license the CPU related IP much less ALL the CPU related IP? Well, it would impact AMD. Thus any potential purchase of Intel would have regulators trying to determine whether the purchasor would continue to license to AMD or not. Additionally, how would this affect other industries that rely on AMD products that contain IP licensed from Intel? The other consideration would be whether the purchasor of Intel would continue to sell CPUs to anyone that wanted to purchase their CPUs.

Regards,
SB

 

If ARM IPO's they are just going to become a small ISA licensing firm, with the big customers deciding ISA evolution by committee (which in a decade or less won't include Apple, they are going to ditch ARM regardless).

Due to lots of reasons the winner take all aspects of the economy are increasing. Trying to fight that for ARM won't make the small players more competitive, it will just drag ARM down with them.

PS. I don't see trust busting increasing significantly, you could find enough support to break up Microsoft perhaps ... but Apple, no way.

 

IMHO the majority of the ecosystem will survive one way or the other, because there are and will be alternative solutions.

 

Not sure about what you mean with Apple offering "more than Arm offers". If you mean just the custom cores, there's plenty of companies which have successfully done just that for years with each company at some point going back to Arm cores at least for now (seeing Arm spread to more markets it might be time for Qualcomm for example to re-evaluate if they should go back to custom cores).

 

As MfA noted above, Apple could eventually move away from ARM. Considering how far they've improved CPU efficiency based on an ARM based architectural license, it shouldn't be too hard to figure that they can if they want go for any possible alternative.
Yes it takes time, but there's already moving quite a bit from different corners for RISC-V designs and while we don't know how matters will evolve around ARM, there are definitely signs that a variety of companies are seeking for alternative solutions.

 

If Apple were to move away from ARM why on Earth would they necessarily move to RISC-V, or any other externally-derived alternative? They have more money than God and can afford to design their own proprietary ISA & architecture from scratch, and given the way they tend to work why would they not do that?

It would be interesting though. Just like old times.

 

For the same reason they haven't done it so far. "Any possible alternative" by the way INCLUDES the possibility of a complete in house design from scratch.

 

Yup. It's the same reason why genuinely new instruction set architectures aren't introduced more often. The simple truth is they are hellishly complicated and when your goal is performance/watt/cost, starting from scratch means developing the hardware and the entire dev tool toolchain - and right now Apple lean heavily on LLVM which helps it stay competitive. Using a common architecture and common compiler technology (LLVM) means you benefit from general advances derived from everybody sharing the same technology versus being some technical isolationist.

Taking a proven technological and architectural base and adapting the broad strokes to your needs is a lot of less risky than tossing it all over and starting with a blank canvas.

Apple may have a lot of money but even with money sometimes you just cannot buy the expertise you would need. Some people aren't motivated my money.

 

Apple sells hardware and software together which give them a unique position. Their walled garden also helps encourage their users to stay with them and they have a following that is extremely loyal so they can just tell them that dongles are great or backward compatibility is not needed and they go along with it. Nvidia would have to make something really competitive and then convince a separate company to do something with it and split the profits. I don't see them as the same. But you are correct that Nvidia could sink a bunch of resources into it and lose a lot of money in the process.