Nvidia has approached Softbank and is considering buying ARM Holdings

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

uzzi38

Platinum Member
Oct 16, 2019
2,565
5,575
146
Will update with articles as they come out (they haven't yet but news is public). For now just a couple of Tweets:



This is a real significant move for Nvidia, having a huge role in ARM core designs could have major implementations down the line. Most importantly, it secures them a platform and ecosystem down the line.

Oh and I guess Nvidia becomes the defacto standard for GPU IP for mobile instead of Mali. That too.

EDIT: Bloomberg article here: https://www.bloomberg.com/news/arti...ompany-arm-is-said-to-attract-nvidia-interest
 
Last edited:

Meghan54

Lifer
Oct 18, 2009
11,504
5,028
136
Maybe seeing that mess on the horizon is why Softbank decided they might want to unload it.

Maybe part of it, but a large part of all this is Softbank's recent rather poor decisions by Softbank on investments, such as Softbank's large investment in WeWork, which isn't exactly playing, or paying, out as expected. Softbank has been doing not so well in their investment decisions, so to recoup some losses, makes sense to sell something.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
@Meghan54

Oof I forgot about SoftBank investing in WeWork. What a money pit. Now ARM Holdings is enduring a mess that's probably knocking down their market value. Not a good time for them!
 

soresu

Platinum Member
Dec 19, 2014
2,617
1,812
136
My theory is that Nvidia wants to sell their GPU IP to mobile vendors. They tried this back in 2013, but as far as I can tell nobody took them up on it:
In all probability partly for the reason that neither Sony or MS want anything to do with them any more.

Only time well tell if even Nintendo sticks with them post Switch after that TX1 hardware security bug disaster.

I think another reason is that nVidia were overly confident in their likely higher pricing and did not even try to competitively bid against the other players in mobile - much as with MS and their heavy handed approach to devs which seems to result in less quality exclusives for Xbox.

Something to also bear in mind, despite being a very powerful GPU to be paired with an ARM CPU core in 2015, the TX1 drew a lot of power at full whack, so they certainly didn't have any big advantage over the likes of Adreno even after Kepler/Maxwell reduced the size of their cores/SMX's.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
On a related note then, is it possible that Nintendo could use a Samsung produced Arm core/AMD gpu Soc for the next switch? Such a beast is already being produced for use in Phones...
 

blckgrffn

Diamond Member
May 1, 2003
9,111
3,029
136
www.teamjuchems.com
On a related note then, is it possible that Nintendo could use a Samsung produced Arm core/AMD gpu Soc for the next switch? Such a beast is already being produced for use in Phones...

This seems to be the predominant rumor, but who knows?

It seems like it would increase port opportunities if they at least were on RDNA based GPUs, and a more modern ARM core on a more modern process (8nm?) could likely hit the same power envelope and still could easily double the CPU performance as well, probably putting it at another mid generation step above but closer to the One S and PS4 than the SX and PS5, much like the WiiU was (in many respects) a mild step up from the 360/PS3.

It would be assumed that nvidia would at least be chasing that business but who knows? As discussed in another thread, nvidia has practically come to a full stop on the ARM based platforms.

To some degree that ties back into my earlier comment - if nvidia doesn't have a full stack platform aren't they just going to get frozen out of the market at some point?

As others have noted, they haven't seemed to play well with others in the past.
 

Doug S

Platinum Member
Feb 8, 2020
2,203
3,405
136
It would be assumed that nvidia would at least be chasing that business but who knows? As discussed in another thread, nvidia has practically come to a full stop on the ARM based platforms.

To some degree that ties back into my earlier comment - if nvidia doesn't have a full stack platform aren't they just going to get frozen out of the market at some point?


Sure, but why did Nvidia come to a full stop in ARM based platforms? Due to their much lower volumes, they were the first Android OEM to conclude they couldn't justify the cost of designing custom ARM cores, since then Qualcomm and most recently Samsung have reached the same conclusion (though they had the volumes, their decision was more because ARM designed cores were just good enough that Qualcomm and Samsung couldn't beat them by enough to matter)

With all custom design efforts ended, ARM designed cores is now par for the course in the Android SoC world. If you want to differentiate in that market you need to do it in other ways. If you are Qualcomm by having the best in class modem. If you are Samsung, by being the first to throw new features against the wall and see what sticks. If you are HiSilicon by being Chinese. If you are Mediatek, by being cheap. If you are NVidia, presumably by offering the best in class GPU.

Nvidia's real problem back when they exited the market, and especially now, is the lack of a modem. By the time they could get anything out the door 5G will be a requirement, but who is going to sell them a 5G modem? Without that, having the best GPU or even owning ARM doesn't help them in by far the most lucrative market for ARM designs.
 

soresu

Platinum Member
Dec 19, 2014
2,617
1,812
136
though they had the volumes, their decision was more because ARM designed cores were just good enough that Qualcomm and Samsung couldn't beat them by enough to matter
More like they were already beaten and pursuing a custom core just for the sake of differentiation was not worth the R&D investment when they could likely see ARM's internal core roadmap for at least 2-4 years into the future.

Samsung's Mongoose cores were barely ahead in raw perf but significantly behind in perf/watt - it just wasn't remotely logical to continue down that path in mobile form factors where battery life really counts.

Qualcomm Kryo though - I would say that Apple caught them completely off guard with the performance tier of A7 onwards and they knew that Kryo needed at least a significant redesign, and as with Samsung if they knew a 20-25% increase per year was coming after A73 it was going to be good enough for them given that their customers do not directly compete with Apple Axx performance anyway.
 
  • Like
Reactions: Tlh97

ThatBuzzkiller

Golden Member
Nov 14, 2014
1,120
260
136
With all custom design efforts ended, ARM designed cores is now par for the course in the Android SoC world. If you want to differentiate in that market you need to do it in other ways. If you are Qualcomm by having the best in class modem. If you are Samsung, by being the first to throw new features against the wall and see what sticks. If you are HiSilicon by being Chinese. If you are Mediatek, by being cheap. If you are NVidia, presumably by offering the best in class GPU.

Huawei/HiSilicon's main advantage is their vertical integration. In a lot of ways they emulate AT&T before it's break up. Both produced radio systems and the cellular devices themselves. Carriers could solely source their supplies from Hauwei and they'd have a complete end-to-end system. Huawei's handsets themselves don't have any special value but their radio equipment is highly valuable among many carriers that enable these mobile network services ...

Nvidia's real problem back when they exited the market, and especially now, is the lack of a modem. By the time they could get anything out the door 5G will be a requirement, but who is going to sell them a 5G modem? Without that, having the best GPU or even owning ARM doesn't help them in by far the most lucrative market for ARM designs.

They could always get the 5G modems from a small player like Mediatek but the main problem then becomes the supported number radio frequency bands since it could still limit their market potential. A 5G solution from Mediatek might not necessarily have as good of a coverage compared to a Qualcomm solution in American markets because they lack some of the necessary carrier supported frequency bands ...
 
  • Like
Reactions: prtskg

mxnerd

Diamond Member
Jul 6, 2007
6,799
1,100
126
ARM Co-Founder: Sale to Nvidia Would Be a Disaster:

“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.”

 

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
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

But would they really? Due to the software ecosystem NV lock-in in Deeplearning is also a fact and people aren't switching away from it because it's simply to painful.
 

mxnerd

Diamond Member
Jul 6, 2007
6,799
1,100
126
But would they really? Due to the software ecosystem NV lock-in in Deeplearning is also a fact and people aren't switching away from it because it's simply to painful.
Yeah, it will be hard to switch. Maybe the best they can do is trying to block the deal.

In the past, it's just smartphone chip makers (Samsung, Qualcomm, or MediaTek,) & smartphone companies using ARM processors, now even Amazon, Apple, AMD, etc. are using the ARM chips with their own designs.
 
Last edited:

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
Yeah, it will be hard to switch. Maybe the best they can do is trying to block the deal.

Fully agree. Much easier and cheaper to block the deal or buy ARM yourself / in a joint venture.

I'm sure if NV buys ARM alone, the others would start projects to move away. But that will be costly and will take >5 years. So yeah blocking or joint venture will be far cheaper.
 

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,686
136
But would they really? Due to the software ecosystem NV lock-in in Deeplearning is also a fact and people aren't switching away from it because it's simply to painful.
And what exactly would be the alternative for someone like Qualcomm? Just accept the fact that they'll slowly die?

Imagine AMD licensing Ampere from Nvidia in 2020. The same would apply for CPU cores for Nvidia vs. Qualcomm and others. They would get last gen tech at best, scraps more likely.

Much easier and cheaper to block the deal or buy ARM yourself / in a joint venture.
This indeed.
 
  • Like
Reactions: Tlh97

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,619
136
But would they really? Due to the software ecosystem NV lock-in in Deeplearning is also a fact and people aren't switching away from it because it's simply to painful.
Two things about that: Only now this lock-in is really paying off for Nvidia, which tells me up to very recently the lock-in has boosted the use of ASICs instead at bigger companies and held back adoption of DL at smaller companies. And all exascale supercomputers avoid the Nv lock-in, which should pay off for the non-Nv software ecosystem and everybody trying to avoid the lock-in.
 
  • Like
Reactions: Tlh97

soresu

Platinum Member
Dec 19, 2014
2,617
1,812
136
Deeplearning is also a fact and people aren't switching away from it because it's simply to painful.
ARM have greatly increased perf on ML workloads with their CPU cores since they focused on that.

However most mobile ML tasks are destined to be run on specialised inference accelerators or a combo of that and DSP in the case of Snapdragon (Hexagon uArch).

These specialised are much more power efficient at ML workloads vs even the current state of the art ARM core, and likely even Matterhorn (supposedly 10x perf previous gen through MatMul).

The sheer unassailable mass of ARM apps written for each of the major mobile platforms is the real reason that no one is switching - but that doesn't mean that nVidia can't find a way to make their former fellow licensees miserable by cutting off supply to new CPU core designs and forcing them to start fresh.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136

Dribble

Platinum Member
Aug 9, 2005
2,076
611
136
So, nobody in those talks speaks sense, just money. Got it.

Dark times ahead if nV ends up owning ARM.
It could be worse, it could be Apple.

Samsung and Qualcomm are at least as bad as Nvidia. Then there is the firms that would just buy it to break it up and sell it on.

Really I don't see any of the prospective buyers being any good.
 

Roland00Address

Platinum Member
Dec 17, 2008
2,196
260
126
It could be worse, it could be Apple.

Not really, for Apple has nothing they want out of it. Since Apple controls the software stack, and you can buy an indefinite license to ARM version number if you make your own chips the only advantage of Apple owning this is future ARM features that they did not buy the license for. The thing is that works both ways where Apple can always add new stuff that is not part of the official ARM architecture.

Apple does not need ARM like at all (more than they already have as a current relationship) and you can't really buy ARM and extort Apple for the already existing licenses apple has already bought an indefinite contract for.

And Apple owning ARM and raising prices for everyone else will be seen as anti-competitive and will definitely invite anti-trust. There are just some ways you can't make money in this world, it is not completely "free" and that is a good thing.

Samsung and Qualcomm are at least as bad as Nvidia. Then there is the firms that would just buy it to break it up and sell it on.

Really I don't see any of the prospective buyers being any good.

A group of owners who would all have minority share buying ARM would as a good thing for ARM is good for the industry by having a unified instruction set and basic designs that anyone can make. Likewise companies like Qualcomm and Samsung could innovate on top of this. Then again in the last 7 or so years Qualcomm and Samsung have not been innovating for the base ARM designs for the big cores have been so good compared to what Qualcomm and Samsung have designed.

Samsung has designed faster stuff but it is not energy efficient taking way too much power for this faster stuff. (Also not die efficient but that is a lesser concern.) Thus Samsung has been buying 3rd party chips for more than half the phones they ship.

Likewise Qualcomm used to make superior chips compared to ARM prior to 64 bit (v8 instruction set), but for the last few years Qualcomm has been taking the default ARM cores and just swapping a few parts preferring the energy efficient ways of building the chip to the point you can say they are almost generic ARM cores but not literally ARM generic cores.

Shrug due to how the licensing works with ARM there is no obvious great buyer for it. (This is a good thing for the industry for it prevents predatory behavior where people pull a Vader and change the nature of past agreements.)
 
  • Like
Reactions: Tlh97

moinmoin

Diamond Member
Jun 1, 2017
4,934
7,619
136
Pretty prescient of Apple to have started talking about "Apple Silicon" at this time. Currently it refers to processors with custom Arm cores, but if the development indeed continues like some media report Apple may be interested in slowly taking the "Arm" part out of the equation over the long haul instead moving to newer Arm versions.
 

Doug S

Platinum Member
Feb 8, 2020
2,203
3,405
136
Pretty prescient of Apple to have started talking about "Apple Silicon" at this time. Currently it refers to processors with custom Arm cores, but if the development indeed continues like some media report Apple may be interested in slowly taking the "Arm" part out of the equation over the long haul instead moving to newer Arm versions.

Why? The only thing they'd lose would be access to future versions of the ARM ISA their existing architecture license doesn't cover (and for all we know, it specifically covers "all future versions" if Apple's lawyers thought ahead to cover their butt in case of a sale of ARM)

If they were stuck on ARMv8.6 or whatever they'd be fine for a long long time, and wouldn't be missing out on much - most of the future extensions to ARM are likely to be aligned more for servers. It already has everything it needs (and more) for phones/PCs.

I think the "Apple Silicon" branding is because Apple is designing both the CPU and GPU, along with the NPU, IPU, etc. It mostly is "Apple" silicon, even if the ISA was defined by ARM. This is IMHO a strategy to differentiate from current Macs ("Intel inside") to future Macs. Of course you'd only want to call that out if they were better, which is another reason I think their performance will beat that of x86 Macs across the board. They want people to know that they are responsible for that, not ARM.

Moving to another architecture might free them from the modest ARM licensing cost, but they'd gain huge new costs having to support a new software platform all by themselves. They'd need a really good reason to consider leaving ARM, and there's no way they haven't protected their trillion plus dollar business from a rogue future owner of ARM Ltd.
 
  • Like
Reactions: Tlh97 and scannall

Dribble

Platinum Member
Aug 9, 2005
2,076
611
136
Not really, for Apple has nothing they want out of it.
A lot of patents. Full control of the design of their cpu's which are based on ARM and they are planning to use for everything and are key to the success of this 2 trillion dollar company. The cost is also insignificant to a company the size of Apple.
 

Doug S

Platinum Member
Feb 8, 2020
2,203
3,405
136
A lot of patents. Full control of the design of their cpu's which are based on ARM and they are planning to use for everything and are key to the success of this 2 trillion dollar company. The cost is also insignificant to a company the size of Apple.

Apple already has full control over the design of their CPUs. The only control ARM has is adding new instructions, and I'm sure Apple carries some influence for that. Apple has enough CPU patents that they can fight off any legitimate lawsuit (and patent libraries don't defend against lawsuits from NPEs/patent trolls)

Just because Apple can easily afford it doesn't mean paying $30 billion or whatever for patents and the ability to dictate additions to the ISA is a good idea. You don't reach a market cap of $2 trillion by being foolish with money. You could use the same argument that Apple "could easily afford" buying say Goodyear or PetSmart, which would be about as useful to them as owning ARM.