Nvidia abandons deal to buy ARM *spawn*

Qualcomm has been using Arm cores for several gens now instead of their own. Nuvias core will the first custom one for 'em in quite a while

 

Don't quote me to explain any difference, but Qualcomm's Kryo are at least according to QCOM (semi) custom CPUs based on ARM CPU IP. The SD888 carries the KRYO 680 CPU. Before Kryo they used to have Krait and in the dark ages Scorpion.

***edit: in order to avoid any misunderstandings; IMHO QCOM SoCs are the best in the entire Android "world". And while I would support the possibility of the NUVIA team to create a quite good CPU solution, there's from one side more to hw then just the CPU and from the other side good hw with mediocre sw cannot succeed. The target for the NUVIA team development is clearly low end laptops and there good and stable windows drivers are going to be essential. Any of their future attempts against any possible NVIDIA attempt (assuming with ARM in the latter's portofolio) isn't going to be easy both on a hw as well as on sw level. NV doesn't need to do much, as their driver support and experience is already guaranteed.

 

Nobody is buying an iPhone, iPad or Mac because of the instruction set architecture or CPU design. Apple don't even reveal the clock speeds in their devices, which they market as a complete product: hardware, OS, applications, services and sofware platform. Nothing Apple does is in direct competition with ARM or which Nvidia.

If Nvidia want to make ARM better, with the possible consequence that Apple spent less time as much proprietary work, that sounds like a win.

 

Apple could save a lot of money not buying out the two most advanced nodes and making fuck huge ICs.

 

Apple's unique approach to any of the above you mention, doesn't make it sell its devices in a vacuum. A high end Apple smartphone competes in real time against a high end Samsung smartphone as an example. If tomorrow Samsung manages for any reason to sell 3x times as many high end smartphones as today, then Apple will have a very serious problem because the pool of potential high end smartphone consumers isn't unlimited.

NV obviously intends with the ARM acquisition to yield both for itself as for ARM a specific growth margin for both their future. NV has resources and experience to move ARM IP into the desktop which other Android players don't have.

 

What you're describing is a competitive market, which is what Apple exist in. There containing to be a competitive market, or an even more competitive markets are not a cause for concern under these types of regulatory consultations.

These is not about whether company A likes what combat B might want to do, these needs to be a demonstrable negative consequence of the acquisition to the wider market. If one player is disadvantaged but the market itself becomes more competitive with more options, that is good and not grounds for complaint. It is unfair competition that is frowned upon and only that will impact the final decision.

 

If a legal system allows N company or companies to file a complaint they will do so. The FTC has presented its points against the acquisition and NVIDIA has recently replied to it. Personally I wouldn't lay my hand into fire that any semiconductor that has filed a complaint so far, doesn't have any reason to worry with NVIDIA acquiring ARM. If there's any kind of legal way or mean to secure all of NVIDIA's claims/promises then I can't find any serious reason to oppose the acquisition exactly because IMHO ARM will most likely have a quite questionable future if they remain independent.

When the mobile market exploded selling IP to large semiconductor manufacturers was the alpha and omega for success for it. Nowadays where those semis increasingly rely only on architectural licenses such as Apple for example both ARM's income and future growth comes with a question mark. Even more so if any of the big players in the market move completely away from any sort of ARM IP and design their own solutions from scratch.

There a viable alternative is to move to higher end markets exactly as Apple did with the M1 and so far we haven't seen anything viable yet from anyone.

 

Perhaps in countries with a culture where frivolous lawsuits are encouraged and/or successful, but the UK and the EU are not territories where regulators tolerate this.

Again, the goal of this process is not a commercial popularity content but to make a determination about whether an acquisition or merger will negatively impact the market. If a party (e.g. Apple) does not like it but has not legitimate complaint and can not compellingly evidence concerns how the market would be impacted, which might involve revealing commercially sensitive future plans, then they are simply wasting their time.

To reiterate, just because somebody might be bad (or good) for Apple, does not mean that the same can be said of the larger market.

 

Which has what to do exactly with the conversation at hand? ARM is the only company that resides in the UK, but is owned by Japanese Softbank currently and is proposed to be acquired by NVIDIA which is a US company recently scrutinized for it by the US FTC. While it's also being besides the point the British government allowed ARM being sold to Softbank in the first place, despite ARM being the largest profitable technology powerhouse in the UK at the time.

If someone would have the urge to patent stupidity his chances would be higher in the US than in the UK, but that's also besides the point.

Apple isn't the only semi which could be interested in said acquisition, nor is Apple the be all end all in the mobile market. The list of possible semi manufacturers that might have filed a complaint against the acquisition is long and they are legally entitled to do so and here it's quite irrelevant if it's Intel, Qualcomm, TI, Renesas or anyone else. Again as I said if NVIDIA keeps what it promises in its answer to the FTC then there's nothing in my mind that would speak against it.

Mediatek at least is IMHO not on that list as it seems in favor of the acquisition.

 

Proving bad faith for a bad argument beyond a reasonable doubt is almost impossible and unless they are making a deposition under threat of perjury the only penalty for outright lying would be diplomatic, not legal.

No need to lie though, weasel words and selective truth suffices.

 

Because the CMA (the UK regulator) is one of the regulatory authorities that need to approve the acquisition. I'm not re-treading the same things that were said earlier in the thread - just scroll back.

Providing false information to an investigation into an acquisition or merger is absolutely a statutory offence under Article 117 of the Enterprise Act 2002. The reason that EU and UK market regulation consultations tend to be less dramatic than some other markets, is because both dole out fairly harsh punishments for peddling bullshit and there is no "reasonable doubt" provision. I think you've been watching too many court dramas. There are other potential consequences as well such as being disqualified from submitting evidence in other investigations.

I'm not a lawyer and neither of you so leave it at just reading what the law says.

 

They used to do full custom like Krait and Scorpion and whatnot, but that's ancient history. They've been using Cortex-derivatives with minimal customization (cache stuff mostly iirc) since Snapdragon 835 (and 810 was just bog standard Cortex's, 820 was custom though)
https://www.anandtech.com/show/10948/qualcomm-snapdragon-835-kryo-280-adreno-540/2

 

Israel is a hotbed for microprocessor design and research facilities for other high-tech companies. Intel setup shop in the 1970s and IBM, Apple and Microsoft have big R&D facilities there.

 

Don't they have their own scheduler?

 

Not really a surprise, especially the location.

 

My point was/is that a semi like QCOM has the resources to go custom at any time they wish. Alas if whatever they're cooking now with the NUVIA team onboard isn't a full custom approach. The question would rather be whether the result can only be used for low end laptops or if it's scalable all the way down to ULP mobile SoCs.

I think yes, but I'll gladly stand corrected.

 

Bweh, finally the clown show is over.
Gotta hope this spurs another round of custom core cooking championship since mainline Cortex clapped previous efforts pretty roughly.