Nvidia abandons deal to buy ARM *spawn*

[pascal]

As long as Nvidia continues to be bold it could be a good thing!
Imagine vertical and horizontal consistent product lines.
From IoT to mobile, game box, to personal computing, to data centers and HPC.

This could be a Techno Reset!

Let's focus on Software Architecture. There is much to be done.
Develop the software once and run and scale it everywhere.
Today AI is far from general AI, data governance is a nightmare, security is weak, etc...

 

[Rootax]

Yeah, I kind of agree. Like them or not, nVidia engineers are pushing the envelop on the gpu side. I would love to see them in the arm cpu market (and I mean having full control, not just licensing IP like they did in the past).

 

[Entropy]

That already exists, it's called RISC-V.

How's that coming along?

Depends on who is asking, and their standards for success. It is very, very far from ARM in impact and eco-system, obviously. A major reason for that is that ARM, and the associated eco system, exist in its current form. If Nvidia buys ARM, you’ll see the strongest such as Apple, roll their own. Some will try to band together behind RISC-V. Some will try to come up with a new ISA, et cetera. It is likely to be a bit of a mess, and set the industry back overall.
If the story as it goes is correct, SoftBank first approached Apple, who immediately declined buying ARM on the grounds that a deal couldn't possibly go through on anti-trust grounds. SoftBank then widened the offering and the Nvidia deal came up. But as far as I can see the same anti-trust concerns apply.

 

[JasonLD]

I’d rather see the chinese government, or TSMC buy ARM.

That is far worse than Nvidia acquiring ARM, not even sure why you are suggesting that.

Or an industry consortium, but the politics of that would be tricky. But better.

I don't see Apple taking a part of that since they already have perpetual license and they do not like sharing their architecture detail with others. Also, many conflicting interests as far as how to push ARM architecture forward may prove to be more hinderance than beneficial.

 

[Sxotty]

Seems like this deal will never happen now due to international issues.

 

[ChuckeRearmed]

Seems like this deal will never happen now due to international issues.

Well, it depends on the regulators and arguments that the opposing forces will make. Despite being relatively inventive, Nvidia is notorious for it proprietary technologies.

China might block as ARM becoming american will lead to possible restriction of access due to american sanctions. So China will lose access to ARM licenses for example.

UK might also be involved (selling ARM to SoftBank was really stupid lol) again.

But they expected the deal to happen within 18 months which is almost two years.

 

[Kaotik]

Well, it depends on the regulators and arguments that the opposing forces will make. Despite being relatively inventive, Nvidia is notorious for it proprietary technologies.

China might block as ARM becoming american will lead to possible restriction of access due to american sanctions. So China will lose access to ARM licenses for example.

UK might also be involved (selling ARM to SoftBank was really stupid lol) again.

But they expected the deal to happen within 18 months which is almost two years.

Also the US:

Slaughter’s elevation signals a tougher approval process for vertical deals. Before taking over the agency, Slaughter criticized new guidelines issued last year by the FTC and the Justice Department outlining how the agencies would evaluate vertical deals. She said the guidelines overemphasize the potential benefits of such mergers and are “inexplicably mute” about the harms.

In December, Slaughter and Chopra said companies should no longer rely on the guidelines as an indication of how the FTC will police vertical deals.

“Moving forward, we need to aggressively enforce against the harms of vertical mergers,” they wrote. “We look forward to turning the page on the era of lax oversight and to beginning to investigate, analyze, and enforce the antitrust laws against vertical mergers with vigor.”

https://www.bloomberg.com/news/arti...oft-qualcomm-protest-nvidia-s-arm-acquisition

 

[Scott_Arm]

I feel like x86 needs to go away some time soon and we need something like arm to replace it. Not sure what gets us there. Nvidia buying ARM probably wouldn't be the best.

 

[iroboto]

I feel like x86 needs to go away some time soon and we need something like arm to replace it. Not sure what gets us there. Nvidia buying ARM probably wouldn't be the best.

are there any promising research papers on other architectures that could extract much more performance for the same amount of silicon use?

 

[Scott_Arm]

are there any promising research papers on other architectures that could extract much more performance for the same amount of silicon use?

Not sure. I’ve only recently been reading about arm. I don’t fully understand why it uses so little power other than just a general sense that it minimizes wasting transistors as a design philosophy. Also it doesn’t waste as many clock cycles. No idea if there are other contenders out there.

Edit: Read this the other day https://debugger.medium.com/why-is-apples-m1-chip-so-fast-3262b158cba2

 

[digitalwanderer]

Not sure. I’ve only recently been reading about arm.

No offense, but I found this sentence absolutely hysterical considering your name!

 

[DavidGraham]

Edit: Read this the other day https://debugger.medium.com/why-is-apples-m1-chip-so-fast-3262b158cba2

So just more decoders and and potent OoOE? I feel this is just a major over simplification.

I would really like to know more about the subject, the M1 is currently the king of single threaded performance, with just 3.2GHz clocks, it managed going toe to toe with 5GHz AMD and Intel CPUs.

 

[digitalwanderer]

Caveman Jim has really been in to ARM and RISC V stuff since leaving AMD, (James Prior). I've been following some of the stuff he's been involved with on Facebook and while I don't quite get it I am excited by the possibilities.

Or am I confusing ARM and RISC?

 

[Scott_Arm]

So just more decoders and and potent OoOE? I feel this is just a major over simplification.

I would really like to know more about the subject, the M1 is currently the king of single threaded performance, with just 3.2GHz clocks, it managed going toe to toe with 5GHz AMD and Intel CPUs.

Yah, I'm sure it's an oversimplification. The blog post is really meant for the layman, so I'm sure there's a lot of technical stuff that's missing.

 

[Scott_Arm]

Caveman Jim has really been in to ARM and RISC V stuff since leaving AMD, (James Prior). I've been following some of the stuff he's been involved with on Facebook and while I don't quite get it I am excited by the possibilities.

Or am I confusing ARM and RISC?

arm is an isa that follows a risc design philosophy. RISC-v is another isa that’s open source and designed around the same principles. I think that’s how it goes.

Edit: Apparently at the microcode level intel and amd CPUs operate like RISC designs. The just have complex decoders to turn complex x86 instructions into simpler microcode instructions. There are a lot of other things like compiler output, registers to look at.

 

[DegustatoR]

The just have complex decoders to turn complex x86 instructions into simpler microcode instructions.

Which is more or less the main part due to which an x86 CPU will always be inferior in perf/watt when compared to a pure RISC design on the same production process.

 

[PSman1700]

Dont worry about NV its apple that will take over everything slow but secure.

 

[Kaotik]

Dont worry about NV its apple that will take over everything slow but secure.

Apple has the exact same issues as NVIDIA does when it comes to who should own Arm.
Except phone soc manufacturers might triple their efforts to block it in case of Apple.

 

[pharma]

Most companies that oppose the merger can't expect Arm to operate perpetually with low margins to their benefit.

Behemoths like Google and Microsoft are opposed to the deal, which requires approval from authorities in at least four governments, including the U.S., U.K., E.U., and China. Qualcomm (NASDAQ:QCOM) is fighting the acquisition, as the company relies heavily on Arm for microprocessor intellectual property.

Apple also is concerned as Arm’s IP powers 90% of the world’s mobile processors/smartphones, with both Qualcomm and Apple as major customers. Should the acquisition be approved, it's easy to imagine Nvidia creeping into their territory.

Most of the naysayers, who have themselves been accused of antitrust and controlling the market as monopolies for decades, are concerned about whether they’ll have equal access to Arm technology after the acquisition closes.

These companies want Arm to be "Switzerland" — but this may be unrealistic as remaining neutral has resulted in thin profit from designs and licensing. What company will own Arm unless it strategically benefits the company’s core business?
...
In terms of market share of licensing and designs, Arm is incredibly successful. Yet, despite near omnipotence in mobile, televisions, and ADAS automotive systems, the top line is only $2 billion annually and the bottom line is $50 million in profit.

Why The Nvidia-Arm Acquisition Should Be Approved (NASDAQ:NVDA) | Seeking Alpha

 

[Kaotik]

Most companies that oppose the merger can't expect Arm to operate perpetually with low margins to their benefit.

Why The Nvidia-Arm Acquisition Should Be Approved (NASDAQ:NVDA) | Seeking Alpha

It's not about margins, it's about keeping it neutral for implementers, which can't be guaranteed when one implementer competing with the rest owns and develops it.