Poll

Is the Nvidia acquisition of ARM good or bad for the industry?

Good
9 (10.5%)
Bad
49 (57%)
Neutral
28 (32.6%)

Total Members Voted: 86

Author Topic: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?  (Read 10633 times)

0 Members and 1 Guest are viewing this topic.

Offline brucehoult

  • Super Contributor
  • ***
  • Posts: 3019
  • Country: nz
  • Formerly SiFive, Samsung R&D
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #75 on: September 20, 2020, 01:46:35 am »
They have allowed it, but you had to pay a fortune for an architecture licence to be allowed to change it. Only a few companies ever bought those. I imagine there are some modified ARMs out there in custom chips we just don't know about.
No, that's not correct. An ARM architecture license allows the company to design their own CPU cores, They are *not* allowed to modify the licensed ARM instruction set in any way -- nothing removed, nothing added. ARM executives are on record saying this. For example in this panel discussion (which I was at):
About 15 years ago I proposed making changes to the status register of the ARM7TDMI, a bit before the M series cores started appearing. Most people in our company immediately said we couldn't. Others said to me quietly that if you really read the architecture licence properly ARM could not stop us. Some key people inside the company really liked the idea of not using ARM, and I never pushed the point, so I don't know how it might have worked out.

I know and love the ARM7TDMI -- that was the dominant chip in mobile phones from every manufacturer back when I was writing a Java native compiler for BREW phones back just before iPhone and Android hit.

The status register might be a grey area if you're not adding or removing instructions.

What kind of change did you propose? Adding an Intel-style Half carry bit or Parity bit something like that? Back then the 32 bit status register had 20 undefined/reserved bits. So yeah you could add something like Half carry or Parity without technically changing the instruction set because instructions that set them would still be the same instruction. You wouldn't be able to add new instructions to branch on those bits directly, but you could always copy the status register to a GPR and then mask it with TST or shift that bit to the MSB and BMI etc. Which would be a lot lower overhead than not having those status bits if you needed them.
 

Online daqq

  • Super Contributor
  • ***
  • Posts: 2193
  • Country: sk
    • My site
Believe it or not, pointy haired people do exist!
+++Divide By Cucumber Error. Please Reinstall Universe And Reboot +++
 

Offline SiliconWizard

  • Super Contributor
  • ***
  • Posts: 9825
  • Country: fr
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #77 on: December 06, 2021, 05:37:58 pm »
Now that gets interesting.
 

Offline brucehoult

  • Super Contributor
  • ***
  • Posts: 3019
  • Country: nz
  • Formerly SiFive, Samsung R&D
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #78 on: December 06, 2021, 08:43:31 pm »
Now that gets interesting.

It's always seemed like a really hard acquisition to get past the regulators in multiple countries.

Meanwhile there have been 15 months and counting of ARM customers having an incentive to investigate alternatives, just in case. More than that when you count nervousness about Softbank increasing license fees to get their $30 billion investment back.
 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 5815
  • Country: nl
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #79 on: December 10, 2021, 04:12:15 am »
The only companies which should be happy with the acquisition failing are the handful of ones using ARM for housekeeping while competing in areas NVIDIA makes products (cars and supercomputers) and the companies which think they can do faster custom cores in house without ARM help (Qualcomm with the NUVIA, reportedly Samsung is building a new team too, I imagine Amazon thinks they can just DIY it too).

For small licensees this is a disaster, the regulators got this ass backwards.
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 6875
  • Country: gb
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #80 on: December 10, 2021, 05:50:06 am »
Now that gets interesting.

It's always seemed like a really hard acquisition to get past the regulators in multiple countries.

Meanwhile there have been 15 months and counting of ARM customers having an incentive to investigate alternatives, just in case. More than that when you count nervousness about Softbank increasing license fees to get their $30 billion investment back.
I assume it has been a great time for you.
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 6875
  • Country: gb
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #81 on: December 10, 2021, 05:52:34 am »
The only companies which should be happy with the acquisition failing are the handful of ones using ARM for housekeeping while competing in areas NVIDIA makes products (cars and supercomputers) and the companies which think they can do faster custom cores in house without ARM help (Qualcomm with the NUVIA, reportedly Samsung is building a new team too, I imagine Amazon thinks they can just DIY it too).

For small licensees this is a disaster, the regulators got this ass backwards.
That needs some explanation. If the deal went through ARM would soon be dead, and of no value to the small licensees. Sure nVidia have promised to keep an open licensing model, but the industry knows what promise like that are worth long term.
 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 5815
  • Country: nl
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #82 on: December 10, 2021, 06:43:06 am »
If the deal went through ARM would soon be dead, and of no value to the small licensees.

Fantasy, there is absolutely no profit in that for NVIDIA. Killing ARM will accomplish almost nothing in their major markets ... 40 billion down the drain for nothing.

The biggest markets affected where they directly compete would be supercomputers ... the performance of housekeeping cores in Cerebras/Fugaku is irrelevant, switching ISA for new generation products would be inconvenient but that's all it would be.
« Last Edit: December 10, 2021, 06:45:07 am by Marco »
 

Online westfw

  • Super Contributor
  • ***
  • Posts: 3804
  • Country: us
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #83 on: December 10, 2021, 07:41:59 am »
Quote
The only companies which should be happy with the acquisition failing
I was/am afraid that NVIDIA has zero interest in the embedded ARM cores (ie Cortex-M), and that they'd go the same way as Intel's ARM core processors...
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 6875
  • Country: gb
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #84 on: December 12, 2021, 04:43:33 pm »
If the deal went through ARM would soon be dead, and of no value to the small licensees.

Fantasy, there is absolutely no profit in that for NVIDIA. Killing ARM will accomplish almost nothing in their major markets ... 40 billion down the drain for nothing.

The biggest markets affected where they directly compete would be supercomputers ... the performance of housekeeping cores in Cerebras/Fugaku is irrelevant, switching ISA for new generation products would be inconvenient but that's all it would be.
This has nothing to do with any actions nVidia might take. Everyone will just run away when a silicon vendor is in charge of ARM. The only reason ARM grew into its current position is because it was not tied to a silicon vendor. When DEC, TI, Intel and others were trying to get ARM into the industry's foreground they all understood this, and none of them wanted to take control.
 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 5815
  • Country: nl
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #85 on: December 12, 2021, 05:20:53 pm »
When they tried ARM was failing to compete as anything but a microcontroller and if ARM doesn't find a way to make silicon macro development more profitable,  that's the situation where they will end back in.

Softbank's investment made it seem that providing relatively performant core macros was profitable, but it really wasn't. NVIDIA profits from having their GPUs in everything, let them and most everyone else can profit from high performant CPU cores. It's not ideal, but ARM IPO'ing and being forced to scale back R&D isn't ideal either.
« Last Edit: December 12, 2021, 05:25:18 pm by Marco »
 

Offline coppice

  • Super Contributor
  • ***
  • Posts: 6875
  • Country: gb
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #86 on: December 12, 2021, 06:02:13 pm »
When they tried ARM was failing to compete as anything but a microcontroller and if ARM doesn't find a way to make silicon macro development more profitable,  that's the situation where they will end back in.

Softbank's investment made it seem that providing relatively performant core macros was profitable, but it really wasn't. NVIDIA profits from having their GPUs in everything, let them and most everyone else can profit from high performant CPU cores. It's not ideal, but ARM IPO'ing and being forced to scale back R&D isn't ideal either.
One of the key things that drove ARM's success was reasonable fees. Anyone seeing the ARM instruction set as a gravy train ends up very disappointed. Things like RISC-V have made this more true than ever.

 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 5815
  • Country: nl
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #87 on: December 12, 2021, 06:42:59 pm »
Post Softbank the reason to use ARM macro cores was also their relatively good performance, that helped companies like Mediatek lot. The extra revenue stream NVIDIA gains from becoming the standard in graphics, ANN and compute across multiple platforms can sustain that level of investment more reliably than an IPO.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 14633
  • Country: us
  • DavidH
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #88 on: December 12, 2021, 08:26:54 pm »
If the deal went through ARM would soon be dead, and of no value to the small licensees.

Fantasy, there is absolutely no profit in that for NVIDIA. Killing ARM will accomplish almost nothing in their major markets ... 40 billion down the drain for nothing.

NVidia would not be able to resist seeking rents by tying their own IP to ARM when licensing, and discontinuing licensing of ARM IP which their competitors are using.  They were not buying ARM for the profits, but to cut off competitors whether it damaged customers or not.
 
 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 5815
  • Country: nl
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #89 on: December 12, 2021, 11:01:25 pm »
NVidia would not be able to resist seeking rents by tying their own IP to ARM when licensing, and discontinuing licensing of ARM IP which their competitors are using.

Governments can simply require them to extend existing licenses. Development would likely stop even if NVIDIA is forced to spin off Mali, but the flavour of GPUs in SoC's isn't all that important. Licensees which use Mali will switch and NVIDIA will profit mostly from the ecosystem effects of having a semi-standard for graphics, compute and ANN accelerators across multiple markets.
 

Online nfmax

  • Super Contributor
  • ***
  • Posts: 1482
  • Country: gb
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #90 on: February 08, 2022, 03:42:49 pm »
And the sale is abandoned: https://www.theguardian.com/business/2022/feb/08/nvidia-takeover-arm-collapses-softbank

Anyone up for subscribing to the IPO next year?  ;)
 

Offline Bassman59

  • Super Contributor
  • ***
  • Posts: 2491
  • Country: us
  • Yes, I do this for a living
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #91 on: February 08, 2022, 05:14:16 pm »
And the sale is abandoned: https://www.theguardian.com/business/2022/feb/08/nvidia-takeover-arm-collapses-softbank

Anyone up for subscribing to the IPO next year?  ;)

NVDA is pretty much flat today, so obviously "The Market" expected this, as the stock had been trending down for a couple of months. (Though that's the general market direction too.)

What I found odd was that the British government claimed "ARM is British! British jobs!" etc etc, but that apparently wasn't important when SoftBank bought it.
 

Offline Sal Ammoniac

  • Super Contributor
  • ***
  • Posts: 1532
  • Country: us
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #92 on: February 08, 2022, 05:44:02 pm »
And the sale is abandoned: https://www.theguardian.com/business/2022/feb/08/nvidia-takeover-arm-collapses-softbank

Anyone up for subscribing to the IPO next year?  ;)

NVDA is pretty much flat today, so obviously "The Market" expected this, as the stock had been trending down for a couple of months. (Though that's the general market direction too.)

What I found odd was that the British government claimed "ARM is British! British jobs!" etc etc, but that apparently wasn't important when SoftBank bought it.

Yeah, I did find that odd, especially since Nvidia is a U.S. company. I wonder how they would have reacted if a Chinese company tried to buy ARM?
Complexity is the number-one enemy of high-quality code.
 

Offline Marco

  • Super Contributor
  • ***
  • Posts: 5815
  • Country: nl
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #93 on: February 08, 2022, 06:04:35 pm »
Anyone up for subscribing to the IPO next year?  ;)

Maybe to dump it almost immediately, but that's a hell of a gamble on sentiment. The fundamentals are awful.

Long term just buy Apple instead for a stock to hold, this was better news for Apple than Arm.
 

Offline Bassman59

  • Super Contributor
  • ***
  • Posts: 2491
  • Country: us
  • Yes, I do this for a living
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #94 on: February 09, 2022, 03:16:59 pm »
And the sale is abandoned: https://www.theguardian.com/business/2022/feb/08/nvidia-takeover-arm-collapses-softbank

Anyone up for subscribing to the IPO next year?  ;)

NVDA is pretty much flat today, so obviously "The Market" expected this, as the stock had been trending down for a couple of months. (Though that's the general market direction too.)

What I found odd was that the British government claimed "ARM is British! British jobs!" etc etc, but that apparently wasn't important when SoftBank bought it.

Yeah, I did find that odd, especially since Nvidia is a U.S. company. I wonder how they would have reacted if a Chinese company tried to buy ARM?

Instead of the Japanese company that already owns it.
 

Offline mac.6

  • Regular Contributor
  • *
  • Posts: 210
  • Country: fr
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #95 on: February 15, 2022, 09:25:44 pm »
It's probably a good thing for the industry that Nvidia gives up, but for ARM it will be a blood bath (or at least for ARM employees).
The breakup clause money will go directly into softbank pockets, and ARM results are not good enough for a good IPO, so it will mean a lot of cost cutting measures...
 

Offline ejeffrey

  • Super Contributor
  • ***
  • Posts: 2936
  • Country: us
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #96 on: February 16, 2022, 06:02:50 am »
I don't think anyone sees RISC/V as a magic bullet. It doesn't offer spectacular speed, or superior low power performance. Its purely a route to save the per device fees that go to ARM right now.
but that is a mirage, as the ARM MCUs are currently way cheaper than any RISC-V, so the 'fees claims' fail to make it to to end user.
Critical mass and economies of scale matter more than licensing fees.

It matters a lot for people putting cores into other chips.  As far as I understand the arm licensing fees are a percentage of the die cost.  If you make a small microcontroller with an M4 core on a 45 nanometer process you pay very little.  If you make a 1 Tb/s network switch on 7 nm and want to put the same M4 core on it to load configuration registers you pay a lot more.  I definitely know people who are using arm cores in situations like this where they would prefer to go to risc v. 
 

Offline brucehoult

  • Super Contributor
  • ***
  • Posts: 3019
  • Country: nz
  • Formerly SiFive, Samsung R&D
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #97 on: February 16, 2022, 06:49:07 am »
It matters a lot for people putting cores into other chips.  As far as I understand the arm licensing fees are a percentage of the die cost.  If you make a small microcontroller with an M4 core on a 45 nanometer process you pay very little.  If you make a 1 Tb/s network switch on 7 nm and want to put the same M4 core on it to load configuration registers you pay a lot more.

One suspects such contract terms are going to be renegiotiated or modified in the next couple of years.
 

Offline SiliconWizard

  • Super Contributor
  • ***
  • Posts: 9825
  • Country: fr
Re: Nvidia Acquiring ARM for $40 Billion -- Good or Bad for the Industry?
« Reply #98 on: February 16, 2022, 05:34:46 pm »
It's probably a good thing for the industry that Nvidia gives up, but for ARM it will be a blood bath (or at least for ARM employees).
The breakup clause money will go directly into softbank pockets, and ARM results are not good enough for a good IPO, so it will mean a lot of cost cutting measures...

Because acquisitions are not usually immediately followed by restructuring and cost reduction measures? So either way...
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf