News Apple pays Arm less than 30 cents per chip in royalties, new report says

[Admin]

Apple pays Arm less than 30 cents per chip in royalties, according to a new report — adding up to less than 5% of Arm's sales.

Apple pays Arm less than 30 cents per chip in royalties, new report says : Read more

 

[JamesJones44]

Apple actually already does use RISC-V for some of their IC chip designs. I wouldn't put it past Apple to switch to RISC-V more heavily if they felt terms were unfavorable given they are only using the instruction set and not chip designs.

 

[ekio]

Yes and that’s why ARM stock price is very overrated.

Better bet on risc-v that will not crumble if they don’t get multiple bilions to just run idle…

 

[Bluetooth!]

Apple is not using ARM’s chips, it’s only using the ISA. Also ARM has used some of the new ISA commands that Apple has made. Thus, the price is not that low and not compareable with those usimg ARM’s chip designs.

 

[JTWrenn]

I say put the screws to them now rather than later. They don't have anything they can jump to fast enough to not lose money so you have a shot now...if contract is coming up. Wait another 5 to 10 and you may have risc-v or some other options up and running and end up in a bind.

This also really shows that Apple needs to be broken up. Their market power and capital give them too much power in negotiations and break competition by them getting every sweet heart deal. Anti trust is dead.

 

[TCA_ChinChin]

Apple actually already does use RISC-V for some of their IC chip designs. I wouldn't put it past Apple to switch to RISC-V more heavily if they felt terms were unfavorable given they are only using the instruction set and not chip designs.

Really? I didn't know they did that even for some of their less complex ICs. Thats cool to know.

 

[hecksagon]

In this case I can almost justify that. Apple only licenses the instruction set. The chip design itself is entirely Apple's.

 

[Bluetooth!]

In this case I can almost justify that. Apple only licenses the instruction set. The chip design itself is entirely Apple's.

Apple also did show with their own chip designs, that ARM was just as good as an ISA as X86. The attention ARM has gotten after the M chips came from Apple are probably wort a lot to ARM!

 

[thestryker]

I'd love to know what the royalties look like on the Qualcomm architecture license if Arm is willing to fight them in court while giving Apple this deal. My guess is that Arm is either trying to maintain the current royalties from Qualcomm using Arm designs instead of their own or Qualcomm's deal is even better than Apple's.

 

[Bluetooth!]

I'd love to know what the royalties look like on the Qualcomm architecture license if Arm is willing to fight them in court while giving Apple this deal. My guess is that Arm is either trying to maintain the current royalties from Qualcomm using Arm designs instead of their own or Qualcomm's deal is even better than Apple's.

The fight with Qualcomm is that they bought a startup with a deal with ARM, that QCom wants to keep, but ARM said was void after the takeover

 

[bit_user]

Apple has also reportedly looked into the open-source RISC-V architecture (which wouldn't require any royalty payments at all) — potentially to give it leverage with Arm.

@JarredWaltonGPU , sorry to bother you, but could you please let Andrew know it's an open standard, not open source? He frequently makes this mistake, and it only spreads confusion among readers.

Also, I think the article doesn't sufficiently distinguish between the architecture license, under which Apple designs its own cores vs. the full core implementations being used in the SoCs currently made by Qualcomm and MediaTek. I think this explains why ARM has been so desperate to renegotiate the terms of Qualcomm's architecture license, as it moves back to designing its own cores (i.e. via Nuvia).

If it's really true that the finished core designs only cost Qualcomm about twice as much as an architecture license would, I'm actually quite surprised! There's a world of difference in how much it costs ARM to design cores vs. just letting people use their patents - and that gets to the heart of the difference between license types.

 

[thestryker]

The fight with Qualcomm is that they bought a startup with a deal with ARM, that QCom wants to keep, but ARM said was void after the takeover

Nothing you said here is accurate except that Qualcomm purchased Nuvia.

 

[bit_user]

Apple actually already does use RISC-V for some of their IC chip designs.

There's a world of difference between tiny, RISC-V microcontrollers and the massively complex P & E cores primarily responsible for power Apple's devices! Beyond the relative simplicity, the microcontrollers aren't accessible to end users or 3rd party developers.

I wouldn't put it past Apple to switch to RISC-V more heavily if they felt terms were unfavorable given they are only using the instruction set and not chip designs.

I would. It wouldn't happen any time soon (i.e. this decade). It took them many years and huge amounts of resources to move their developers and customers onto the same platform as their phones already use. You're talking about repeating that whole ordeal, only way worse, because now it'd extend to their phones & watches, too!

I think neither they nor their customers or developers have the appetite for it. There's no way customers wouldn't be put off by having to switch architectures again, too.

Really? I didn't know they did that even for some of their less complex ICs. Thats cool to know.

So does Nvidia, but it's immaterial. Their GPUs have embedded microcontrollers. They used to license ARM M cores for those, but swapped in their own RISC-V designs about 5 years ago. Even so, Nvidia is firmly in the ARM camp, for the externally-programmable parts of their products and even went on to try and buy the company!

 

[bit_user]

I'd love to know what the royalties look like on the Qualcomm architecture license if Arm is willing to fight them in court while giving Apple this deal.

The difference is that ARM wants Qualcomm's customers to pay royalties for using chips containing Qualcomm-designed ARM cores. In Apple's case, this wouldn't matter because the only devices they go into are also made by Apple.

The fight with Qualcomm is that they bought a startup with a deal with ARM, that QCom wants to keep, but ARM said was void after the takeover

No, not exactly. Nuvia's license is non-transferrable, and there's apparently no debate about that. It's not uncommon for a contract to have a clause saying that the contract becomes null & void if the licensee is acquired. This is a common safety measure to keep big companies from buying up small ones as a way of acquiring more favorable licenses than the big one would be able to negotiate on its own. Another reason is that the acquiring company could be a direct competitor that you wouldn't want to have a license under any circumstances.

The debate is that Qualcomm still had an architecture license from when it was doing its own core design. Qualcomm claims that Nuvia can continue to use that pre-existing Qualcomm license, but ARM says it can't. The new one ARM wants Qualcomm to use instead is way worse for Qualcomm and its customers.

My guess is that Arm is either trying to maintain the current royalties from Qualcomm using Arm designs instead of their own or Qualcomm's deal is even better than Apple's.

Agreed. ARM needed to find some way to improve its financial performance, relative to other design firms like AMD and Nvidia.

They apparently have a "plan B", which is to offer their cores on chiplets, potentially cutting out the "middle man" of SoC integrators like Qualcomm and MediaTek. This lets end users pair an ARM chiplet with their own custom (or 3rd party) IP, although I don't know if that's a very viable option for smartphone makers or just something aimed mostly at other markets.

 

[JarredWaltonGPU]

@JarredWaltonGPU , sorry to bother you, but could you please let Andrew know it's an open standard, not open source? He frequently makes this mistake, and it only spreads confusion among readers.

Also, I think the article doesn't sufficiently distinguish between the architecture license, under which Apple designs its own cores vs. the full implementations being used in the SoCs currently made by Qualcomm and MediaTek. I think this explains why ARM has been so desperate to renegotiate the terms of Qualcomm's architecture license, as it moves back to designing its own cores (i.e. via Nuvia).

If it's really true that the finished core designs only cost Qualcomm about twice as much as an architecture license would, I'm actually quite surprised! There's a world of difference in how much it costs ARM to design cores vs. just letting people use their patents - and that gets to the heart of the difference between license types.

Your open-source complaint has been fixed. As for Qualcomm... doesn't it do its own Snapdragon architecture, or Kryo or whatever? (I don't follow smartphones that much.) MediaTek does license the Arm cores, though.

 

[thestryker]

Your open-source complaint has been fixed. As for Qualcomm... doesn't it do its own Snapdragon architecture, or Kryo or whatever? (I don't follow smartphones that much.) MediaTek does license the Arm cores, though.

Qualcomm used to, but they switched over to using Arm designs after I think the Snapdragon 820/821. The first Kryo was in house then they just co-opted the name for the future SoCs.

 

[bit_user]

Your open-source complaint has been fixed.

Thanks!

As for Qualcomm... doesn't it do its own Snapdragon architecture, or Kryo or whatever? (I don't follow smartphones that much.) MediaTek does license the Arm cores, though.

Ugh, Kryo is such a troublesome thing. Qualcomm uses its own names for the cores as a way to obfuscate what they are and stymie comparisons with SoCs from other manufacturers.

Once upon a time, Kryo cores were 100% in-house Qualcomm designs. The last such phone SoC with custom cores was the generation with Snapdragon 820 as its flagship (and refined in the 821, circa 2016). This was their first attempt at 64-bit cores. Qualcomm seemed to have been running behind with those cores, because the Snapdragon 810 had standard ARM Cortex-A57 + A53 cores.

"The 820 was pivotal for another reason too: It introduced Qualcomm’s first custom 64-bit CPU core, Kryo. Creating a custom CPU (or GPU/DSP/ISP) is one way for SoC vendors to differentiate their products and establish themselves as innovators. Snapdragon 810’s use of stock ARM cores could be construed as a step backwards then after previous Snapdragon SoCs used Qualcomm’s custom Krait CPUs. Apple’s prior introduction of a custom 64-bit CPU, which caught everyone by surprise, only added fuel to the fire."

Source: https://www.anandtech.com/show/10948/qualcomm-snapdragon-835-kryo-280-adreno-540
​

For anyone interested in the details, here's an in-depth analysis of those cores:

Kryo: Qualcomm’s Last In-House Mobile Core

CPU design is hard.

chipsandcheese.com

In the conclusion, they quip:

"To frustrate anyone looking for information on the original Kryo, Qualcomm kept using the Kryo name even though future chips no longer used in-house CPU designs."
​

Starting with the Snapdragon 835 (circa 2017), Qualcomm used lightly-tweaked ARM Cortex designs:

"The Snapdragon 835's Kryo 280 performance and efficiency cores make use of ARM's Built on ARM Cortex Technology license which allows Qualcomm to make requests to ARM to change some aspects of the architectures of newly released cores and implement these changes exclusively into Snapdragon SoCs. As such the S835's CPU cores were derivatives of ARM's Cortex A73 and Cortex A53 CPU IPs."

Source: https://www.anandtech.com/show/12114/qualcomm-announces-snapdragon-845-soc
​

Down road, Qualcomm used this partnership with ARM to create the Cortex-X1, which was apparently so successful that ARM decided to continue the Cortex-X series on its own.

Here's a decoder ring, for working out which Cortex cores a given Kryo number is based on:

Kryo - Qualcomm - WikiChip

Kryo is a series of semi-custom ARM cores from Arm integrated by Qualcomm in their Snapdragon SoCs.

en.wikichip.org

Unfortunately, that stops at Snapdragon 865 (circa 2019). Any cores beyond that, you can find listed here:

How to understand Kryo CPU numbering in Qualcomm Snapdragon processors

Qualcomm's Snapdragon processors use Kryo CPUs. These CPUs have their own number system, separate to the Snapdragon numbering. Confused? Don't be.

www.androidauthority.com

 

[JamesJones44]

I would. It wouldn't happen any time soon (i.e. this decade). It took them many years and huge amounts of resources to move their developers and customers onto the same platform as their phones already use. You're talking about repeating that whole ordeal, only way worse, because now it'd extend to their phones & watches, too!

I think neither they nor their customers or developers have the appetite for it. There's no way customers wouldn't be put off by having to switch architectures again, too.

While I agree it would take time, worrying about consumer and developer backlash is a non-issue. Apple already did this with x86/x64 for Mac with almost zero backlash. In fact from a consumer point of view it was a big win.

Selling RISC-V over ARM might not come with the same benefits as moving on from Intel/x86/x64 but I don't think Apple will have a problem spinning the message.

 

[George³]

I don't know why Apple is exalted in this case. They haven't been able to design even one modem for so many years and that's why they continue to use modems from Qualcomm. Are you also sure about what you are sharing here about the architecture of their SoCs?

 

[thestryker]

I don't know why Apple is exalted in this case. They haven't been able to design even one modem for so many years and that's why they continue to use modems from Qualcomm. Are you also sure about what you are sharing here about the architecture of their SoCs?

While I have no love for Apple building a modem isn't as simple as one might think. There's decades worth of IP to license and/or build around and Qualcomm controls most of the recent IP applicable to 5G. I'm quite certain they could shove a modem out the door quickly if they weren't trying to circumvent Qualcomm.

A long way of saying making a modem has basically nothing in common with making a SoC.

 

[bit_user]

While I agree it would take time, worrying about consumer and developer backlash is a non-issue. Apple already did this with x86/x64 for Mac with almost zero backlash.

There was about 15 years between each of their transitions from 68k -> PPC -> x86 -> ARM. A big reason there wasn't more backlash is that it was telegraphed long enough in advance, that most people knew they were coming and could defer upgrades until after the transitions and developers had ample time to prepare. If Apple suddenly turned on a dime and switched to RISC-V now, I'm sure it would bother a lot of users and developers as they're finally getting settled on ARM.

Worse, it would make Apple look like they don't know what they're doing. Furthermore, the RISC-V software ecosystem isn't nearly as mature as ARM was, when Apple started to move in that direction.

Finally, as I pointed out, they've never switched their device ISA - only their computers.

In fact from a consumer point of view it was a big win.

I doubt it. Having to run stuff in emulation usually isn't. I'll grant that x86 was a win, due to the ability to run Windows apps natively. Plus PPC was stagnating, so x86 was a way to get more performance. I also happen to think it was good that they switched to little endian, too.

The main upsides to ARM are that it's more efficient (i.e. runs cooler; better battery life) and that it unifies their computers with their phone platform. Also, it happened at a time when Intel was stagnating, much like IBM did, previously. Switching to RISC-V has no such advantage over ARM, so there really would be no practical upsides to it - just headaches.

I don't think Apple will have a problem spinning the message.

Spin only gets you so far.

 

[bit_user]

I don't know why Apple is exalted in this case. They haven't been able to design even one modem for so many years and that's why they continue to use modems from Qualcomm. Are you also sure about what you are sharing here about the architecture of their SoCs?

Modems are very different than CPU cores. The woes of their modem effort have been well-documented, by now.

 

[Dementoss]

Apple pays Arm less than 30 cents per chip in royalties, according to a new report — adding up to less than 5% of Arm's sales.

Apple pays Arm less than 30 cents per chip in royalties, new report says : Read more

Apple and ARM negotiated a deal, so what? Their problem, if it is one at all, not mine. If it wasn't good enough, ARM's shareholders would have kicked up a fuss.

 

[JamesJones44]

There was about 15 years between each of their transitions from 68k -> PPC -> x86 -> ARM. A big reason there wasn't more backlash is that it was telegraphed long enough in advance, that most people knew they were coming and could defer upgrades until after the transitions and developers had ample time to prepare. If Apple suddenly turned on a dime and switched to RISC-V now, I'm sure it would bother a lot of users and developers as they're finally getting settled on ARM.

I didn't say NOW, no where did I say NOWW. Why would they even bother now if the have an agreement till 2040? The point is, if they started to feel ARM was going to try to shake them down for higher royalties I have no doubt they would start to move on from ARM. Hell even if the started NOW it would take 3 to 5 years just in development tooling and rebuilding their own software, by then it would have been 8 years since flipping to ARM on Mac.

Worse, it would make Apple look like they don't know what they're doing. Furthermore, the RISC-V software ecosystem isn't nearly as mature as ARM was, when Apple started to move in that direction.

How exactly? If they built out their tools just like they did for the x86 to ARM transition there would be very little impact from a developer point of view. In the majority of cases the transition from the developer point of view was completely transparent. Simply download the latest version of Xcode and recompile.

Apple could simple start with one device like the iPhone SE or the Apple Watch. Starting there wouldn't require them to update runtimes like Java, .NET, etc. or LLVM backends for languages like GO and Rust. Apple did this for years with ARM based phones and x86 based Macs. For years you've been able compile Swift/Objective-c codebase for both targets (ARM on iOS devices and x86 for Mac devices).

At any rate, you are trying to argue about an assumption, the assumption that I'm saying they would flip the switch in 6 to 12 months. That's not what I'm saying AT ALL. I'm saying they would start to look at transitioning, that doesn't mean shock the eco system with one massive change.

 

[JamesJones44]

I doubt it. Having to run stuff in emulation usually isn't. I'll grant that x86 was a win, due to the ability to run Windows apps natively. Plus PPC was stagnating, so x86 was a way to get more performance. I also happen to think it was good that they switched to little endian, too.

What stuff in emulation? I haven't had to run Rosetta in two years using Mac, everything has been transitioned. Java has an ARM runtime, .NET has an ARM runtime, Node has an ARM runtime. Go, C++, Rust, Kotlin Native/Multi project/Runtime, Python all have LLVM backends for ARM. Even crappy projects like Electron have ARM based versions now.

So unless you are running something that hasn't been updated in 3 years there is no reason for emulation/translation on the Mac these days.