While nobody has exact details, from what I understand this is not $10-30M one off, but it's some kind of recurring licensing fee, also based on volumes, so could reduce profits on a long term basis....
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Arm Files Lawsuit Against Qualcomm and Nuvia for Breach of License Agreements
- Thread starter jdb8167
- Start date
- Sort by reaction score
There is no slowdown. Microsoft is not ready. The never had to convince all their customers to recompile everything for a new instruction set. And since they won't stop selling Windows on Intel, it might even be impossible to get Windows on ARM off the ground.
I doubt Qualcomm could even buy the cheapest commercial license for that price.
According to Anandtech in 2013:
The ARM Diaries, Part 1: How ARM’s Business Model Works
And prices seem to have gone up since then.
Exclusive: Arm raises prices on chip technology for some customers, sources say
SoftBank Group Corp-owned <9984.T> semiconductor technology supplier Arm Ltd has moved to raise licensing fees for some customers in recent negotiations, four people familiar with the matter told Reuters.
Windows on Arm impossible to get off the ground? Arm pragmatically kicked AMD out of the current generation Surface 9 Pro tablet PC. The notion that Windows on Arm is not "off the ground" unless completely clobbers Intel market share is grossly flawed. If Windows on Arm took 5-6% of the overall Windows PC market share it would be about as big as Mac are. Are the Macs not off the ground? No.
Right now one of the bigger liabilities that Windows on Arm has is pragmatically a higher price point more so than performance. Qualcomm makes the argument that their SoC with modem built in is a lower bill-of-materials (BOM) than Intel/AMD+ discrete modem solution. But that also is a relatively narrow niche of PCs. Qualcomm has to get past the point where using Windows on Arm to push radio sales.
All Windows on Arm has to been on is a growth progression. Microsoft is never going to tell all their customers to dump x86_64 and switch. There was a substantively large revolt when they told folks that they had to dump Intel Gen 7 (and AMD models) and 32-bit boot/driver to get to Windows 11 which is still in the x86 family. Let alone a whole new arch.
Microsoft spent a lot of time , effort , and treasure trying to get older quirky 32 x86 apps running on Arm. That was dubious. But it also the conservative , anti-change elements of their user base that drives that motivation. ( MS also had inertia from Windows Phone OS on Arm and WinRT playing a role. In some respects Microsoft is still paying "launched too early" price. ) Apple dumps legacy software on a regular basis. Probably doesn't loose sleep when folks with that heavy anti-change sentiment bolt over to Windows. Apple isn't trying to sell everything to everybody. Microsoft does. That is a major reason why they are so slow in transition, but also why they are steady with transition also (not likely to quit on a gradually evolving transition. )
Windows on Arm isn't going to lift off the ground like a lightly loaded fighter jet on afterburners. It is more so going to be like a fully loaded , max-sized military cargo plane ( C5 or An225 ). The missions are different.
P.S. Major missing price point price is partially uncorked by finally shipping a decent leading edge SoC in a developer kit
$599 32GB RAM ( no sizable Cellular radio mark up )
Buy Windows Dev Kit 2023 Desktop PC for Arm App Developers - Microsoft Store
Build, debug, and test native Windows apps for Arm with Windows Dev Kit 2023, a compact desktop computer engineered for developer workloads.
www.microsoft.com
If Qualcomm is charging MS for the radio these are being sold at a loss. ( Hence the developer only marketing. )
Yes. ARM chips account for nearly 10% of the entire PC market. 90% of those 10% are Macs. So all Chromebooks and Windows on ARM fight over less than 1% of the PC market. And in this situation Microsoft releases a Surface that's $1000 with an Intel chip and +$300 more with an ARM chip. This thing will never fly! Nobody will buy it.
Don't think Microsoft can afford for it not to work, x86 isn't catching up to Apple in perf/watt even if they can tread water in raw performance terms, and current x86 chips are about at the limit of just pumping in more power. I still think x86 will be all but gone from consumer computing by the end of this decade, Apple are going to keep pushing ahead and Microsoft and Google are stepping into chip development in their own right to keep up.
Profits for whom? Qualcomm or Arm. Currently Qualcomm is using licensed tech for their Phone/Tablet market SoCs. Those are a higher fee per unit than with an architecture license. If Qualcomm takes Nuvia core tech and dumps Tech licensed cores from the top end of their Phone SoC market the profit loss would likely impact Arm more than Qualcomm. Even if Nuvia Arch license unit fees were higher, it seem unlikely that it would be higher than a Arm Tech design license.
Qualcomm asserts that they have (presumably paid for) an architecture license that they really aren't using. How come that hasn't adversely hurt profit margins? Pretty clear how Arm could be hurt by someone stops double licensing (paying extra just in case want to jump back into doing it 'in house". )
There is pretty good chance that Nuvia's SoC took some uncore Arm stuff to speed up deployment. How much of that Qualcomm was going to deploy is open (could shed some fees if remove most of it).
As for Qualcomm profits taking a hit. Over paying $400-700M for Nuvia over a 10 year amortization is about (without NPV) $40-70M a year. If tried to stretch that to 15 years it is still $27-47M per year. Long term hit on profits was already in the bag when they overpaid. The SoCs have slipped to "late 2023" and no really good signs that won't slip into 2024. ( legal 'drama' , weakening PC market , somewhat long Mac upgrade cycles, etc. ) The more time Qualcomm gives Intel/AMD time to adjust, then the more likely the window to "fat margins" is going to narrow (competing in the WindowsPC space).
MediaTek is getting better. Google isn't using Qualcomm radios or SoCs. Apple is working toward dumping Qualcomm radios.
Decent chance that both companies are going to see lower margins. This lawsuit will turn into a game of musical chairs where lower margins are coming either way for both regardless of who 'wins' one round of taking away a chair in this legal contest. Both companies have "emperor's new clothes" problems of denial about lower margins coming.
A lot of nonsense. Apple may have 10% of USA market, but worldwide they aren't near 10%. ( the latest USA dollar adjustments worldwide to Mac prices only reinforces that ). 1% of the market means someone is buying it. Your are arguing a tautology that Arm systems sell ( only 0.2-0.5%) and not sell ( 0% ) at the same time.
Mac Pro's are likely less than 1% of the overall Mac unit volume. Are they "not off the ground" also? No. Just as not true for Windows on Arm. 'Impossible' doesn't have any rational basis as being an appropriate adjective here.
Windows on Arm isn't 100% totally dependent upon Qualcomm either. WoA instances on Azure are also sales of the OS.
The + $300 price point is mostly being driven by the cellular radio. 'Arm' really isn't playing a major role there. It is Qualcomm not really wanting to sell a radio-less SoC. That is a contributing factor to why the Chromebook Arm SoC outsell the Qualcomm SoC in unit volume.
For company work laptop that spends 98% of the time on he road (away from 'home base' Wifi and wired Ethernet and convenient plugs ) that $300 has a much higher value proposition than some 'homebody' tablet that never goes anywhere while playing about as many games/content consumption hours as getting billable work done. It will sell. It won't be a top 10 overall laptop volume seller (outsell the MBA ), but it will sell.
For WoA via cloud services the price is likely going to be cheaper than Intel instances. ( true in other categories of cloud services; so pretty likely going to be true in delivering windows also. )
Yes! A product that can stay un-updated for 1000 and even 2000 days in a row is definitely not off the ground in the PC market. Like a Formula 1 race car is not really part of the car market.
If you want to hear an irrational argument go no further then, WoA must succeed because Microsoft "can't afford for it not to work". I mean that's true, but who said that Microsoft will survive? Not me.
But they are catching up. AMD on TSMC N5 is better. Apple iterating SoC at a pace 18+ months won't necessarily race forward faster than AMD/Intel are moving at 4-6 years from now. Intel is throwing more power consumption at their laptop Gen 12 and 13 SoCs, but AMD isn't (because have access to better fab nodes and are consistently iterating. Mobile hasn't been a primary focus for AMD but as they gotten more financially stable they are starting to turn the ship on that battlefront. ). On desktop CPU and GPU Intel/AMD are throwing out the possible Perf/Watt overall gains out to overclock to higher benchmark metrics. But when set not to 'over reach' they are making progress. There is no good evidence that they are fundamentally 'stuck' on Perf/Watt on some technological basis on any possible SoC design.
AMD/Intel are not going to get to "close enough" perf/watt parity in 1-2 years, but 3-6 years out is not a sure thing. Nor does x86_64 need to carry around all of the 1980's and 1990's baggage to run 32 bit OS in 3-6 years. In 2025 when Windows 10 dies, AMD/Intel can let go of some substantive 'boat anchor' features for mainstream SoCs.
Healthy and real competition between AMD and Intel will keep them moving forward on x86_64. With some outside competitive pressure it likely will get easier to throw some of the boat anchor, old stuff into non mainstream SoCs ( and clean up the mainstream SoCS. )
And if Apple keeps sticking to their "only Apple GPU drivers only" path for next 5-6 years they aren't going to capture broad swathes of PC market they don't have now either.
Probably rolls well into at least the next decade as the major player. Windows 10 doesn't even go away until 2025. Apple will dump macOS on Intel by end of the decade. But that is only going to create some refugees onto Windows on x86_64. Laptops have taken large share away from desktops, but the smaller, more 'die hard', remnants of desktop are going to be harder to move.
Arm will proliferate into a higher volume of non-Windows devices ( more phones , tablets , heatsets , etc) , but Windows + any arch SoC has problems there anyway. x86 isn't really the major problematical issue there.
In the Windows space , as long as the top vendors (i.e., at the moment HP , Dell , Lenovo) are "sell everything to everybody" vendors if any one of the three says " we are quitting x86_64 entirely" then the other two will say "buy from me we still sell everything for everybody". Either you have to get all of them to quit all at once at the same time or you have a transition problem. Microsoft cannot unilaterally make that call. Have a chorus of system vendors with completing goals to manage.
All of that will slow the decline. There will be substantive share loss , but relatively rapid implosion? Probably not. There is no "slam dunk" easy "whole market" Windows win here for Qualcomm , or any other Arm vendor.
All you say is probably true, but I think Intel/ AMD are probably closer to throwing in the towel themselves than you might think. They're just two companies which have to fund the entirety of x86 development between them. Even despite their deep pockets that's a big ask, and if sales volumes do start to shrink meaningfully that compounds the cost problem very quickly. This is even more true if Microsoft/ Google really grab at the lucrative high end with the chips they are developing. When it becomes clearly cheaper/ easier to start making Arm (or RISC V) chips than continue with x86 I don't think it will take either company that long to make the decision (tbh I'd be amazed if they aren't already developing such chips behind the scenes). Undoubtedly there will be a long tail of applications, particularly in enterprise systems, but I think development will wind down and narrow in scope surprisingly quickly once it starts.
There is no future in this decade or even in the next where ARM takes over X86 in the desktop and even laptop market. Apple will probably gain market share with Macs but I don't see it going over 15-20%... ever...
Windows will remain the dominant OS for desktops and laptops.
Qualcomm has basically no competitor in Windows on Arm (as noone seems interested to invest in this segment) and is behind in performance and not competitive in price.
Developers are not interested in WoA. Emulation has roughly a 30% impact on performance, so Qualcomm should makes chips that are over 30% faster than X86 and not more expensive just to keep up...
This does not mean that ARM will only be a Mac technology, but that it will take decades before the situation radically changes, if ever... And I say this as a big fan of ARM (the tecnology, not the company) who hopes Windows on ARM could develop as soon as possible and become dominant at some point.
Windows will remain the dominant OS for desktops and laptops.
Qualcomm has basically no competitor in Windows on Arm (as noone seems interested to invest in this segment) and is behind in performance and not competitive in price.
Developers are not interested in WoA. Emulation has roughly a 30% impact on performance, so Qualcomm should makes chips that are over 30% faster than X86 and not more expensive just to keep up...
This does not mean that ARM will only be a Mac technology, but that it will take decades before the situation radically changes, if ever... And I say this as a big fan of ARM (the tecnology, not the company) who hopes Windows on ARM could develop as soon as possible and become dominant at some point.
And ? But what is "x86 development" ?
From an old Ryzen article about the general architecture.
The AMD Zen and Ryzen 7 Review: A Deep Dive on 1800X, 1700X and 1700
"x86" never makes it past the decode stage. The vast bulk of the processor core is all done using micro-opcode processing. Everything past the Micro-op queue above for the most part isn't "x86" . (some semantics aspects leak through , but neither AMD and Intel are implementing it exactly the same thing in interchangeable (between companies ) lockstep. )
So your premise largely falls down to that somehow the x86_64 decoders themselves are somehow going to "bankrupt" Intel and AMD so they will have to quit. That is pretty unlikely,. It is a relatively limited amount of the design work for the core. And it doesn't even run all the time.
On vast majority of code AMD/Intel micro-op cache hit rates are 50-80%. Which means 50-80% the x86 decoder isn't even running. What AMD/Intel need to work on is making the x86 decoder run less and run more efficiently. Part of that would be tossing opcodes that 92+% of users don't even use. x86 is more constipated than complex. Kernel boot modes for 1987 vintage version of DOS ... does that really need to be implemented in hardware? (versus pushing that workload to some QEMU like processor emulator). Six different SIMD modes. Really ... just one good one would be a major problem? Probably not. ( and again push some 1999 vintage SIMD code off to a software emulator).
Will that completely change x86_64 to make it perfectly uniformly regular? No. But refusal to toss any 'bad idea' from the 70s , 80s, 90s from the opcode mix has shifted from being an "asset" to being a liability. Really overdue for a deep dive into just what are folks with modern code are really using now. And emulation is "fast enough" to run old apps in lots more cases in the 2020-2030's than it was back in 2005.
If they shifted focus to optimizing the running of code from 2007 and forward in 2025 then that would be almost a 20 year old code base they would be targeting. It isn't going to that small. Reaching all the way back to 1990 is goofy. Apple isn't. Apple isn't even reach back 10 years. ( dumped all the mac 32 bit apps down the drain at this point. ). That is actually a substantive chunk of the gap. ( 90% of time looking backward as opposed to spending 90% of time looking forward. )
I thnk x86 is 'messy' but the so called 'complex' part isn't the big problem at this point. Clinging too hard to the distant past is the bigger issue.
Not really. It is just a decoder.
ARM has raised the stakes in the trial.
www.semianalysis.com
Arm’s Nuclear Option – Qualcomm Must Cancel Next-Generation Products If Arm Succeeds
Arm denies allegations of changing its business modelIf you have been keeping up with the biggest soap opera in the tech world, you know Arm sued Qualcomm. Qualcomm more recently made some pretty b…
www.semianalysis.com
These quotes look like serious trouble for Qualcomm in the near term:ARM has raised the stakes in the trial.
Arm’s Nuclear Option – Qualcomm Must Cancel Next-Generation Products If Arm Succeeds
Arm denies allegations of changing its business modelIf you have been keeping up with the biggest soap opera in the tech world, you know Arm sued Qualcomm. Qualcomm more recently made some pretty b…
It looks to me like they would need intervention by a judge to force Arm to get a Nuvia SoC validated. Qualcomm might have a case based on their own ALA license but that will have to be decided by a jury which will take years. Until then, Arm is refusing to let Qualcomm validate their SoCs. I hope that Qualcomm's RISC-V work is progressing quickly.
No, they want them to pay to use the designs. If they don’t pay, THEN they can destroy them
The "base station" (server) probably could continue if just pretend it is still chasing the "server market" that Nuvia was originally licensed for. That isn't why Qualcomm paid ~ $1B for the company , but there isn't really good justification for size of an amount anyway.
Qualcomm significantly needs a 'value add' differentiator in PC market. Depending upon what Apple does may need an AR/VR push also.
It isn't really much of an across the board nuclear option when Qualcomm's ALA really doesn't have much product if remove the 'fig leaf' that it is suppose to be air cover for the Nuvia work. The current Snapdragon , AR/VR , phones , etc stuff is Tech licenses. ( perhaps some narrow corner case on a niche feature).
It is "nuclear" on the Nuvia stuff, but that is the primarily what the suit is all about. It isn't really aimed at all of Qualcomm's licenses ( Perhaps nuclear in a 'neutron bomb' sense where it is suppose to mainly kill people and leave most of the buildings/infrastructure still standing. )
That RISC-V doesn't make much sense as a viable short-intermediate term option. Qualcomm just went through gyrations to keep keep 32-bit Android Arm apps afloat in the Snapdragon 8 gen 2
"... the AP packs 4x Cortex-A715 cores. The quad-cluster arrangement rumour wasn't entirely incorrect, though. However, Arm added 32-bit app support to the Snapdragon 8 Gen 2's Cortex-A510 cores, eliminating the need for the previous-generation Cortex-A710 cluster. ..."
Qualcomm Snapdragon 8 Gen 2 to feature a tri-cluster CPU with four Cortex-A715 cores
Kuba Wojciechowski has confirmed the Qualcomm Snapdragon 8 Gen 2 will feature 1x Cortex-X3, 4x Cortex-A715 and 3x Cortex-A510 cores in a tri-cluster arrangement. The previously rumoured 2x Cortex-A710 cluster is no longer needed because Arm added 32-bit app support to the Cortex-A510.
( previously it was rumored that there was two 715 and two 710 . (where the 710's are being held around for 32-bit apps. China and other places were non-Google Store is prevelant there is no requirement for 64-bit app submissions to the store. ).
If Qualcomm is putting in substantial effort trying to hold onto regional 32-bit ARm app store markets... are they really in a position to dump it all any time soon?
RiSC-V would be just as far off into the distance future as the conclusion of the trial. It is not going to shorten the timeline.
p.s. Odd qualcomms spec sheet is not clear . Conflicts about whether 710 are there or not . If there just more 32-bit gyrations .
Snapdragon 8 Gen 2 deep dive: Everything you need to know
All you'll ever need to know about Qualcomm's latest Snapdragon 8 Gen 2 mobile platform for next-gen phones can be found here.
Last edited:
If companies can't rely on ARM, they will look for an alternative. It seems that Google is also interested in RISC-V. In fact, Google will be giving a talk on Android and RISC-V at the upcoming RISC-V summit.
RISC-V Summit 2022: Keynote: The Android Open Source Project...
View more about this event at RISC-V Summit 2022
If companies can't rely on ARM, they will look for an alternative. It seems that Google is also interested in RISC-V. In fact, Google will be giving a talk on Android and RISC-V at the upcoming RISC-V summit.
RISC-V Summit 2022: Keynote: The Android Open Source Project...
View more about this event at RISC-V Summit 2022riscvsummit2022.sched.com
Android runs on x86. Android has never been adverse to alternatives instruction set implementations. But as some viable short-to-intermediate term mass exodus vehicle off of Arm ... that isn't coming. RISC-V generally isn't outperforming Arm. Nor is the software foundation breadth really mature.
RISC-V is going to make progress over time , but it is very likely to be like how Arm progressed in cracking the server market. There were several years where Arm Server packages were coming "real soon now" over and over again while the software foundation caught up to the hype train.
RISC-V inside a Qualcomm network , storage, or modem controller is going to be much easier for Qualcomm to pull of where there is zero end user person's software involved. Narrow 'firmware' software stack that can be easily targeted and can relatively easily be made totally decoupled from legacy software.
But what was really "unreliable" about Arm here? The contract says you have to check with them before doing acquisitions. Qualcomm did , but didn't like the "answer" so said "F you" and did what they wanted to do anyway. Hence, they got sued. How is the it even remotely "unreliable Arm" the primary issue there ?
The primarily 'unreliable' issue with Arm is not their contract structure , but ownership who is far more focus on making money than on long term stewardship of Arm IP ecosystem. Softbank paid way, way too much for Arm and trying to recover sunk costs that really can't be recovered without negatively impacting Arm long term. The Nvidia buyout would have been even worse ! ( Can see how Nvidia is having revenue flow problems now... imagine if they have thrown 10's Billions out the window with no practical return. They'd would be even more hard up than Softbank is on being a vulture capitalist on trying to pimp maximal cash out of Arm. )
Minus Arm IPO-ing at an irrationally high price point and the Arm management devoted to chasing that price point long term, this dust up with Qualcomm will likely be just a hiccup. Arm vs RISC-V is on track to turning out similar to what happened between Linux and BSD Unix on the path to being the dominate "open" Unix option. Linux has a 'benevolent dictator" in Linus that helped keep some (or at least greater ) coherence than BSD (which fragmented off into mostly compatible, but different focus , camps: FreeBSD , NetBSD , OpenBSD . RISC-V has the BSD balkanization trap that they may fall into. Is that more risky than Arm's overvaluation problem? eh maybe , maybe not.
P.S. Arm didn't take over the "2006-2010" crown that x86 held. Nor likely will they in the future. x86_64 probably will not shrink completely away. Nor will RISC-V sweep both of them away. Probably going to go a future where there is lots more parity and more healthy competition at the "raw iron" foundation level.
Last edited:
RISC-V is working on a solution: profiles.
riscv-profiles/profiles.adoc at main · riscv/riscv-profiles
RISC-V Architecture Profiles. Contribute to riscv/riscv-profiles development by creating an account on GitHub.
github.com
What holds Linux together is Linux's stricter licensing, not Linus Torvald. In the consumer space, BSD-derived OSes are far more popular than Linux OSes. Apple, Playstation, and Nintendo hardware all use BSD-derived OSes.
As Dylan Patel said a couple of months ago, it looks like ARM will soon change its licensing policy.
ARM hates Qualcomm for 2 reasons:
1. Qualcomm lobbied heavily against the ARM + Nvidia deal
2. Qualcomm's Nuvia is a major threat to ARM's business.
ARM is trying its hardest to prevent Nuvia from coming to the market. If it does come to market, ARM wants to a far bigger slice of Qualcomm's profits than before. Hence, ARM sued Qualcomm to block Nuvia designs.
Qualcomm lobbied the US government to block the Nvidia deal. It's very obvious why. Qualcomm has very high custom ARM-core ambitions from their purchase of Nuvia. They want to be the Apple Silicon of the PC/Android world.
Nvidia owning ARM would have injected resources into stock ARM cores and made stock ARM cores more competitive against whatever Qualcomm is planning to do.
Qualcomm wants stock ARM cores to not be competitive in the near future which would make them the defacto high-end consumer ARM supplier.
1. Qualcomm lobbied heavily against the ARM + Nvidia deal
2. Qualcomm's Nuvia is a major threat to ARM's business.
ARM is trying its hardest to prevent Nuvia from coming to the market. If it does come to market, ARM wants to a far bigger slice of Qualcomm's profits than before. Hence, ARM sued Qualcomm to block Nuvia designs.
Qualcomm lobbied the US government to block the Nvidia deal. It's very obvious why. Qualcomm has very high custom ARM-core ambitions from their purchase of Nuvia. They want to be the Apple Silicon of the PC/Android world.
Nvidia owning ARM would have injected resources into stock ARM cores and made stock ARM cores more competitive against whatever Qualcomm is planning to do.
Qualcomm wants stock ARM cores to not be competitive in the near future which would make them the defacto high-end consumer ARM supplier.
I think Qualcomm, like many other companies, feared a change in the licensing model, not Nvidia's money. Qualcomm could compete with Nvidia + ARM with fair terms, but not with these new terms.