This is maddingly frustrating. I *really* wanted an ARM processor for software development that could compete with Apple silicon and maybe even x86. Nuvia seemed to be leading to this. ARM is squandering an opportunity. This reminds me of AT&T and BSD
Burning through venture capital to develope and improve the ARM technology is not a sustainable business model. As we see with ARM, the VCs might get bored with spending a lot of money without much - if any - of a return. Qualcom wants to pay as little as possible to ARM and buys a startup to get the IP they need cheaper than negotiating the licenses with ARM and any third party involved so they then can sell their chips to OEMs and make them pay for licenses. Good for Qualcom :D ARM on the other hand wants to earn as much money as possible so the VCs and other shareholders get their RoI and have enough left to improve so they stay competitive. Good for ARM :D Not too long ago, maybe 10 or 15 years, we expected new and exiting hardware out of China based on MIPS and/or OpenSPARC. Turns out IP alone does not make a chip, you need more than that. That's where ARMS support for developers comes in and this support has to be payed for , too.
I think that it is there. You just need to assemble your system. Gigabyte has a dual core motherboard for Amphere processors. You put two 128 core Amphere Altra Max CPUs into the board. Plug a PCIe graphic display card. Plug another PCIe SSD. Plug 16 pieces of DIMM may be 512G RAM. No Apple silicon can match your development machine. You can have it in a 2U setup.
Great breakdown of the case. I was waiting for RISC-V to be mentioned at the end. If anything this might push future chip startups to lean in that direction. If ARM wins, the Nuvia purchase would most likely have a negative impact on any future startup looking to do something similar to Nuvia. At least with RISC-V there's a much higher chance of being aquired if they can help a chip maker be less reliant on ARM.
Probably not. RISC-V is so modular to the extent that binaries are no longer portable between different chips because they each support slightly different subsets of the RISC-V ISA. The ARM CEO mentioned this himself in an interview IIRC.
From what I heard, one of the issues with RISC-V is that while the architecture is open, alot of the extensions are closed source (or paid, one of those)
@@albertsun3393 everything that linux or other OSes needs are open Simd and visualization extensions open too. So, everything that are mature and needed by most users eventually will be turned into standard extension. For experimental staff riscv has a lot of places(instruction encodings, mmu types, internal interrupt numbers) that marked "for vendor use" Main goal of riscv is to give vendors linux capable system that they can add problem specific parts(instructions, interrupts...)
Funny how Qualcomm - the IP and licensing stickler - doesn't understand ARM's licensing terms. Qualcomm's been IP extorting the handset business for years with their modems and non-FRAND licensing. You'd think they'd be smart enough to just cough up the cash they way they expect everyone else to do for their modems.
Correct. But Qualcomm has legal depth, deep pockets, and will not be moving. A reasonable Judge might say 3-5-10% is incidental and normal in the trade. OEM royalty -- forget it QC will not have a bar of it. ARM has also failed to improve faster relative to others- it just keeps adding fluff or blocks that others do better. The shrink to circular gates/wells is a gamechanger, and QC has the pockets to do a RISC5 port. ARM wants to move into the QC space, but it will have the edge over QC id they can favor sweet OEM deals. So will QC roll over and die, let ARM eat its lunch - I think not. NVIDIA uses R5 core-bits in its GPU's. An alternative is QC does not blink, air the ugly court case, then rabbit out of a hat, we gone R5, and the rest of ARM's customers move.
Arm this is brilliant, this is exactly what we need to get everyone to move to the open source RISC-V architecture. This is essentially AT&T vs BSD Unix all over again... hold my beer, I'm going to make some popcorn. 🍿
RISC-V is not open *source*, "just" open *specifications*. You'd still need cores, of which many are closed, and many come with royalties and licenses again. It's a slippery slope. :)
@@CyReVolt The patients for the original PowerPC architecture are already starting to expire, and whatever is still patient encumbered the OpenPOWER Foundation is trying to make available, for instance there are already two reference Power ISA v3.0 softcore implementations that are entirely open source, Libre-SOC and Microwatt. Nvidia and Mellanox are both founding members of the OpenPOWER Foundation, and since Nvidia was denied the ability to purchase Arm you can bet that they are going to double down on evangelizing the Power ISA. The OpenPOWER Foundation is now actually organized under the Linux Foundation. The goal is to create a new open platform akin to the original IBM compatible PC after Compaq reverse engineered IBM's BIOS. Both RISC-V and OpenPOWER are little endian RISC implementations with objectives that almost entirely overlap, so these two project need to come together to facilitate these shared common goals.
I would put it this way: Arm PC is a Competitor to X86 AND keeps PC competitiv against (Arm)Mac so actually this lawsuit foremost is beneficial for Apple by delaying the maturity of serious ARM Desktop PCs, right? Can't help but have the feeling that ARM is burrying itself with this... smells like adobe, Nintendo, OneWheel, BMW,.... "the old business models are more and more failing but this anti consumer-behavior is doomed to collapse hard and ugly..." nostradamos ;)
I am kind of against ARM entering the PC market, especially desktops given that their SOC design limits upgradeability and repairability. But on thin & light laptops it can be quite beneficial especially for battery life.
And nothing of value is lost, instead the market might be forced to embrace RISC-V rather than ARM. And RISC-V follows a somewhat similar simple instructions approach but it an "open standard instruction set architecture" so companies can do whatever they want to do with it (and have been doing so already). It was starting to get competitive with ARM already, getting more resources behind it because ARM is digging its own grave just means faster progression for RISC-V.
@@DragonOfTheMortalKombat more than that, having no BIOS GUI (except some rare implementations of UEFI or petitboot) is extremely detrimental for many PC users as they will have to enter the deep water of U-boot just to change some boot options. Having to play with device trees and having to maintain images for every single implementation of the same SOC is a lot of support work and that's one of the reason why a lot of devboards from the past decade turned to be quickly unusable (or unsafe to use) even with FOS operating systems.
I wouldn't hold my breath for arm pc/laptop. Not having standardized platform, means continued mess with bootloaders per device, firmware per device, and lack of open drivers, and a pile of e-waste no one supports because it is not making money.
@@DragonOfTheMortalKombat ARM is strictly an IP company. They're not entering any market. It's up to the companies that license the ARM architecture to create processors what they do with them.
yes they were. when you buy something groundbreaking that can potentially disrupt the market (5x gains) you can bet the rest of the market would try to kill you. also literally.
ARM is poisoning themselves with this lawsuit and my guess is that this is what Qualcomm is banking on. There's a middle ground which should be reached and I would hope that is what yes achieved. If I were negotiating terms as ARM, I would provision that any ALA contracts give joint custody of the generated IP which could be relicensed as a TLA design. ARM designs would continue to improve and acquire more capabilities and the TLA licenses would provide a mass market solution. The companies which built their systems using ALA would have a first to market advantage and the entire ecosystem could mature and become more power per instruction efficient. Long term we just to see advances which create more powerful and efficient designs. As the licensor, ARM still has a path to profit by encouraging a higher adoption of their cores. They are biting the hands which feed them if they win this suit.
Now here's my question: what *exactly* does the Qualcomm ALA say? Specifically, does it say "code must be Qualcomm IP only" or "invalid in the event of a merger or acquisition" or anything to that effect?
20:30 From what it sounds like, it absolutely seems to me like Arm is poisoning the well. If I had some influence in a company coming up with a long-term roadmap for some SoCs, this behavior from Arm would absolutely have me making contingencies to pivot to RISC-V if I wasn't already.
Qualcomm is so stupid. They had a World Class CPU Design team... That was beating ARM designs for years.. But all got fired after their Server CPU didn't sell... Now they spend Billions on Nuvia and possibly can release nothing 😅
So the TLDR is Qualcomm didn't get the appropriate license that covered everything what Nuvia dose before merging the two companies, and now trying to use their own incomplete license which had lower royalty rate on Nuvia's design. Of course ARM will sue them, especially now ARM are starved of cash after acquisition deal with NVIDIA was killed partially due to opposition from Qualcomm.
I noticed this background and recognized it and for a moment I thought I was going crazy and probably misremembered STH's Patrick new desk. IDK - anyway I look at this it looks bad for ARM. If they win, they will have such leverage over their customers that no-one will want to make new business with them. And no new companies will start. If they lose - well it is bad for ARM, because money. And they might be more cagey about licensing stuff. Damn Softbank. It is going to pop up RISCV. It would do more if it was more mature, but it will work in its favor regardless of what happens now.
Sounds like ARM just wants to double dip. Regardless of the outcome this absolutely poisons the well. It shows no matter how much work gets done on top of ARM's IP that ARM still views it as theirs and they'll reneg whenever they want once your work becomes more valuable than it was when previously negotiated. That's like selling your house to a 2nd party then suing a 3rd party when they rebuy that house for a higher price after the 2nd party put a bunch of work into it. Arm's been spending too much time with Apple and wants their cut every step down the line.
Not sure how I feel about this, besides RISC-V getting more traction. Being an OpenSource guy, I think it would be nice to see so more open cpu efforts, without ending up with "too many options to choose from" side-effects.
Whether cores are open is up to their owners though, RISC-V being just specifications. Vendors may still charge royalties and license fees, and SiFive sponsoring *this* video is... telling, huh? 🙃
@@CyReVolt Difference is that you are free to design your own cpu. you don't have to pay arm royalties for using the architecture nor an additional tax for creating your own core. It would be no different than manufacturers paying Qualcomm to buy snapdragon chips. And if you want to use snapdragon SoC and modify the core, you should rightfully pay them for it. The benefit is that it cuts out the middleman from charging royalties to qualcomm, and their customers for using that IP. Think of it like this, imagine if x86 became a free and open architecture for anyone to design their own cores. it would be fair to allow nvidia to design their own cpu using x86, but not fair to let them use existing Ryzen or Intel cores to make their cpu, unless they paid for that IP, otherwise, nobody would want to foot the bill of development
Hi Ian, I'm surprised that you didn't mention the new Board members of Arm and what impact that may have. Adding Dr Paul Jacobs, the son of the legendary Dr Irwin Jacobs - (both ex Qualcomm CEO), is a smart move. Also adding Rosemary Schooler brings some Intel sales knowledge too. Add that to the appointment of the Apple designer Tony Fadell and it seems that the US team is gaining strength to deal with the Qualcomm issue. The tough year ahead for tech will hopefully force them to concentrate engineering effort on the future recovery and not get too embroiled in enriching lawyers who appear to have err'd on both sides. I'm no fan of Arm mgmt but prefer the semi leaders to recognise how much collaboration is necessary between them.
I am a professional digital IC Designer that has worked for the likes of e.g. Intel, and when it comes to companies you do not wish to work with, it's exactly 2 names: Qualcomm and ARM. Besides the fact that I think that ARM IP isn't even that great (the RTL code does not impress me much tbh), their licensing ideas become more and more ridiculous. Qualcomm on the other hand by now is seemingly more a law firm than a chip company, which is a shame. Because their silicon is usually pretty advanced and great, but the way Qualcomm treats basically anyone outside of their own walls is a 100% model of how NOT do do it. I seriously hope RISC-V rises to a point where ARM becomes worthless or at least so pressured that they come back to earth with what they are demanding. They are trying to milk this instruction set to the limit, but frankly, competation is on the rise. So especially alianating your architecture license customers like that will backfire big time. It just takes Qualcomm, Apple or even Intel/AMD to really embrace RISC-V and ARM is dead. And being an extendable RISC-architecture, it is really mostly about re-engineering the decoder. Many intstructions are similar enough to turn a very efficient ARM design into a RISC-V design.
Great video with a unique inside view of how ARM's IP business works! Don't most, if not all, IP agreements contain very specific wording on what happens when a licensee gets acquired?
I'm very bullish on Arm's architecture, their IPC and perf/watt is just insane given the tiny caches. But this case just doesn't seem right, Arm should not be allowed any claim to Nuvia's IP. It reminds me of Oracle vs Google/Android except ISA compatibility instead of API compatibility. I hope Arm's case gets thrown out quickly
ARM has a huge advantage over AMD and Intel in those regards to power. A customer using their IP can say "OK folks recompile your sh%t for 64-bit only we're throwing away legacy". Tell that to the Windows crowd.
@@ColdPotato It's not about perf. x86 has no issue with that. But all the legacy (or not that legacy, avx512 hint hint) crap, just adds a big constant overhead. And this gets really noticeable at small wattages, because this "legacy" becomes a significant chunk of power wastage the lower you go. I.e. the overhead of instruction decoder might be unnoticeable at 4GHz+ with 32+ high performance cores running simd stuff, but once you try to go low-power it starts to matter.
Seems pretty even overall... hopefully the case gets settled quickly, with Qualcomm and ARM agreeing to some moderately priced, additional license agreements.
I love the origin of ARM, I like the history of it. I dislike the company that purchased it and wants to put a dark cloud over itself. Requesting that qualcomm destroy all of it's technology relating to nuvia, most likely either leading to a strong arm in licensing renewal that hurts everything down the line or a complete destruction of this technology. It's looking like it's a maximize profits at all costs and or destroy everything in your way to achieve that goal. I would love if this accelerates risc-v and it becomes the defacto standard overtaking arm. Risc has proven itself time and again, so it's a good technology that should be pursued and not have a ruthless company that will hurt everyone to get what it wants. I know licenses can be tricky and companies want to get their fair share, but this scorched earth approach is insidious and i'm glad people are taking note and it's making people wary of wanting to continue with arm.
If they can pull off good performance and power efficiency, RISC-V chips could become "the new ARM" in the end, in being the most-used chip solution, which would be great! =D
@@greecemobile7610 For now. There's a bunch of money getting poured into RISC-V right now, but it's not at the same level that, say, Intel, Apple, and AMD spend on their chip architecture RND. Honestly, I think that what it will take is for one of the big players to get burned by ARM badly enough to start pumping some serious fuck you money and resources into RISC-V to try and use it as a viable alternative architecture for _something_ at a large scale.
Arm architecture is definitely the future for the coming decade. Qualcomm pushing Nuvia to go desktop is the logical next step, simply because of the power to performance ratio that x86 just hasn't been able to demonstrate. Arm sees the opportunity that Qualcomm+Nuvia unlock and wants to dip their hand into the honey jar, Pooh style. As far as arm servers are concerned, the future to me seems to be pointing in the direction that Graviton is going imho.
I haven't read the brief, but it seems like ARM simply want to force Qualcomm to renegotiate. The fact that they're continuing to support them (and that this is "orthogonal to the case") hints at this, although their support may be contractually obliged. I can imagine they had a chat beforehand and said "Look, sorry fellas but we need to sue you so we can get a bit more money out of this deal you've pivoted on." I expect Qualcomm will renegotiate for something costing somewhere between Nuvia's agreement and their own current agreement(s). Basically, it seems like a strong-arm negotiation tactic. (No pun intended!)
I read what legal filings that are on line. Nowhere did ARM state in their deal with Nuvia that if Nuvia is acquired that the deal they made with ARM is off. Both sides have weaknesses in their cases. Destroying 3 years of development is not going to happen. The Judge will send it to mediation and eventually Qualcomm will give ARM some money to go away.
This kind of situations will simply lead the future for hardware to be the same as has been for software lately, more open source standards. RISC-V stands to gain a lot from this license and patent fatigue that is sure to grow with such moves by ARM and others.
It feels like any new semiconductor startup will avoid using ARM as a base. If there's any sort of new startup here, they will focus on RISC-V. I'd say any new "NUVIA" type of startup that is doing anything big will avoid using ARM to avoid these types of lawsuits. ARM has already poisoned the well.
I'd be concerned that Arm's behaviour might be anti-competitive as well. Are they essentially saying that they can decide that Qualcomm cannot sell this kind of product. It shouldn't have got to court, ultimately it's going to hurt Arm and push more companies towards RISC-V (in the longer term). Either way, the IP should not be destroyed, human effort shouldn't be wasted in that way just because Arm has a bad take.
RISC-V on Windows PC desktop is inconceivable for the time being and a long time yet, ARM was hard enough. And at least ARM has the advantage of specs being rich and fixed in place, which is necessary when you have binary compatibility to deal with. But as far as server... well that looks like the beginning of the end for ARM. There won't be much visible for the next few years, but under the sheets, things would be brewing. Also interesting how ARM started with microcontrollers essentially... and is now rapidly losing them to RISC-V.
I'm rooting for QCOM, I would love to see some competition in the processor space. however, QCOM needs to adopt a more intel like model where they don't force customers to use QCOM power devices. their power portfolio is weak. Battery chargers suck, PMICs are inefficient, they need better supporting devices to succeed in the long run.
like mommy and daddy are fighting but dont worry guys arm and qcom still love you. lol the real reason is apple doesnt want qcom to take their chip devs. they want to cost qcom more money.
I think Ampere is in a much better position. If they wait, they can use the time to make their systems better. If they make a deal, they can avoid all the pitfalls that Qualcomm created.
From the headlines I thought it was likely that ARM gave Nuvia a favorable licensing (or whatever) deal. As Nuvia, while having some clout, personal and funding wise were trying to get into the Server space. Something a lot more valuable to ARM than whatever royalties they’d get from Nuvia in a few years. From the video it’s sounding that’s not the case or even something kinda opposite is. I’m not read up enough on this case to have a solid opinion but I’m getting Oracle vibes…
Windows doesn’t have a place where to put libraries so it’s actually not there fault and if your emu can’t deal with different paths you should reach reconsider how you designed it.
Is this your new studio? Looks nice! :) Ah, it's Patrick's... Anyway, thank you for the in-depth explanation. Doesn't this cause any problems for an ongoing lawsuit?
It seems that Qualcomm have stuck their heads in the sand with the ARM licensing. They've just done the "it'll be alright" move and probably them and the Nuvia people have "group-thinked" their way to believing that it will. The McDonald's comparison is valid. That said, it shouldn't be difficult to resolve and they should do so for all concerned - except, that is, for the lawyers involved.
Long comment warning! This is interesting (to me at least) because I've often wondered how ARM's ... unique? ... business model will impact the acquisition of startups, as it's quite well known that in business the chaos of little companies is (for better or worse ... mostly worse) temporary, and in the end big fish eat little fish and bigger fish eat big fish until the mammoth sharks dominate E V E R Y T H I N G. Whereas ARM's licensing approach ... doesn't really fit these seas. So this lawsuit ... on the one hand ARM isn't being a bad guy for litigation in that they must protect the integrity of their licensing that is their business model. On the other hand, it is most definitely shooting themselves in the foot because, as said, it's a business model that doesn't really fit these seas, and the more they defend it with loaded arms, the more they scare their customers. If ARM isn't stupid (which, again, they developed a business model that doesn't fit their ambitions, so this is highly questionable) ARM will eschew strict license adherence and find a way to amicably forgive, grandfather in, or renegotiate the licenses in a way that Qualcom "wins", BEFORE this actually gets too far into trial, and the litigation gets dropped. This way they don't scare their customers and show future customers that ARM isn't a complete pushover and will defend its business, but can also be a good guy about tricky situations. That ARM is only scary when rampant abuse is involved, and as this isn't a case of rampant abuse, it can come to a happy ending that isn't scary. However, as ARM's business acumen is ... questionable ... I wouldn't place bets on this being the outcome. If ARM pushes this, right or wrong, in the end they're going to lose money, even if the win the award of being a Technically Correct Pedant. Because RISC-V is showing that ARM isn't the only way, and the last thing that ARM needs is to scare its customers into joining hands to pave the road for RISC-V. But even if that's what happens, that's ARM's problem. Whether ARM or RISC is better for consumers is a different topic entirely. LOL But if *I* were ARM, I would not only be looking to Good Guy in a new licensing agreement for Qualcom that both sides can live with, but I'd be re-examining and iteratively improving my licensing business model to be flexible and friendly to reduce the RISC of this ARMs race.
@@kevikiru usually after a lawsuit is filed, both parties are more inclined to reach a settlement. Both have things to lose if the lawsuit will follow all the way through.
At 16:38 did you mean to say Apple instead of Arm?
2 роки тому+1
I guess the next logical step for Qualcomm (and many other companies) is RISC-V. ARM proved it is not impossible to introduce brand new hardware platform for mobile, consumer and server market.
3:00 5x gains, yeah, nobody is gonna allow them to disrupt current market that much. thats gonna be a hard fight to win. ps we live in the world where 5-15% "gains" in performance per year made "the norm" by tech giants, to milk the market and spend as low as possible on R&D. slow progress = less price drop on old products = bigger margins = more profits. fast progress is very bad for business of milking people's wallets.
Dennard scaling ended years ago, new processes cost far more to develop and only logic scales now, not cache memory or io, while higher frequency with power efficiency needs smart design. No need for conspiracy theories, in the past most performance gains came from process improvement.
@@RobBCactive thats not a conspiracy theory. Lets take Nvidya/AMooDe. Nvidya makes GaaS = wants PC gaming dead. AMooDe makes PS5-Xbox = wants PC gaming dead. result? = overpriced af GPUs = PC gaming essentially IS dead at this point, nobody can afford current gen gaming, so game devs make games for obsolete hardware. AMooDe has Vcache tech to kill Intel yet chose to release 7000 CPUs overpriced aaaand without 3D Vcache to not compete much and to not drop prices much on older CPUs it still makes. with GPU prices we all are familiar. + a Steam Deck exists with old AMooDe tech running AAA games for less then 600 bucks, meanwhile smartphone SoCs can run only 10 y.o. graphics and stupid people are still buying that morally obsolete crap for 1000 bucks+. Now imagine a Steam Deck with latest AMD hardware, Zen4, 3D Vcache, RDNA3, all that tech already exists and the only reason we don't see it is such a device would literally kill ps5/xbox market, which is a market AMooDe sells its chips to. And imagine it in a smartphone form factor. it would kill everything on a smartphone market. AAA gaming instead of 10 y.o. morally obsolete crap in a pocket. it also would severely damage the PC market because why buy PCs if you can have everything in your pocket. in 1 device. no need for gaming PCs unless they start to offer truly next-gen 5x performance + photorealistic graphics - all of that. That is what AMooDe sits on pretty much right now. And thats facts, not a a conspiracy theory.
@@rawdez_.... Valve is building a steam deck 2 with amd's new apu design Hell amd has been building up the dragon range series of apus with explicit intention of making power efficient low cost gaming (extremely good gpu systems) And you do realise that ps5 and xbox consoles are usually sold at cost or a loss and are made up through game sales? They will always cost less than a pc due to there being less focused on upfront profit Not to mention gpu crisis was almost entirely due to the mining boom - 4080 just failed in it's launch to the point Nvidia is considering a price cut for it. If anything amd is bending over backwards in every market to gain market share - and yes they make far more selling to servers and pc crowd than consoles because of the bulk agreements You also forgot intel in there - why are they with holding their tech if it's all a conspiracy? They literally can't profit off console gaming
@@rawdez_this is literally one of the wildest theories i have heard till late I would rather believe gpu crisis would be because of nvidia and amd playing the market for everything it's worth (which while isn't intentional but what's happened) You really have to re think this lad
@@aravindpallippara1577 >Valve is building we need smartphone makers building smartphones with AMD tech, not Valve building another gaming device (which is cool but not gonna help much by itself). Unless Valve make it in a smartphone form factor (e.g. with detachable controls) + with phone functionality and release it in 2023. also I really doubt AMD would allow anyone use all what they have - 3d-vcache and RDNA3 in a not heavily cut-down version just because they can't allow something better than a ps5/xbox on the market. >crisis was almost entirely due to the mining boom you look so big yet still believe in marketing fairy tales when in fact AMD/NVIDIA were way too happy to make billions on that. who are really to blame miners or those who sold overpriced pickaxes and didn't want to drop prices when mining has died? >forgot intel in there - why are they with holding their tech what tech? they get to keep selling their overheating overpriced CPUs and slow overpriced GPUs with bad drivers, lol. ps at this point we need Qualcomm and their 5x perf-per-watt improvement but with RDNA3-level or better graphics. Adreno is a trash-level GPUs. and with a price tag thats not even close to 1000 bucks but way lower. now THAT would be a real progress.
I don't understand why the FTC doesn't step in at least within any division of ARM based in America... They are a monopoly on reduced instruction sets with the only alternative being RISC-V.
Companies using arm were already looking at risc V. This surely only serves to accelerate that effort in all companies. No surprise sifive are sponsoring, they must be on an upswing for every arm lawsuit
The thing about poisoning the well is that ARM is suing on of its customers which is generally a bad thing even if there are legal reasons for doing so in this case. I do think this move was a to force a negotiating position but such business deals are always best behind closed doors as the ugliness of arguments is kept from the public eye. I would argue that the worst case and something will likely happen to some degree going forward is a modification of how ARM licenses its IP. Worst case is that ARM stops offering new ALA agreements and increases pricing on their TLA agreements. The other wild card in this disagreement is that ARM too was attempting to be acquired by nVidia. Discovery in the ARM vs. Qualcomm disagreement has the potential to reach into various documents exchanged between ARM and nVidia as part of their failed transaction.
If that is true and ARM can claim IP through the support it gave a company that is licensing it's tech, then that can prove to be a very dangerous precedent for all support contracts in R&D.
It's actually pretty simple if Nuvia did have some closure in theirs lincese contract, that nobody except Nuvia have right to use the tech of ARM, then 100% Qualcomm will get a problem, but when the license agreement is not transparent, then Qualcomm always can say that they as owners of Nuvia have full right and all IPs which belonged to Nuvia belong to Qualcomm.
I read the ARM account fillings in UK, they protect themselves and their tools by being online only, and that it's not given like a pice of software. So I think they could turn off the access for this pretty easily.
@@michaeljarcher Basically a software lock. And what about the ARM rogue subsidiary in China? Could Qualcomm turn to them to license the missing pieces of their design?
I have to admit that I am confused as to why Qualcomm didn't just licence ARM v9 in the first place since they already have an experienced team of engineers in house. That would have saved them from all this legal messing around and would probably have cost a whole lot less than buying out a company (let alone the legal fees).
QCOM has been exclusively working with MSFT for four years on Windows on ARM. It isn’t completed. It seems neither side is serious on this effort. The exclusivity ends soon that opens the door to more vendors to jump in.
This is such a rookie M&A mistake, I'm almost shocked. Almost because I've seen this level of legal incompetence a couple times at major tech firms in the last couple years. I'd love to know if this was Arm's in house council or outside council that 'oofed'. It's not even legal's fault as this kind of thing is covered in most MBA/CFA programs.
When VIA purchased Cyrix and Centaurthey were allowed to make x86 chips because of Cyrix's and Centaurs preexisting rights. AFAIK Intel didn't make them sign a new agreement.
What's this news that Charlie from SemiAnalysis is hinting at ? Behind a paywall it is unfortunately but seemingly it's hinted that Qualcomm just shut down Nuvia ?? Is that true ?
ARM are shooting themselves in the foot here in my opinion, if you have an existing business relationship - settle it without throwing around lawsuits - especially if said partner is already licensed to the gills. Unfortunately I think this just comes down to new management looking to get some quick "wins" via brute force. EDIT: Also, if ARM are going to argue that IP developed by Nuvia with ARM licensing was not actually owned in whole by Nuvia, then why did not ARM not step in during the acquisition?
Given what TTP said about the extreme conditionality of ARM licenses, I don't think ARM themselves wants to compete in that space - since it would break their "pay per market" business model.
Qualcomm doesn't stand a chance, companies have tried this tactic of buying out others for the arm license and have failed. Its one of the reasons why all the big manufactures are desperate to buy arm. Rather than qualcomm trying to fight it they need to ask ARM how much for the correct IP and give them money for it.
What court is this brought up in? It's hysterical for me that ARM would require to destroy knowledge. How did it happen that they are opposed to development of technology and not just to how this technology is sold? Why can't Nuvia get knowledge about ARM cores and use it however they want? Isn't it ARM's problem that they did not foresee Nuvia being bought by someone? 24:00 I wonder if that restriction of target market only applies to the end product. Would be funny if ARM allowed to develop knowledge of producing CPUs only if that knowledge is used only for specific market segment. I'd imagine there are a lot of generic problems in CPU development if not most of them. Would it be good idea for Qualcomm to demonstrate that all of the important knowledge which Qualcomm acquired is generic and should not get restricted by ARM? I'm guessing any "verified" or "validated" designs which Qualcomm got are not critical for the end product they want to make (except some additional costs), this is what I would consider a non-generic knowledge which ARM could be able to restrict.
Apple hinted that they would use "Apple Silicon" in their server side components of "Apple Intelligence". Does this mean that they now have an ARM server license? If not, will ARM go after them like they are going after Qualcomm? Something tells me they wont.
To those engineer and creative individual, it must have been a headache. Their hard work become investor property and they only received a monthly salary + they are subject to be fired(but they contributed more than the investor). Arm will never invent since they play this royalties business scheme and if they did, 10 years already passed and we still looking at an arm architecture with a miniscule improvement instead of inventing new way, just like INTEL.
To be honest, I think ARM will win. Quite simply, ARM created a package deal for Nuvia to build a specific processor for a specific market and this deal was made and agreed to in writing. If there are NO provisions in the agreement for in the event Nuvia got sold to someone else(Qualcomm) and that someone else automatically got the same package as if they were Nuvia, then the deal is void. That is the way the judge will see it. Its Qualcomms fault for not negotiating with ARM before purchasing Nuvia. That is also the way the judge will see it.
I feel like ARM would have been better off being ok with the acquisition. Qualcom could have scaled and sold units. Obviously their agreement reflected the scalability. I'm sure ARM would have preferred Nuvia to scale at the higher royalty.
From what I've read & You say here Qualcoms Money people have gone through the books of Nuvia said whow that's a big licence fee going to ARM we have ARM licenses and then gone to the league team said if we don't negotiate a licence with ARM will our current Licences cover us using ip league team said 90-98% so money team then decided to save money and not purchase the necessary licences on the grounds ARM get lots if cash from Qualcom & A court case would loom bad. ARM intern said fine not willing to talk we will take you to court maybe that will get you to talk to us and that's where we are now. The only winners from this will loyers.
I do wonder if Softbank will kill ARM. If their architecture/IP is so expensive and complex to maintain perhaps it is cheaper in the long term to think outside the box and evolve the open source RISC-V architecture. If ARM is laying off its staff then they are dying anyway. If not that it ultimately may be cheaper for Qualcom to buy ARM then negotiate with it. Benjamins speak to Softbank and if the economy goes south they may need the $'s.
33:45 - that's me!
This is maddingly frustrating. I *really* wanted an ARM processor for software development that could compete with Apple silicon and maybe even x86. Nuvia seemed to be leading to this. ARM is squandering an opportunity. This reminds me of AT&T and BSD
Same :-(
Hope things resolve
It not big issue it's only licensing issue and soon this issue solve by Qualcomm and Arm. Qualcomm Oryon processor come in 2023.
Burning through venture capital to develope and improve the ARM technology is not a sustainable business model. As we see with ARM, the VCs might get bored with spending a lot of money without much - if any - of a return.
Qualcom wants to pay as little as possible to ARM and buys a startup to get the IP they need cheaper than negotiating the licenses with ARM and any third party involved so they then can sell their chips to OEMs and make them pay for licenses. Good for Qualcom :D
ARM on the other hand wants to earn as much money as possible so the VCs and other shareholders get their RoI and have enough left to improve so they stay competitive. Good for ARM :D
Not too long ago, maybe 10 or 15 years, we expected new and exiting hardware out of China based on MIPS and/or OpenSPARC. Turns out IP alone does not make a chip, you need more than that. That's where ARMS support for developers comes in and this support has to be payed for , too.
ARM is just pissed that they won't be making money off their pre-synthesized cores anymore
I think that it is there. You just need to assemble your system. Gigabyte has a dual core motherboard for Amphere processors. You put two 128 core Amphere Altra Max CPUs into the board. Plug a PCIe graphic display card. Plug another PCIe SSD. Plug 16 pieces of DIMM may be 512G RAM. No Apple silicon can match your development machine. You can have it in a 2U setup.
Great breakdown of the case. I was waiting for RISC-V to be mentioned at the end. If anything this might push future chip startups to lean in that direction. If ARM wins, the Nuvia purchase would most likely have a negative impact on any future startup looking to do something similar to Nuvia. At least with RISC-V there's a much higher chance of being aquired if they can help a chip maker be less reliant on ARM.
Probably not. RISC-V is so modular to the extent that binaries are no longer portable between different chips because they each support slightly different subsets of the RISC-V ISA.
The ARM CEO mentioned this himself in an interview IIRC.
@@afc8981Not really. Ofcourse the ARM CEO would say that.
@@afc8981 nope, for os capable cores there is standard list of extensions (imafdc), and every riscv linux distro assumes that
From what I heard, one of the issues with RISC-V is that while the architecture is open, alot of the extensions are closed source (or paid, one of those)
@@albertsun3393 everything that linux or other OSes needs are open
Simd and visualization extensions open too.
So, everything that are mature and needed by most users eventually will be turned into standard extension.
For experimental staff riscv has a lot of places(instruction encodings, mmu types, internal interrupt numbers) that marked "for vendor use"
Main goal of riscv is to give vendors linux capable system that they can add problem specific parts(instructions, interrupts...)
Funny how Qualcomm - the IP and licensing stickler - doesn't understand ARM's licensing terms.
Qualcomm's been IP extorting the handset business for years with their modems and non-FRAND licensing. You'd think they'd be smart enough to just cough up the cash they way they expect everyone else to do for their modems.
Correct. But Qualcomm has legal depth, deep pockets, and will not be moving. A reasonable Judge might say 3-5-10% is incidental and normal in the trade. OEM royalty -- forget it QC will not have a bar of it. ARM has also failed to improve faster relative to others- it just keeps adding fluff or blocks that others do better. The shrink to circular gates/wells is a gamechanger, and QC has the pockets to do a RISC5 port. ARM wants to move into the QC space, but it will have the edge over QC id they can favor sweet OEM deals. So will QC roll over and die, let ARM eat its lunch - I think not. NVIDIA uses R5 core-bits in its GPU's. An alternative is QC does not blink, air the ugly court case, then rabbit out of a hat, we gone R5, and the rest of ARM's customers move.
Arm this is brilliant, this is exactly what we need to get everyone to move to the open source RISC-V architecture. This is essentially AT&T vs BSD Unix all over again... hold my beer, I'm going to make some popcorn. 🍿
RISC-V is not open *source*, "just" open *specifications*. You'd still need cores, of which many are closed, and many come with royalties and licenses again. It's a slippery slope. :)
@@CyReVolt The patients for the original PowerPC architecture are already starting to expire, and whatever is still patient encumbered the OpenPOWER Foundation is trying to make available, for instance there are already two reference Power ISA v3.0 softcore implementations that are entirely open source, Libre-SOC and Microwatt. Nvidia and Mellanox are both founding members of the OpenPOWER Foundation, and since Nvidia was denied the ability to purchase Arm you can bet that they are going to double down on evangelizing the Power ISA. The OpenPOWER Foundation is now actually organized under the Linux Foundation. The goal is to create a new open platform akin to the original IBM compatible PC after Compaq reverse engineered IBM's BIOS. Both RISC-V and OpenPOWER are little endian RISC implementations with objectives that almost entirely overlap, so these two project need to come together to facilitate these shared common goals.
@@PixelPi thank you for the insight !
@@dinoscheidt Dido
@@PixelPi we need it behind a strong CopyLeft licence
Hey I know that guy! Fun times Ian! I cannot wait until you are back in town.
Appreciate the coverage and insights. Heard a little bit about the situation here and there, but never really looked into it.
I would put it this way: Arm PC is a Competitor to X86 AND keeps PC competitiv against (Arm)Mac so actually this lawsuit foremost is beneficial for Apple by delaying the maturity of serious ARM Desktop PCs, right?
Can't help but have the feeling that ARM is burrying itself with this... smells like adobe, Nintendo, OneWheel, BMW,....
"the old business models are more and more failing but this anti consumer-behavior is doomed to collapse hard and ugly..."
nostradamos ;)
I am kind of against ARM entering the PC market, especially desktops given that their SOC design limits upgradeability and repairability. But on thin & light laptops it can be quite beneficial especially for battery life.
And nothing of value is lost, instead the market might be forced to embrace RISC-V rather than ARM. And RISC-V follows a somewhat similar simple instructions approach but it an "open standard instruction set architecture" so companies can do whatever they want to do with it (and have been doing so already). It was starting to get competitive with ARM already, getting more resources behind it because ARM is digging its own grave just means faster progression for RISC-V.
@@DragonOfTheMortalKombat more than that, having no BIOS GUI (except some rare implementations of UEFI or petitboot) is extremely detrimental for many PC users as they will have to enter the deep water of U-boot just to change some boot options.
Having to play with device trees and having to maintain images for every single implementation of the same SOC is a lot of support work and that's one of the reason why a lot of devboards from the past decade turned to be quickly unusable (or unsafe to use) even with FOS operating systems.
I wouldn't hold my breath for arm pc/laptop. Not having standardized platform, means continued mess with bootloaders per device, firmware per device, and lack of open drivers, and a pile of e-waste no one supports because it is not making money.
@@DragonOfTheMortalKombat ARM is strictly an IP company. They're not entering any market. It's up to the companies that license the ARM architecture to create processors what they do with them.
New recording gear? The quality is great!
I was visiting Patrick and used his studio!
@@TechTechPotato Apple sued Gerard Williams, CEO of Nuvia. What happened to that lawsuit after Qualcomm acquired Nuvia?
@@TechTechPotato Ahhh gotcha!
I bet Qualcomm wasn't expecting that they'd be acquiring a lawsuit when they bought Nuvia. Yikes.
yes they were. when you buy something groundbreaking that can potentially disrupt the market (5x gains) you can bet the rest of the market would try to kill you. also literally.
Ironically, Apple sued Gerard Williams, CEO of Nuvia. What happened to that lawsuit after Qualcomm acquired Nuvia?
@@Xevos701 Everybody forgot about that I guess.
This doesn't have to do with Nuvia. It has to do with Qualcomm themselves. Nuvia didn't do anything wrong.
@@Xevos701 imagination should sue apple
ARM is poisoning themselves with this lawsuit and my guess is that this is what Qualcomm is banking on. There's a middle ground which should be reached and I would hope that is what yes achieved. If I were negotiating terms as ARM, I would provision that any ALA contracts give joint custody of the generated IP which could be relicensed as a TLA design. ARM designs would continue to improve and acquire more capabilities and the TLA licenses would provide a mass market solution. The companies which built their systems using ALA would have a first to market advantage and the entire ecosystem could mature and become more power per instruction efficient. Long term we just to see advances which create more powerful and efficient designs. As the licensor, ARM still has a path to profit by encouraging a higher adoption of their cores. They are biting the hands which feed them if they win this suit.
s/yes/gets/
Now here's my question:
what *exactly* does the Qualcomm ALA say? Specifically, does it say "code must be Qualcomm IP only" or "invalid in the event of a merger or acquisition" or anything to that effect?
20:30 From what it sounds like, it absolutely seems to me like Arm is poisoning the well. If I had some influence in a company coming up with a long-term roadmap for some SoCs, this behavior from Arm would absolutely have me making contingencies to pivot to RISC-V if I wasn't already.
Thanks for the fair and level headed coverage and bringing in someone with a law background for stove well thought out commentary
So how did this resolved now.. after a year later? I am aware of the new issues they have, but was this one resolved or not?
Qualcomm is so stupid. They had a World Class CPU Design team... That was beating ARM designs for years.. But all got fired after their Server CPU didn't sell... Now they spend Billions on Nuvia and possibly can release nothing 😅
Just imagine how Nvidia would have leveraged the ARM IP in the courts
Nvidia owning ARM is the worst time line
right on to a shelf so they have even less competition
So the TLDR is Qualcomm didn't get the appropriate license that covered everything what Nuvia dose before merging the two companies, and now trying to use their own incomplete license which had lower royalty rate on Nuvia's design. Of course ARM will sue them, especially now ARM are starved of cash after acquisition deal with NVIDIA was killed partially due to opposition from Qualcomm.
I noticed this background and recognized it and for a moment I thought I was going crazy and probably misremembered STH's Patrick new desk.
IDK - anyway I look at this it looks bad for ARM. If they win, they will have such leverage over their customers that no-one will want to make new business with them. And no new companies will start. If they lose - well it is bad for ARM, because money. And they might be more cagey about licensing stuff. Damn Softbank.
It is going to pop up RISCV. It would do more if it was more mature, but it will work in its favor regardless of what happens now.
Sure looks pretty much like it
My brain was literally spinning.
This was a true nerd fest with Mr. Geerlings cameo at the end 😁
So essentially Qualcomm found a loophole and ARM doesn't like it so since we're both right let's go to court and find out who's more right. 😋
well if it's pay to win, qualcomm wins lol
Sounds like ARM just wants to double dip.
Regardless of the outcome this absolutely poisons the well. It shows no matter how much work gets done on top of ARM's IP that ARM still views it as theirs and they'll reneg whenever they want once your work becomes more valuable than it was when previously negotiated.
That's like selling your house to a 2nd party then suing a 3rd party when they rebuy that house for a higher price after the 2nd party put a bunch of work into it. Arm's been spending too much time with Apple and wants their cut every step down the line.
Not sure how I feel about this, besides RISC-V getting more traction. Being an OpenSource guy, I think it would be nice to see so more open cpu efforts, without ending up with "too many options to choose from" side-effects.
Whether cores are open is up to their owners though, RISC-V being just specifications. Vendors may still charge royalties and license fees, and SiFive sponsoring *this* video is... telling, huh? 🙃
@@CyReVolt Difference is that you are free to design your own cpu. you don't have to pay arm royalties for using the architecture nor an additional tax for creating your own core. It would be no different than manufacturers paying Qualcomm to buy snapdragon chips. And if you want to use snapdragon SoC and modify the core, you should rightfully pay them for it. The benefit is that it cuts out the middleman from charging royalties to qualcomm, and their customers for using that IP.
Think of it like this, imagine if x86 became a free and open architecture for anyone to design their own cores. it would be fair to allow nvidia to design their own cpu using x86, but not fair to let them use existing Ryzen or Intel cores to make their cpu, unless they paid for that IP, otherwise, nobody would want to foot the bill of development
Hi Ian,
I'm surprised that you didn't mention the new Board members of Arm and what impact that may have. Adding Dr Paul Jacobs, the son of the legendary Dr Irwin Jacobs - (both ex Qualcomm CEO), is a smart move. Also adding Rosemary Schooler brings some Intel sales knowledge too. Add that to the appointment of the Apple designer Tony Fadell and it seems that the US team is gaining strength to deal with the Qualcomm issue.
The tough year ahead for tech will hopefully force them to concentrate engineering effort on the future recovery and not get too embroiled in enriching lawyers who appear to have err'd on both sides.
I'm no fan of Arm mgmt but prefer the semi leaders to recognise how much collaboration is necessary between them.
This was filmed before the new members were announced.
ARM is just a community, kick QUALCOMM out of the community please!
let them develop their own chips!
I am a professional digital IC Designer that has worked for the likes of e.g. Intel, and when it comes to companies you do not wish to work with, it's exactly 2 names: Qualcomm and ARM.
Besides the fact that I think that ARM IP isn't even that great (the RTL code does not impress me much tbh), their licensing ideas become more and more ridiculous.
Qualcomm on the other hand by now is seemingly more a law firm than a chip company, which is a shame. Because their silicon is usually pretty advanced and great, but the way
Qualcomm treats basically anyone outside of their own walls is a 100% model of how NOT do do it.
I seriously hope RISC-V rises to a point where ARM becomes worthless or at least so pressured that they come back to earth with what they are demanding.
They are trying to milk this instruction set to the limit, but frankly, competation is on the rise. So especially alianating your architecture license customers like that will backfire big time.
It just takes Qualcomm, Apple or even Intel/AMD to really embrace RISC-V and ARM is dead. And being an extendable RISC-architecture, it is really mostly about re-engineering the decoder.
Many intstructions are similar enough to turn a very efficient ARM design into a RISC-V design.
Great video with a unique inside view of how ARM's IP business works! Don't most, if not all, IP agreements contain very specific wording on what happens when a licensee gets acquired?
Is it possible for me to design a manycore GPU on sifive since I want to get into HLSL and software rendering
26:20 but what about greggs? some places in the UK you dont even need to get to the other end of the street to find another
I don't think Patrick knows about Greggs. Next time he comes to the UK I'll take him
Wait a minute! this background looks familiar, are you guys collaborating with Patrick?
@28:12 - That support was paid for by Nuvia; ARM can't take that back or ask for another payment for said support.
I'm very bullish on Arm's architecture, their IPC and perf/watt is just insane given the tiny caches. But this case just doesn't seem right, Arm should not be allowed any claim to Nuvia's IP. It reminds me of Oracle vs Google/Android except ISA compatibility instead of API compatibility. I hope Arm's case gets thrown out quickly
ARM has a huge advantage over AMD and Intel in those regards to power. A customer using their IP can say "OK folks recompile your sh%t for 64-bit only we're throwing away legacy". Tell that to the Windows crowd.
@@ColdPotatoanything that's not x86 has that advantage - of course forgot ibm mainframes, yeah they don't have it either
@@aravindpallippara1577 Not so sure, modern x86 itself just breaks down into smaller instructions/uops so it's not like it used to be.
@@ColdPotato It's not about perf. x86 has no issue with that. But all the legacy (or not that legacy, avx512 hint hint) crap, just adds a big constant overhead. And this gets really noticeable at small wattages, because this "legacy" becomes a significant chunk of power wastage the lower you go. I.e. the overhead of instruction decoder might be unnoticeable at 4GHz+ with 32+ high performance cores running simd stuff, but once you try to go low-power it starts to matter.
They’re not claiming Nuvia’s IP, they’re saying that the terms of the license was broken
Seems pretty even overall... hopefully the case gets settled quickly, with Qualcomm and ARM agreeing to some moderately priced, additional license agreements.
I love the origin of ARM, I like the history of it. I dislike the company that purchased it and wants to put a dark cloud over itself. Requesting that qualcomm destroy all of it's technology relating to nuvia, most likely either leading to a strong arm in licensing renewal that hurts everything down the line or a complete destruction of this technology. It's looking like it's a maximize profits at all costs and or destroy everything in your way to achieve that goal. I would love if this accelerates risc-v and it becomes the defacto standard overtaking arm. Risc has proven itself time and again, so it's a good technology that should be pursued and not have a ruthless company that will hurt everyone to get what it wants. I know licenses can be tricky and companies want to get their fair share, but this scorched earth approach is insidious and i'm glad people are taking note and it's making people wary of wanting to continue with arm.
This is down to greed and strategy by both parties. It’s not likely either party will benefit from this conflict.
If they can pull off good performance and power efficiency, RISC-V chips could become "the new ARM" in the end, in being the most-used chip solution, which would be great! =D
but unfortunately risc-v is 5 years behind arm
The newer the ISA the better. You can argue that MIPS is decades ahead of ARM but what does that age advantage get you?
@@greecemobile7610 For now. There's a bunch of money getting poured into RISC-V right now, but it's not at the same level that, say, Intel, Apple, and AMD spend on their chip architecture RND. Honestly, I think that what it will take is for one of the big players to get burned by ARM badly enough to start pumping some serious fuck you money and resources into RISC-V to try and use it as a viable alternative architecture for _something_ at a large scale.
@@Kevin-jb2pv exactly
Arm architecture is definitely the future for the coming decade. Qualcomm pushing Nuvia to go desktop is the logical next step, simply because of the power to performance ratio that x86 just hasn't been able to demonstrate. Arm sees the opportunity that Qualcomm+Nuvia unlock and wants to dip their hand into the honey jar, Pooh style. As far as arm servers are concerned, the future to me seems to be pointing in the direction that Graviton is going imho.
I haven't read the brief, but it seems like ARM simply want to force Qualcomm to renegotiate. The fact that they're continuing to support them (and that this is "orthogonal to the case") hints at this, although their support may be contractually obliged. I can imagine they had a chat beforehand and said "Look, sorry fellas but we need to sue you so we can get a bit more money out of this deal you've pivoted on." I expect Qualcomm will renegotiate for something costing somewhere between Nuvia's agreement and their own current agreement(s). Basically, it seems like a strong-arm negotiation tactic. (No pun intended!)
I read what legal filings that are on line. Nowhere did ARM state in their deal with Nuvia that if Nuvia is acquired that the deal they made with ARM is off. Both sides have weaknesses in their cases. Destroying 3 years of development is not going to happen. The Judge will send it to mediation and eventually Qualcomm will give ARM some money to go away.
This kind of situations will simply lead the future for hardware to be the same as has been for software lately, more open source standards. RISC-V stands to gain a lot from this license and patent fatigue that is sure to grow with such moves by ARM and others.
It feels like any new semiconductor startup will avoid using ARM as a base. If there's any sort of new startup here, they will focus on RISC-V. I'd say any new "NUVIA" type of startup that is doing anything big will avoid using ARM to avoid these types of lawsuits. ARM has already poisoned the well.
Great Summary. Love the Server at Home addition form legal angle.
Oh, I didn't even know you were a youtuber(until now), I thought you were just one of Leonard French's Patrons.
Any update on this? It's relatively quiet.
I'd be concerned that Arm's behaviour might be anti-competitive as well. Are they essentially saying that they can decide that Qualcomm cannot sell this kind of product.
It shouldn't have got to court, ultimately it's going to hurt Arm and push more companies towards RISC-V (in the longer term). Either way, the IP should not be destroyed, human effort shouldn't be wasted in that way just because Arm has a bad take.
RISC-V on Windows PC desktop is inconceivable for the time being and a long time yet, ARM was hard enough. And at least ARM has the advantage of specs being rich and fixed in place, which is necessary when you have binary compatibility to deal with.
But as far as server... well that looks like the beginning of the end for ARM. There won't be much visible for the next few years, but under the sheets, things would be brewing.
Also interesting how ARM started with microcontrollers essentially... and is now rapidly losing them to RISC-V.
Could you compare the technology and licensing of these chips with Graviton?
I'm rooting for QCOM, I would love to see some competition in the processor space. however, QCOM needs to adopt a more intel like model where they don't force customers to use QCOM power devices. their power portfolio is weak. Battery chargers suck, PMICs are inefficient, they need better supporting devices to succeed in the long run.
Nice to see a collab between you two 👍
Wow… Patrick, Ian, and Jeff in the same room! I’d be star struck.
like mommy and daddy are fighting but dont worry guys arm and qcom still love you. lol
the real reason is apple doesnt want qcom to take their chip devs. they want to cost qcom more money.
I think Ampere is in a much better position. If they wait, they can use the time to make their systems better. If they make a deal, they can avoid all the pitfalls that Qualcomm created.
I see you were filming at STH studio 👍 Edit: never mind didn't wait to end of introduction
From the headlines I thought it was likely that ARM gave Nuvia a favorable licensing (or whatever) deal. As Nuvia, while having some clout, personal and funding wise were trying to get into the Server space. Something a lot more valuable to ARM than whatever royalties they’d get from Nuvia in a few years.
From the video it’s sounding that’s not the case or even something kinda opposite is. I’m not read up enough on this case to have a solid opinion but I’m getting Oracle vibes…
Good luck to the judge, because holy hell what a mess this situation is.
Windows doesn’t have a place where to put libraries so it’s actually not there fault and if your emu can’t deal with different paths you should reach reconsider how you designed it.
Does Qualcomms translation layer work on Linux?
ARM standing in its own way is never going to change.
Nope, ARM is only a community of partners.
Just kick QUALCOMM out, let them develop their own processor !
@@lucasrem en.wikipedia.org/wiki/Arm_(company) Here you go
Is this your new studio? Looks nice! :) Ah, it's Patrick's... Anyway, thank you for the in-depth explanation. Doesn't this cause any problems for an ongoing lawsuit?
why you care about a studio?
you rent a studio, or make content at home ?
lawsuit ? Studio???
It seems that Qualcomm have stuck their heads in the sand with the ARM licensing. They've just done the "it'll be alright" move and probably them and the Nuvia people have "group-thinked" their way to believing that it will. The McDonald's comparison is valid. That said, it shouldn't be difficult to resolve and they should do so for all concerned - except, that is, for the lawyers involved.
Long comment warning! This is interesting (to me at least) because I've often wondered how ARM's ... unique? ... business model will impact the acquisition of startups, as it's quite well known that in business the chaos of little companies is (for better or worse ... mostly worse) temporary, and in the end big fish eat little fish and bigger fish eat big fish until the mammoth sharks dominate E V E R Y T H I N G. Whereas ARM's licensing approach ... doesn't really fit these seas. So this lawsuit ... on the one hand ARM isn't being a bad guy for litigation in that they must protect the integrity of their licensing that is their business model. On the other hand, it is most definitely shooting themselves in the foot because, as said, it's a business model that doesn't really fit these seas, and the more they defend it with loaded arms, the more they scare their customers. If ARM isn't stupid (which, again, they developed a business model that doesn't fit their ambitions, so this is highly questionable) ARM will eschew strict license adherence and find a way to amicably forgive, grandfather in, or renegotiate the licenses in a way that Qualcom "wins", BEFORE this actually gets too far into trial, and the litigation gets dropped. This way they don't scare their customers and show future customers that ARM isn't a complete pushover and will defend its business, but can also be a good guy about tricky situations. That ARM is only scary when rampant abuse is involved, and as this isn't a case of rampant abuse, it can come to a happy ending that isn't scary. However, as ARM's business acumen is ... questionable ... I wouldn't place bets on this being the outcome. If ARM pushes this, right or wrong, in the end they're going to lose money, even if the win the award of being a Technically Correct Pedant. Because RISC-V is showing that ARM isn't the only way, and the last thing that ARM needs is to scare its customers into joining hands to pave the road for RISC-V. But even if that's what happens, that's ARM's problem. Whether ARM or RISC is better for consumers is a different topic entirely. LOL But if *I* were ARM, I would not only be looking to Good Guy in a new licensing agreement for Qualcom that both sides can live with, but I'd be re-examining and iteratively improving my licensing business model to be flexible and friendly to reduce the RISC of this ARMs race.
What a shame it would be if Qualcomm is made to delete the Nuvia core… I hope it does not come to that, that would make zero sense for both sides…
Most likely they will reach an agreement outside of the court.
Before the lawsuit, Arm had already tried to negotiate with Qualcomm and they had not reached a compromise
@@kevikiru usually after a lawsuit is filed, both parties are more inclined to reach a settlement. Both have things to lose if the lawsuit will follow all the way through.
At 16:38 did you mean to say Apple instead of Arm?
I guess the next logical step for Qualcomm (and many other companies) is RISC-V. ARM proved it is not impossible to introduce brand new hardware platform for mobile, consumer and server market.
I think you're right it's about money, but also there's a strategy of how much money can be made if the soc became commonly used in Windows.
👍 Have you any reviews/videos on Tachyum as a player in market for server chips? Cheers.
ua-cam.com/video/siN1eQ0SfxU/v-deo.html
3:00 5x gains, yeah, nobody is gonna allow them to disrupt current market that much. thats gonna be a hard fight to win.
ps we live in the world where 5-15% "gains" in performance per year made "the norm" by tech giants, to milk the market and spend as low as possible on R&D. slow progress = less price drop on old products = bigger margins = more profits. fast progress is very bad for business of milking people's wallets.
Dennard scaling ended years ago, new processes cost far more to develop and only logic scales now, not cache memory or io, while higher frequency with power efficiency needs smart design.
No need for conspiracy theories, in the past most performance gains came from process improvement.
@@RobBCactive thats not a conspiracy theory.
Lets take Nvidya/AMooDe. Nvidya makes GaaS = wants PC gaming dead. AMooDe makes PS5-Xbox = wants PC gaming dead.
result? = overpriced af GPUs = PC gaming essentially IS dead at this point, nobody can afford current gen gaming, so game devs make games for obsolete hardware.
AMooDe has Vcache tech to kill Intel yet chose to release 7000 CPUs overpriced aaaand without 3D Vcache to not compete much and to not drop prices much on older CPUs it still makes. with GPU prices we all are familiar.
+ a Steam Deck exists with old AMooDe tech running AAA games for less then 600 bucks, meanwhile smartphone SoCs can run only 10 y.o. graphics and stupid people are still buying that morally obsolete crap for 1000 bucks+.
Now imagine a Steam Deck with latest AMD hardware, Zen4, 3D Vcache, RDNA3, all that tech already exists and the only reason we don't see it is such a device would literally kill ps5/xbox market, which is a market AMooDe sells its chips to.
And imagine it in a smartphone form factor. it would kill everything on a smartphone market. AAA gaming instead of 10 y.o. morally obsolete crap in a pocket. it also would severely damage the PC market because why buy PCs if you can have everything in your pocket. in 1 device. no need for gaming PCs unless they start to offer truly next-gen 5x performance + photorealistic graphics - all of that.
That is what AMooDe sits on pretty much right now. And thats facts, not a a conspiracy theory.
@@rawdez_.... Valve is building a steam deck 2 with amd's new apu design
Hell amd has been building up the dragon range series of apus with explicit intention of making power efficient low cost gaming (extremely good gpu systems)
And you do realise that ps5 and xbox consoles are usually sold at cost or a loss and are made up through game sales? They will always cost less than a pc due to there being less focused on upfront profit
Not to mention gpu crisis was almost entirely due to the mining boom - 4080 just failed in it's launch to the point Nvidia is considering a price cut for it.
If anything amd is bending over backwards in every market to gain market share - and yes they make far more selling to servers and pc crowd than consoles because of the bulk agreements
You also forgot intel in there - why are they with holding their tech if it's all a conspiracy? They literally can't profit off console gaming
@@rawdez_this is literally one of the wildest theories i have heard till late
I would rather believe gpu crisis would be because of nvidia and amd playing the market for everything it's worth (which while isn't intentional but what's happened)
You really have to re think this lad
@@aravindpallippara1577 >Valve is building
we need smartphone makers building smartphones with AMD tech, not Valve building another gaming device (which is cool but not gonna help much by itself).
Unless Valve make it in a smartphone form factor (e.g. with detachable controls) + with phone functionality and release it in 2023.
also I really doubt AMD would allow anyone use all what they have - 3d-vcache and RDNA3 in a not heavily cut-down version just because they can't allow something better than a ps5/xbox on the market.
>crisis was almost entirely due to the mining boom
you look so big yet still believe in marketing fairy tales when in fact AMD/NVIDIA were way too happy to make billions on that. who are really to blame miners or those who sold overpriced pickaxes and didn't want to drop prices when mining has died?
>forgot intel in there - why are they with holding their tech
what tech? they get to keep selling their overheating overpriced CPUs and slow overpriced GPUs with bad drivers, lol.
ps at this point we need Qualcomm and their 5x perf-per-watt improvement but with RDNA3-level or better graphics. Adreno is a trash-level GPUs. and with a price tag thats not even close to 1000 bucks but way lower. now THAT would be a real progress.
I don't understand why the FTC doesn't step in at least within any division of ARM based in America... They are a monopoly on reduced instruction sets with the only alternative being RISC-V.
Companies using arm were already looking at risc V. This surely only serves to accelerate that effort in all companies. No surprise sifive are sponsoring, they must be on an upswing for every arm lawsuit
The thing about poisoning the well is that ARM is suing on of its customers which is generally a bad thing even if there are legal reasons for doing so in this case. I do think this move was a to force a negotiating position but such business deals are always best behind closed doors as the ugliness of arguments is kept from the public eye. I would argue that the worst case and something will likely happen to some degree going forward is a modification of how ARM licenses its IP. Worst case is that ARM stops offering new ALA agreements and increases pricing on their TLA agreements.
The other wild card in this disagreement is that ARM too was attempting to be acquired by nVidia. Discovery in the ARM vs. Qualcomm disagreement has the potential to reach into various documents exchanged between ARM and nVidia as part of their failed transaction.
If that is true and ARM can claim IP through the support it gave a company that is licensing it's tech, then that can prove to be a very dangerous precedent for all support contracts in R&D.
ARM is essentially claiming that they own all of Apples CPU IP because Apple licensed ARM.
It's actually pretty simple if Nuvia did have some closure in theirs lincese contract, that nobody except Nuvia have right to use the tech of ARM, then 100% Qualcomm will get a problem, but when the license agreement is not transparent, then Qualcomm always can say that they as owners of Nuvia have full right and all IPs which belonged to Nuvia belong to Qualcomm.
ARM for Server? Did Huawei already have that?
Huawei, Amazon, Ampere, Baidu, Marvell.
Could the required support from ARM be like a software key to enable certain key features of the core design into each chip?🤔
I read the ARM account fillings in UK, they protect themselves and their tools by being online only, and that it's not given like a pice of software. So I think they could turn off the access for this pretty easily.
@@michaeljarcher Basically a software lock. And what about the ARM rogue subsidiary in China? Could Qualcomm turn to them to license the missing pieces of their design?
At 11:27 do you mean Apple and Nvidia?
I have to admit that I am confused as to why Qualcomm didn't just licence ARM v9 in the first place since they already have an experienced team of engineers in house. That would have saved them from all this legal messing around and would probably have cost a whole lot less than buying out a company (let alone the legal fees).
In the video, I explain that you don't simply license a version of Arm. You license a version for a particular market.
Qualcomm’s, and ARM’s, engineers was not able to keep up with Apple.
QCOM has been exclusively working with MSFT for four years on Windows on ARM. It isn’t completed. It seems neither side is serious on this effort. The exclusivity ends soon that opens the door to more vendors to jump in.
the nuvia acquisition sounds like a great idea for me to run as a linux desktop
This is such a rookie M&A mistake, I'm almost shocked. Almost because I've seen this level of legal incompetence a couple times at major tech firms in the last couple years. I'd love to know if this was Arm's in house council or outside council that 'oofed'. It's not even legal's fault as this kind of thing is covered in most MBA/CFA programs.
When VIA purchased Cyrix and Centaurthey were allowed to make x86 chips because of Cyrix's and Centaurs preexisting rights. AFAIK Intel didn't make them sign a new agreement.
What's this news that Charlie from SemiAnalysis is hinting at ? Behind a paywall it is unfortunately but seemingly it's hinted that Qualcomm just shut down Nuvia ?? Is that true ?
ARM are shooting themselves in the foot here in my opinion, if you have an existing business relationship - settle it without throwing around lawsuits - especially if said partner is already licensed to the gills.
Unfortunately I think this just comes down to new management looking to get some quick "wins" via brute force.
EDIT: Also, if ARM are going to argue that IP developed by Nuvia with ARM licensing was not actually owned in whole by Nuvia, then why did not ARM not step in during the acquisition?
I would love to see a powerful ARM powered laptop or desktop. We already have server arm cpus so i understand qualcomm's direction
Given what TTP said about the extreme conditionality of ARM licenses, I don't think ARM themselves wants to compete in that space - since it would break their "pay per market" business model.
@@SuperSmashDolls There are apple cpus, but since they are extremely expensive and apple they don't count
Qualcomm doesn't stand a chance, companies have tried this tactic of buying out others for the arm license and have failed. Its one of the reasons why all the big manufactures are desperate to buy arm. Rather than qualcomm trying to fight it they need to ask ARM how much for the correct IP and give them money for it.
Lol, those multi-NIC miniPCs in the background did make me think of him...
I bet this lawsuit will increase interest for RISC-V R&D.
Having SiFive as a sponsor of this is great. I hope they catch up to ARM quickly
What court is this brought up in?
It's hysterical for me that ARM would require to destroy knowledge. How did it happen that they are opposed to development of technology and not just to how this technology is sold?
Why can't Nuvia get knowledge about ARM cores and use it however they want? Isn't it ARM's problem that they did not foresee Nuvia being bought by someone?
24:00 I wonder if that restriction of target market only applies to the end product. Would be funny if ARM allowed to develop knowledge of producing CPUs only if that knowledge is used only for specific market segment. I'd imagine there are a lot of generic problems in CPU development if not most of them.
Would it be good idea for Qualcomm to demonstrate that all of the important knowledge which Qualcomm acquired is generic and should not get restricted by ARM? I'm guessing any "verified" or "validated" designs which Qualcomm got are not critical for the end product they want to make (except some additional costs), this is what I would consider a non-generic knowledge which ARM could be able to restrict.
Have you ever heard of IP?
@@johnm91326 have you ever read my post til the end?
Apple hinted that they would use "Apple Silicon" in their server side components of "Apple Intelligence". Does this mean that they now have an ARM server license? If not, will ARM go after them like they are going after Qualcomm? Something tells me they wont.
Rene Haas is close with Apple.
You have upgraded your setup and quality substantially or I have missed a few videos
I'm in ServeTheHome's studio!
Would this lead to companies looking more closely at RISC-V?
I think so
Big companies, not yet. Startups absolutely. I feel like ARM will lose a lot of startups licenses.
To those engineer and creative individual, it must have been a headache. Their hard work become investor property and they only received a monthly salary + they are subject to be fired(but they contributed more than the investor).
Arm will never invent since they play this royalties business scheme and if they did, 10 years already passed and we still looking at an arm architecture with a miniscule improvement instead of inventing new way, just like INTEL.
To be honest, I think ARM will win.
Quite simply, ARM created a package deal for Nuvia to build a specific processor for a specific market and this deal was made and agreed to in writing.
If there are NO provisions in the agreement for in the event Nuvia got sold to someone else(Qualcomm) and that someone else automatically got the same package as if they were Nuvia, then the deal is void.
That is the way the judge will see it.
Its Qualcomms fault for not negotiating with ARM before purchasing Nuvia.
That is also the way the judge will see it.
Fantastic breakdown and explanation
I feel like ARM would have been better off being ok with the acquisition. Qualcom could have scaled and sold units. Obviously their agreement reflected the scalability. I'm sure ARM would have preferred Nuvia to scale at the higher royalty.
From what I've read & You say here Qualcoms Money people have gone through the books of Nuvia said whow that's a big licence fee going to ARM we have ARM licenses and then gone to the league team said if we don't negotiate a licence with ARM will our current Licences cover us using ip league team said 90-98% so money team then decided to save money and not purchase the necessary licences on the grounds ARM get lots if cash from Qualcom & A court case would loom bad.
ARM intern said fine not willing to talk we will take you to court maybe that will get you to talk to us and that's where we are now.
The only winners from this will loyers.
Really good video. This analysis is great
I do wonder if Softbank will kill ARM. If their architecture/IP is so expensive and complex to maintain perhaps it is cheaper in the long term to think outside the box and evolve the open source RISC-V architecture. If ARM is laying off its staff then they are dying anyway. If not that it ultimately may be cheaper for Qualcom to buy ARM then negotiate with it. Benjamins speak to Softbank and if the economy goes south they may need the $'s.
You're looking well. Good to see 👍
I wonder if I could build a company that makes RISC-V processors customized for your desire.