If Nvidia wanted they could enter the X86 CPU market. They don't have a license for it like Intel or AMD. However Via does have a X86-64 license. So why don't they buy Via or license their X86 license?Having 3 players in the X86 CPU market would be great. Rather than just Intel or AMD.
Why risking going into a new market when your main one generates tons of profit
>So why don't they buy Via or license their X86 license?They tried buying ARM and got paused by the UK government. Also, I'm pretty sure the ownership of an X86 license is not transitive precisely to prevent these type of retarded anti-competitive acquisitions.
>>107360898x86 is dead tech, arm is where it's at
>>107361301Yeah but if Nvidia bought ARM they would have control of the whole ARM CPU market so I understand that. But right now all Via does is sublicense their X86 license to Zhaoxin so Chinese can make their own CPU's. If X86 license is not transitive they could go the same route as Zhaoxin and sublicense it from Via.
>>107360898Nothing special about X86. Also they are doing fine, why would they expand into a different field?
CPUs stopped scaling. They should have done it 20 years and and it's too late now.
>>107360898Why?like seriously: why?Also the US gov basically owns nvidia now and Isreal (aka intel) owns the US gov so why would they buy themselves?
>>107361346They tried to make one in 2011 but due to licensing ended up making ARM chips instead. My guess is that Via wasn't ready to make a joint venture with them back then since they were just launching their first ever dual core CPU the Nano X2 and probably had some hopes for it. In 2013 Via partnered with Zhaoxin so if Nvidia was developing their CPU's in 2013 they may have actually made an X86 CPU.
Isn't there enough tel aviv in intelaviv for you already, do you want backdoors for your backdoors
>>107361324arm is slow as fuck
>>107360898It would be way better for them to make Risc5 CPUThat would allow them to make low consumption hardware.
VIA probably has licensing terms such as "If you get bought you lose the license"
>>107362728I mean VIA got the X86 license by buying up Cyrix. Unless Intel then updated their license once VIA bought them. They could at least license the X86 license from VIA like Zhaoxin has done.
>>107361904>source: I made it up
>>107361346>CPUs stopped scalingso they can easily catch up then
>>107361324This has been said for the last 20 years yet almost all windows pc's are still X86.
>>107365650Windows on Arm is unironcally pretty good nowadays, though. 90% of anything you might need is native and the translation layer werks well enough for the rest. Arm Macs were the beginning of the end for x86 as a mainstream platform, once Apple does something successfully everyone else is bound to follow.
>>107360924/thread
>>107360898There are three players, VIA still makes x86 hardware
>>107361324Nope.x86 is two vendors that essentially have the same instructions, so it's almost universal as a target.arm by design is hundreds and hundreds of variations.x86 will be there forever, as the fastest CPU platform.
>>107362755More like source: every benchmark
>>107360898Do they have any reason to do that? They're already making shitloads of money and joining the CPU market would probably be very difficult.
nvidia should just use all their AI power to reverse engineer it
>>107360898How customized are the ARM CPUs used in their Jetson lineup? Say, compared to Apple's M series?
>>107365720You're basically saying that the web browser is native and ARM Windows works great as a proverbial Facebook machine, right?The problem is that Windows in and of itself sucks major ass so if you're not getting the full compatibility, upgradability and performance of the x86 platform then there's no reason to put up with Windows. If you want to run ARM just use another OS.
>>107366289>If you want to run ARM just use another OS.And that's why people are flocking to MacOS. No other laptop has been able to compete with Macbooks since M1.
>>107366320I mean, I'm pretty sure there aren't any other major ARM laptops around. What else is there, the Windows Copilot garbage of course but that's using Windows. Is there any boutique ARM laptop out there running Linux by default?
>>107360898Nvidia is invested in the ARM ecosystem. They tried to buy ARM Holdings for $40 billion back in 2020.
>>107360898They tried to buy Via. It almost went through and then it was discovered buying them would not grant them usage of an x86 license. They are non-transferable. If you buy a company that owns it, that license is terminated and Intel retains control over it. You'd have to go back to Intel to get a new license. And Intel don't usually agree to that.https://www.intel.com/pressroom/archive/releases/2003/20030407corp.htm Even this new Intel Nvidia team up, Nvidia has no shot of getting that license.
>>107366336Not by default, but Linux can run on ARM Thinkpads well enough
>>107366376literally why is this fucked up duopoly allowed?
>>107361904here's apple's new M5 on top vs the 9950X3D below
>>107366376>If you buy a company that owns it, that license is terminated and Intel retains control over it.wtf kind of jew shit is this
>>107366450>>107366405You guy aren't familiar with Intel really was lol This is old school kikery
>>107366405x86 only got licensed to cyrix and AMD back in the 80s because IBM forces Intel to. Hell, the only reason AMD was able to make their own CPU was because AMD exploited a legal loophole where Intel couldn't copyright numbers (i386 formerly 80386).
>>107365720Then why do their laptops run at over 100°C on medium load?
use case for ARM on the desktop?
via's licence isnt transferable
>>107366450Bro let me introduce you to the company that locked you out of features your cpu had unless you bought a literal gift card
>>107366469Kek, I just found out it was Intel Haifa that brought us the cursed raptorlake generation. Israel directly contributed to the fall of intel mindshare
>>107360898How much does patents play a role here. It seems me like every semi complex industry is chock full of patent blocking and avenue for growth, or rather just keeps competition or open source 20 years behind all the time? Is there any truth to this or am I just a schizo?
>>107366593Patents play all of the role here
>>107366376How did Via then get that license since they got it when they bought Cyrix. Or is it that because of the settlement only Nvidia wouldn't get the license but some other company would.
>>107361324Shalom
>>107368463It's because Intel immediately changed how the licensing worked after failing to stop Via gaining it. Cyrix didn't have a license. Texas Instruments, which owned Cyrix had one. Cyrix was sold to National Semiconductor who also had one. When Via bought Cyrix they bought it to become a part of its CPU division and the original license Intel had with National Semiconductor was worded in a way that allowed its transfer if NatSemi sold its entire CPU division, which they did, and with that, the x86 license transferred. Intel tried to stop this, they claimed the license was non-transferable under any circumstances and sued but the way the original license with NatSemi was written they couldn't do it. NatSemi exited the x86 business entirely and VIA entered it. Once Intel saw they'd lose in court when the original NatSemi license was discovered they quickly settled. A loophole was exploited by Via and Intel closed by the time Nvidia came along. And then when Nvidia tried to get around it they sued them and banned them from ever obtaining an x86 license.
>>107369446>Muh loicense>muh patentsif Intel had their way and there were no other corporations with permission to make x86 chips, then we would still be stuck with dual core 2.4ghz chips running on 14nmthis is a symptom of the failure of capitalism.
>>107369446I guess this X86 ban also prevents Nvidia from going the Zhaoxin route. Basically have a joint venture where Via licenses their X86 license to them. And maybe even buy say 49% of Via. Not directly own it but have almost controlling share of the company. But with the X86 ban Intel put on Nvidia that is not going to happen.
>>107369506Yeah I'm shedding no tears at Intel's current state thanks to the actions they made before Ryzen. If Ryzen hadn't come out i7's would still be 4C8T and i5's 4C4T and i3's 2C4T. And non workstation laptop variants would all be 2C4T.
>>107369777To be fair to shittel, an undesired effect of their lazy and slow development is that if AMD didn't push the boundaries of performance, we could've stuck with Sandy Bridge for 2 or 3 decades without feeling any slowdown. But of course when CPU performance increases, software performance decreases. So the end result is that the performance of your PC stays the same and you just have to keep buying new CPUs.The only advantage comes when you write optimized software that fully takes advantage of these new powerful CPUs.
>>107360898via is still around?>license their licensei highly doubt intel allows sublicensing of the x86 architecture
>>107360898>However Via does have a X86-64 license. So why don't they buy Via or license their X86 license?If I remember correctly almost all x86 licenses have a clause that says they're not transferable if they're bought. The only reason VIA got theirs is because when they bought Cyrix they had to renegotiate the license with Intel because if Intel wanted to, they couldve denied the transfer.So if Nvidia bought VIA, Nvidia would likely need to negotiate with Intel in order to actually use that x86 license.
>>107369882They do allow. Right now there is a Chinese company called Zhaoxin that makes their own X86 CPU's. And they do it thanks to a joint venture with Via where they use Via's X86 license. They are still behind Intel and AMD. Supposedly their latest CPU's is almost as fast as a ..... i3 8100.
>>107360898>If Nvidia wanted they could enter the X86 CPU market.They could. Would it be wise, tho?>>107361324>x86 is dead tech, arm is where it's atARM has already lost it's microcontroller marketshare to RISC-V. RISC-V will be coming for it's microprocessor marketshare, next. Then x86_64's when it does the same shit at 1/3 the power...
>>107361346>CPUs stopped scaling.I think OpenPiton good for about ½billion cores, what you talking? You need more? Use another chip.
>>107369937China are making great headway with RISC-V...After being kicked in the dick with chips, they doubled down hard into making their own. x86_64 is dead, it's not getting any faster and it's not getting less power hungry... It won't compete.
>>107366019Every benchmark actually shows ARM is faster
>>107360924novidya could use their ai powers to design super duper ultra performance x86 cpu and btfo intlel and ayymd.
>>107366376Simple: VIA should acquire (N)VI(DI)A
>>107370023Reality: Workload dependantThat's mostly because these are not apples:apples things. They are different, they perform different. Generally the "win" for the x86_64 comes from extensions to the instruction sets, and essentially onchip ASIC's to offload the task. They can be added to ARM, and it'll bump up the power draw a little, but it *should* still beat x86_64 there, too. RISC-V however, has even more improvemnts to gain over ARM. And it's growing. There's already cheap SBC's you can run common *nix on. Shitty laptops will be next. Then not so shitty laptops as the shitty laptop grade crap replaces ARM in end user devices like TV's... Then it'll be starting on x86_64...
>>107370028>design super duper ultra performance x86 cpuWhat do you imagine to be the bottleneck that is constraining x86 performance, currently?
>>107370078the amount of coding socks
>>107370030this would be incredibly funny
x86-64 patents have already expired almost 6 years ago you retarded faggot.It's SSE4.3+ and AVX1/2/512 extensions that Intel and AMD are sitting on which VIA does not have.
>>107373543I mean even the latest VIA branded CPU (VIA Nano QuadCore C4650) supported AVX 1 and 2. And the latest Zhaoxin ones support those features as well. And there is no such feature as SSE 4.3 (and VIA cpu's support 4.2). Only AVX 512 isn't supported (but Intel doesn't support that either anymore)
>>107370078CBP-6 concluded this year after 9 years in the limbo and implementing TAGE-SC-L like prediction rate with Seznec's new TAGE-SC design became viable (and Koizumi et al made the amazing RUNLTS over TAGE-SC-L for chasing down the horizon, in efforts to make it practical).Also, with Triangel last year and Prophet this year, we're also seeing L1 temporal prefetchers get much better.I expect that will improve performance a bit even on the same nodes.>>107373641Meant 4.2, typo>I mean even the latest VIA branded CPU (VIA Nano QuadCore C4650) supported AVX 1 and 2Didn't know that. I guess they negotiated with Intel and AMD to be in on the loop?
>>107366490use case for the desktop?
>>107369862>The only advantage comes when you write optimized software that fully takes advantage of these new powerful CPUs.seems expensive. indians sound cheaper.
>>107369969>RISC-V will be coming for it's microprocessor marketshare, next.that won't happen unless apple adopts it, or someone similar. reason being is again because there is nothing standard in risc-v, especially no standard boot up in particular, so there is nothing to target. doesn't matter for microcontrollers, bit of a big deal for microprocessors. works for apple because they're a closed ecosystem run by a major. basically because risc-v is a hobby for academics who have no fucking idea, despite having the ibm/intel example right in front of their fucking faces.
>rust>linux>armtechnology for trannies
>>107373105For you.
I think moving to an open and free architecture would be good. It does seem like things are coming to a head. Hopefully everything shakes out nicely.
>>107361324I see more and more zoomers having this anti-consumer take...do you kids actually understand the current landscapes of both?I build yocto linux images at work, we have multiple arm boards and it's a fucking nightmare, >devicetrees are all different, even for a same gen SoC because it's customized by another vendor each time>not a single board work with upstream uboot, we have to deal with 5 differents one>no uefi so each one has it's own proprietary firmware that work differently for no fucking reason instead of just using uefi like a non-retarded vendor>no acpi>udev-hwdb that you have to fill/fix>want broadcom chip spec? pays up you fucking goy even if you just want to discover their offering>support for most SoC is under 5 years on average>serious vendors only have x86 offering for hardware designed to last 15 years with the supply chain that come with it>pci implementations are usually absolute dogshit>NDA NDA NDA NDA NDA>no spec for anything, reverse engineer everything yourself browhen I compare to uefi/x86 there is just no match>ONE image for all hardware from within the last 20 years>bios and uefi depending on hardware, it's all standardized, bootloader get the correct (hd0,gpt0) hardware location, etc, it just works>want to use grub? uboot? systemd? nothing? you do you pal, it just works anyways because the mbr/gpt header/parts are at the same location for every hardware in the last 40 years>devicetree? the fook is that pal?>want the spec? go to advantech or aaeon website and read up, everything about serial to cpu is documented for free, have questions? just call no NDA>acpi>follow udev-hwdb>you can still buy baytrail-era SoC just fine>serial is standard innodesk stuff you can buy on mouser or digikey with no retarded custom binary blobI really don't understand why the arm world is dogshit on purpose, it does not make any sense, why wouuld you want that too? it will make you life miserable, it's like android but worse.
>>107369997china's been trying to make their own for decades nowbefore risc-v it was mips
>>107375250the arm ecosystem is embarrassingly bad
>>107375250PC-compatible only became a standard by accident. it was just simple enough to clone and also backed by big enough names (IBM/Microsoft) to increase the desire to clone it and purchase clonesnobody consciously set out to standardise a computer and so it's really not as surprising that it hasn't happened with arm stuff.
>>107377282I've managed to get away from arm, I now focus 80% of the time on x86 and my life is so much better ever since>>107377376yeah but it does not really matter how we got there, it's rock-solid, we don't have a single reason to stray away except vendor lock-in, greed and planned obsolence.even if arm cpu btfo intel cpu it does not matter as the arm ecosystem is pure dogshit except if you're dealing with cortex-m0-4 and therefore aren't dealing with firmware/bootloader/os anyways.it's a shame that intel/amd aren't interested in building low-power high-performance chips, x86 and arm from a ISA and uops implementations POV are basically the same, it all come down to product segmentation, apple get all the praises for no reason when they just built a chip in a segment that no one was interested in before.>nobody consciously set out to standardise a computer and so it's really not as surprising that it hasn't happened with arm stuff.and it's a damn shame, the current arm landscape is very hostile towards users and developers unlike x86... the x86 duopoly is rather very healthy and both companies are working together to get shit down, see fred, avx10, chktag and AMX.
