Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Nvidia to Acquire Arm for $40B

nvidianews.nvidia.com

Post reply on HN
Print view View original

761–770 of 784 posts

Re: Nvidia to Acquire Arm for $40B

#761

Post by klelatti »

tibbydudeza wrote: post #749
smolder wrote: post #602

Earlier quoted context omitted.

AMDs answer to Ampere won't be shabby based on info about next generation consoles. They're also widening the gap on CPUs with Zen to where ARM won't have an easy time making inroads on server/workstation. On the Intel side, the process obstacles have been tragic, but they have plenty of hot products and plenty of x86 market share to lose, or in other words, plenty of time to recover CPU performance dominance.

As Linus said ARM will only be widely adopted if there is availability of affordable hardware that developers can actually buy. Apple has pulled this off about 4 times because of their small market share and willingness to deprecate old hardware, software and the close control of the hardware they release.| In the PC world - x86 will remain with us for a LONG time to come.

Which is why Arm in Macs is a crucial moment for the whole of the Arm ecosystem.

Re: Nvidia to Acquire Arm for $40B

#762

Post by dragontamer »

floatboth wrote: post #736
dragontamer wrote: post #303

Earlier quoted context omitted.

VLIW was the best implementation (20 years ago) of instruction level parallelism. But what have we learned in these past 20 years? * Computers will continue to become more parallel -- AMD Zen2 has 10 execution pipelines, supporting 4-way decode and 6-uop / clock tick dispatch per core, with somewhere close to 200 registers for renaming / reordering instructions. Future processors will be bigger and more parallel, Ice…

Main lesson: we failed to make all the software JIT-compiled or AOT-recompiled-on-boot or something, that would allow retargeting the optimizations for the new generation of a VLIW CPU. Barely anyone even tried. Well I guess in the early 2000s there was this vision that everything would be Java, which is JIT, but lol

Your point seems invalid, in the face of a large chunk of HPC (neural nets, matrix multiplication, etc. etc.) getting rewritten to support CUDA, which didn't even exist back when Itanium was announced.

VLIW is a compromise product: its more parallel than a traditional CPU, but less parallel than SIMD/GPUs.

And modern CPUs have incredibly powerful SIMD engines: AVX2 and AVX512 are extremely fast and parallel. There are compilers that auto-vectorize code, as well as dedicated languages (such as ipsc) which work for SIMD.

Encoders, decoders, raytracers, and more have been rewritten for Intel AVX2 SIMD instructions, and then re-rewritten for GPUs. The will to find faster execution has always existed, but unfortunately, Itanium failed to perform as well as its competition.

Re: Nvidia to Acquire Arm for $40B

#763

Post by deaddodo »

aasasd wrote: post #710
deaddodo wrote: post #677

Earlier quoted context omitted.

VIA isn't privy to the cross-licensing agreement Intel/AMD signed which gives them full access to the other's technology portfolio (since AMD was a second source foundry for Intel, while Cyrix/VIA was not). However, the Court ruled (when they were still Cyrix) that they are allowed to sell any clean-room designed x86 hardware. So while they don't have access to internal architectural documents, they're allowed to rev…

> they are allowed to sell any clean-room designed x86 hardware Uuuuh, presumably the case was about patents, right? I don't see how cleanroom-ing is fine with regard to patents.

Source: https://law.justia.com/cases/federal/district-courts/FSupp/8...

Summary: Cyrix (and subsequently, VIA) have an implied license to Intel's patents; allowing them to develop x86 hardware.

How that differs from AMD: AMD and Intel have a full cross-license on technologies. This means, AMD can utilize Intel resources to integrate AVX-512, a ring-bus, etc (and vice versa). VIA can not. They must develop those technologies independently, in a compatible manner.

Re: Nvidia to Acquire Arm for $40B

#764

Post by deaddodo »

tibbydudeza wrote: post #745
klelatti wrote: post #627

Earlier quoted context omitted.

Very briefly: - They don't license because they can make a lot more money manufacturing the chips themselves. - AMD also has the right to x86 because Intel originally allowed them to build x86 compatible chips (some customers insisted on a 'second source' for cpus) and following legal action and settlements between the two companies over the years there is now a comprehensive cross licensing agreement in place. [1] -…

Afaik they could implement the ISA but not the Intel interconnect bus so AMD licensed Hypertransport (as used in the Alpha) from DEC so you had different motherboards/chipsets.

All second source AMD (and other manufacturer) chips were pin-compatible with Intel chips. Compare the i386/am386, the i486/am486, etc. Even a few non-second source (e.g. the unique uArchs that came later) designs were still pin-compatible, such as the 5x86 and K6/K6-II/K6-III. In fact, the third parties were the first to break away from electrical compatibility with their Super Socket 7 motherboards.

Intel did change sockets as a means to disallow socket-compatibility; forcing consumers into their architecture if they bought their motherboard, but that had no effect on AMD's development. AMD had purchased a significant share of DEC's engineering portfolio and, along with it, their employees. Those employees then developed the K7 (Athlon) architecture around some of the Alpha's technological advantages, which included HyperTransport, a multi-issue FPU (fixing one of the major issues AMD had struggled with and bringing them ahead of Intel), etc.

Re: Nvidia to Acquire Arm for $40B

#765

Post by ineedasername »

nitrobeast wrote: post #683
ineedasername wrote: post #669

Earlier quoted context omitted.

They have yet to do this with rare earth metals, which represent a huge strategic threat to both the US and much of the rest of the world to rely on China for a vast majority of the supply.

It has been said that other countries, including US, in fact have more rare earth deposits. If they choose to relax environmental laws and disregard patents held by Chinese corporations, more rare earth would be produced.

Yes, other countries have them. They're not willing to produce them at the higher cost required to obey environmental laws. I'm not sure how patent laws are a huge issue here: rare earths may have increasing market demand, but they've been extracted and processed for countless years. Only innovations China has made in the last ~20 years would be covered by patents. I'm sure there have been some innovations, but I'm also sure our other mining industries in the US would be happy to leverage their patent portfolio in a patent war if China wasn't willing to work on reasonable licensing terms.

A global strategic bottleneck in China for these things doesn't require patent or environmental law violations, it just requires us to pay more for them. For a critically strategic resource like this, the US should ensure a consistent supply chain independent of geopolitical concerns with China. And if China were to cut off supply then concern for their patents goes out the window too.

Re: Nvidia to Acquire Arm for $40B

#766

Post by floatboth »

dragontamer wrote: post #762
floatboth wrote: post #736

Earlier quoted context omitted.

Main lesson: we failed to make all the software JIT-compiled or AOT-recompiled-on-boot or something, that would allow retargeting the optimizations for the new generation of a VLIW CPU. Barely anyone even tried. Well I guess in the early 2000s there was this vision that everything would be Java, which is JIT, but lol

Your point seems invalid, in the face of a large chunk of HPC (neural nets, matrix multiplication, etc. etc.) getting rewritten to support CUDA, which didn't even exist back when Itanium was announced. VLIW is a compromise product: its more parallel than a traditional CPU, but less parallel than SIMD/GPUs. And modern CPUs have incredibly powerful SIMD engines: AVX2 and AVX512 are extremely fast and parallel. There ar…

I'm not talking about rewrites and GPUs. I'm saying we do not have dynamic recompilation of everything. As in – if we would have ALL binaries that run on the machine (starting with the kernel) stored in some portable representation like wasm (or not-fully-portable-but-still-reoptimizable like llvm bitcode) and recompiled with optimization for the current exact processor when starting. Only that would solve the "new generation of VLIW-CPU needs very different compiler optimizations to perform, oops all your binaries are for first generation and they are slow now" problem.

GPUs do work like this – shaders recompiled all the time – so VLIW was used in GPUs (e.g. TeraScale). But on CPUs we have a world of optimized, "done" binaries.

Re: Nvidia to Acquire Arm for $40B

#768

Post by PixelOfDeath »

a1369209993 wrote: post #630
DCKing wrote: post #531

Earlier quoted context omitted.

I wouldn't use the term "unfair" here. There's also just three x86 licensees in the world and people don't usually consider that an affront. You buy, you control, that's how the world works. But I do think it's important that we recognize that we're going from a position of tremendous competitiveness to a much less competitive situation. And that will be a situation where ARM will be tightly controlled and much less…

> There's also just three x86 licensees in the world and people don't usually consider that an affront. Hi! Counterexample here.

ME TO!

Especially because of the x86 oligopoly I would think that Arm is so much more important as an ecosystem.

Re: Nvidia to Acquire Arm for $40B

#769

Post by dijit »

paulpan wrote: post #578

My initial reaction is that this reminiscent of the AMD-ATI deal back in 2006. It almost killed both companies and comparatively, this deal size is much bigger ($40B vs. $6B) for both a more mature industry and companies involved. $40B is an obscene lot of money objectively and what's the endgame for Nvidia? If it's to "fuse" ARM's top CPU designs with their GPU prowess, then couldn't they invest the money to restart…

I see two ways of it playing out for Nvidia:

1) Adoption of ARM CPU's (AWS Graviton, rPi etc) will cause software to be adapted to ARM anyway, meaning: Nvidia could come out with a full vertically integrated cloud.

or

2) Leveraging full vertical integration with ML based super computers.

Re: Nvidia to Acquire Arm for $40B

#770

Post by stunt »

yogrish wrote: post #294

SoftBank isa true banking company .. invested (Bought) in ARM and selling it now for meagre profit. A company that says it has 300 year vision

Maybe they are forced to choose behind the scene. Or, they know better opportunities and they need to cash out right now.
Post reply on HN