Hacker News Reader

Hacker News

Comment by sradman | Hacker News Reader

Full thread

sradman

In my estimation, Intel has four categories in which it is being outperformed by key competitors:

1. TSMC/Samsung - fabrication

2. AMD/Amazon-Graviton - Cloud Server

3. Apple-M1/AMD - Laptop

4. Nvidia/Amazon - Cloud ML/DL Accelerator

Intel has made giant blunders in the past (e.g. Itanium [1], Atom [2], WiMAX [3]) but I'm not sure that any of these past challenges were equivalent to the current four-front war. I would not count Intel out at this point but it will be several years before we know if they were able to right the ship.

[1] https://en.wikipedia.org/wiki/Itanium

[2] https://en.wikipedia.org/wiki/Atom_(system_on_a_chip)

[3] https://en.wikipedia.org/wiki/WiMAX

View on HN

You forgot their main market, AMD - desktop x86/windows

But the first point on your list is where their real loss is, it's historically the source of their strenght and they've never been losing at this one.

Is that their main market? It’s surely what they’re known for to the general public but I’d have guessed they care more about the Xeon line. But I’d be interested to hear that I’m wrong.

https://www.intc.com/news-events/press-releases/detail/1439/...

pc-centric "client computing group" made 40b revenue last year, over half their 78b total. probably something like 250 million cpus at an average price of $160

> driven by record PC-centric revenue with PC unit volumes up 33 percent YoY led by record notebook sales.

I think we need more details on the CCG split between notebooks and desktops. When I chose "Laptop" as the description of my 3rd category I was trying to choose a label that included the Apple M1 but didn't imply PC-Centric which I associate with Windows. I'm fine with Laptop+Desktop or some other permutation; I wasn't trying to be overly fine grained with the categories.

> The company also announced that its board of directors approved a cash dividend increase of five percent...

A 5% increase in dividend for a technology company is expected for a massive cash-cow with limited future growth. The Financial Report is an interesting read.

> The Financial Report is an interesting read.

Mind sharing a link if you have one?

I'm kind of questioning whether it even makes sense for Intel to try to compete with Nvidia and AMD when it comes to GPUs or deep learning accelerators. They just need to really win in one area in a sense, or at least to have a product with a really compelling price/performance ratio. Though you could argue there's a certain synergy between laptops, desktops and servers (people will generally favor running the same architecture they develop on).

I think Intel SIMD (AVX-512) is still best-of-class for vector math. Run Length Encoded algorithms in column stores like SAP HANA are currently the key use-case (revenue/market wise). What I've learned since the release of Apple M1 is that ISA extensions for Matrix math, as incorporated in Apple's AMX, complement the Vector-centric SIMD ISA and are probably a key battle front for ML/DL. A good Vector/Matrix SIMD-like ISA should cover many ML/DL use cases currently addressed using a discreet accelerator like Apple's Neural Engine and Nvidia GPUs.

Amazon's Graviton ARM Neoverse CPUs only implement ARM NEON SIMD rather than the newer SVE2. I don't know if ARM SVE2 has Matrix math instructions like AMX does. Intel AVX++ might be an important alternative to discreet ML/DL accelerators. DL Training accelerators appear to be focused on RDMA Over Converged Ethernet [1] (RoCE) and I'm assuming that this technology is new enough for Intel to gain a foothold.

I don't know if I understand this space well enough but I think there are enough truths in this explanations to keep me from ruling out Intel.

[1] https://en.wikipedia.org/wiki/RDMA_over_Converged_Ethernet

I wouldn't rule them out either. AMD survived some very bad years. It's not impossible for a company to shrink, trim the fat and grow again. However, having known people who worked at Intel, it's not just a question of lack of innovation, they had some very smart engineers, but it seemed like the culture there was very toxic. Lots of unhealthy internal competition, bureaucracy, infighting, etc. I think that they need to improve their internal culture first, in order for the company to make a real comeback. If anything, I think their recent losses are a serious wake up call. I hope they make the right decisions, because the hardware industry needs competition.

What you're talking about here is the client/inference side, not the server/training side of the ML/DL compute battle. The former is meaningful in a sense, but practically insignificant in relation to the latter. Almost all ML/DL compute GPU, CPU, or otherwise is on the server side of the equation, and if you're relying on vector CPU math for it, well good luck.

Accelerators have a very clear space for themselves on the training side, and no amount of fancy ISA magic is really going to make up for the fact that TPUs and the like beckon. It's where most of the real money is.

Wayne Gretzky: "I skate to where the puck is going, not where it has been.”

Intel Acquires Artificial Intelligence Chipmaker Habana Labs [1]:

> Intel estimates the total addressable market (TAM) for AI silicon by 2024 will be greater than $25 billion, and within that, AI silicon in the data center is expected to be greater than $10 billion in the same timeframe.

Amazon EC2 instances powered by Habana Gaudi [2]:

> Up to 40% better price performance for deep learning models

Matrix CPU math is another plausible future; Intel, AMD, and/or ARM are positioned to drive such an ISA extension. Accelerators, coprocessors, FPGAs, and CPU ISA extensions all seem to be in play; it is too early to pick winners in this nascent market.

[1] https://newsroom.intel.com/news-releases/intel-ai-acquisitio...

[2] https://aws.amazon.com/ec2/instance-types/habana-gaudi/

5. Phones/tablets, back in 2016 and before. https://www.extremetech.com/computing/227816-how-intel-lost-... If they were competitive there, maybe the M1 wouldn't have happened either...

Your list is skewing toward Amazon, but point well taken. If you include Google and Microsoft, they're all chip-designers as well because they can simply license ARM's designs. Imagine if Netflix decided to hire chip designers. All that's left is fabrication and Intel is exiting that.

Netflix runs on AWS, and even if they didn't, I would seriously question their management if they decided to hire chip designers. There's something to be said about being focused and not trying to do everything in-house, or specializing where it makes sense to do so.

Yes. Netflix sells content. Any savings they make on the tech-stack side is going to need to be shared with the companies that they are buying content from.

It makes much more sense for them to focus on making content than making ARM chips.

Why wouldn't they? Maybe not the best of the best of chip designers, but sure, at least a half decent one. Just so you know what kind of chip would be perfect for the kinds of workloads that you have, without divulging secrets to anyone else, all so that you can find the best value for money out there in the market. Honestly, if you didn't do this, I would probably consider you incompetent.

There is no way Intel is going to exit fabrication.

They're using TSMC out of necessity, not desire. Intel's fabs are integral not only to their success and profits, but to the infrastructure of the United States itself. We simply cannot allow all CPU fabrication to be offshored, especially when most of it is close to China (Taiwan / TSMC). Samsung is too busy pumping out smartphone / tablet CPUs and NVIDIA's GPUs, so even if Intel wanted to use them, there's not enough capacity.

Intel is going to have to get their shit together, not just for their own sake, but for strategic manufacturing security interests of the United States as well.

I think it's important to point out that the US has plenty of silicon fabrication plants. They just don't produce the cutting edge 3nm chips, instead they focus on the older chips that make up the bulk of the market. Silicon chips are still the fourth largest American manufacturing export.

Given the expensive involved in bleeding edge manufacturing and the comparatively low volumes of sales, the consolidation of manufacturing capacity is not really surprising.

5. AMD/TSMC - desktop x86

In which category is Intel not being outperformed by competitors?

Money. Intel's operating income is about 66 times bigger than AMD's for example.

Intel will probably never get a run like the last decade, but they have enough market share and money to get themselves back on track.

But AMD isn't the right comparator:

Q1 2021 Revenue Guidance

Intel $17.5bn - 12% YOY

TSMC $13bn + 20% YOY

Reply on news.ycombinator.com