I wish him luck, though.
As Adam Smith said, "there is a great deal of ruin in a nation", and likewise, big companies generally get more opportunities to reinvent themselves than one would expect.
Tired brain read this as "there's a great deal in a ruin of a nation." Close enough, I guess?
Until still have some amazing technology compared to AMD, they're five processes screwed right now but I think they have a chance to succeed. But given the timeline on new CPUs and fab processes, it'll be 4 years before we see the fruits of anything he does.
Competitors maybe passed them in the last couple of years, but Intel is still in a really solid position to turn things around.
Apple's situation was a lot more dramatic. Intel still has a very good market share and can probably take a couple+ years to safely get back on track, in my opinion.
Amongst other things:
- SGX is really far ahead of AMD SEV. The latter is probably easier to sell because it's marketed as "drop in" (encrypts whole VMs), but SEV has been repeatedly hacked or shipped with obvious gaping design holes that they patch later and call features. SGX is a lot more focused, a lot more flexible and frankly a lot more secure.
- Optane NVRAM is completely unique, as far as I know. It's only 10x slower than DRAM which is nothing, but it's persistent! It totally changes the whole IO hierarchy.
- AVX512 / DLBoost are able to hold their own against mid-range GPUs for some AI tasks, which is impressive and useful.
- Their core chips are still very fast. AMD chips are selling more for less, which is a good position for them to be in, and TSMC's process advantage is helping them out for now. But they don't have a truly massive edge in tech like Apple's competitors had when Jobs returned.
- Intel have a long tail of obscure features that AMD doesn't, although it's often hard to know this. SGX and AVX512 are high profile but there are others that are less well known.
I don't count side channels as an issue because it's also an issue for all their competitors, and frankly I found the near single-minded focus on Intel by the security community to be rather misleading. I even read a side channels paper that admitted they suspected AMD had the same problem but they didn't bother to check simply because they didn't have access to any AMD hardware in the first place, which was unusually honest.
Apple's M1 is very impressive in its space, but for high performance cases like servers it's only quad-core - you can't use the little low power cores for much. Apple have shown no interest in making server parts for a long time, and Apple's engineering is bespoke so it tells us nothing about what other ARM vendors can do. So in that space it's still just AMD vs Intel and Intel is a long way from being on its back yet.
this is true, but does not account for the fact that apple M1 is just the first (public) iteration of such cpu.
Then there are companies that have at least managed to stabilize themselves reasonably, e.g. IBM and HP (I wouldn't consider either of them a huge turnaround success, but I would not expect them to head for bankruptcy anytime soon either).
And even companies that did go under, e.g. Kodak, took an enormous amount of time and effort to do so, and not without launching a cryptocurrency first…
If you have $233 billion (a LOT of money) and you want to own a world-class chipmaker, would you rather start from scratch or just buy Intel for cash?
I feel like the view of Intel as a "sinking ship" is an inside-basebally misread of the situation.
If you buy a PC or a laptop or a server today, you're most likely getting an Intel CPU. It's now at least possible to buy AMD in many market segments from major vendors - and of course many of us do - but to the broader consumer and hosting world Intel still dominates.
You can look ahead and extrapolate and say "they can't compete with TSMC right now, and maybe they're going to start falling behind further." Fair analysis. But they dominate the market to an extent that's hard to overstate.
Contrast this with between-Jobs-stints Apple, which was a tiny shrinking company with a niche product and no clear strategy.
Intel is almost in too-big-to-fail territory. They aren't sinking, just because they now have real competition. Yes, they need to do "something" to maintain dominance - but why would you see their situation and conclude they won't? The new CEO here is a sign that they see the trouble ahead, and they're ready to steer around it.
If they actually want to survive long term, there are two paths as I see it:
A) Be legitimately better than AMD, this could include opening up the management engine, much higher performance chips at lower price points, or some sort of space magic utilizing their Altera acquisition.
B) Embrace RISC-V and push it to laptops and desktops HARD, while not pulling the Microsoft Embrace Extend Extinguish™ play. If they go this route then their stock becomes an exceptionally strong buy IMO.
There's a lot of interest in that from a national security perspective anyway.
Part of TSMC's offer is that they won't do design, they won't compete with you on design etc.
They specialize in the best possible fab and service for your design and that's it.
That's kind of a hoax. GUC (Global Uni Chip) is huge, and a nearly-wholly-owned subsidiary of TSMC.
https://en.wikipedia.org/wiki/Global_Unichip_Corporation
TSMC has a large design operation. It simply is part of a subsidiary rather than the parent company, that's all.
There are a lot of shady links between TSMC and GUC. I was once pressured to "collaborate" with GUC as an explicit condition of getting access to the then-bleeding-edge 16nm PDK. I turned it down. A competitor of mine (with whose chief engineer I am friends) had the same screws put to them.
Imagine 100x of that, and that is Samsung.
I want to be able to train NNs with a laptop that can normally last 20 hours.
I realize that WinTel machines arent going anywhere because of entrenched business use -- but between NVIDIA GPUs for heavy workloads and M1s for day-to-day activities -- what is the feeling inside Intel right now? Is this like a Microsoft-Netscape moment in the 90s?
What it should definitely be, however, is a death-sentence for x86. I bet Intel are proud of it, but ultimately they must be at least somewhat jealous of those who don't have to use it.
We don't really know why Keller left. All we can say is that from the outside it looks like he might have just given up in disgust. Having an engineer at the helm again can make a big difference to an engineering organisation.