https://news.ycombinator.com/item?id=42361955
I'd suggest skipping past the long stream about how ETFs work, although it does touch on how in the world the Intel Board has remained in place while driving the company into the dirt.
6 replies (most recent on top)
Corrupt Indian managers with helping from low technical background young white managers working together to flush Intel into toilet sink
Board should be replaced and its current members should never be able to find another job again. They are a disgrace.
Was the board responsible for this ad?
https://www.youtube.com/watch?v=HbkdpyUlJNs
Or was it the completely out of touch Intel culture?
the board did a poor job, I am not sure why most of them are not broke
probably corruption
then intel hired and fired poorly
intel needs to break up and perhaps reform each piece separately
some pieces will thrive while other pieces will fail and disappear
This Board must go!
Scraped some of the comments:
Even as late as the semi-aborted 2018 launch of Cannon Lake it seemed like it was just a routine burp they'd correct with a process respin. Then TSMC quietly reached parity with 7nm, shipped 5nm which was a better process, and by 2021 Apple had jumped ship and Intel was falling behind even AMD.
It was a desperate move to avoid shareholder lawsuits and possible criminal prosecution. Intel had to ship something under the 10nm label before the end of the year, because they had made far too many (false) promises that 10nm would be working Soon.
Then TSMC quietly reached parity with 7nm.
The true embarrassment was when SMIC (read: China) reached 7nm and thereby surpassed Intel last year (or was that 2022?).
As in many large projects, even more so with a large company, the point is not to react when the problems are happening, it is to preempt these problems, foresee them and prevent their happening.
So indeed, by 2018, even though Intel has not yet fallen, it's actually already late. The roots of the problems seems to be earlier, and that's where the CEO, and the Board, should have reacted.
Meanwhile, Intel's chip designers kept targeting an unusable process, and wasted years that they should have been iterating on designs for the fab process that actually worked. Skylake shipped in 2015. They didn't deliver a new CPU microarchitecture on 14nm until 5.5 years later, a year and a half after they shipped that same microarchitecture in a mobile-only form when their 10nm finally started to be somewhat usable (but not fast enough for desktop).
Not participating in the feedback loop of actually shipping left Intel with an oversized P-core design and an E-core design that wasn't well-matched to it, making Alder Lake awkward and slapdash when they finally got 10nm working well enough for desktop CPUs.
Once Intel lost its research focus it became an extractive company extracting the riches that were already there, instead of creating true innovation.
It was well-known publicly that Intel was running their business in a way that maximized the damage any fab troubles would have on their product roadmap. Design pipelines are deep and Intel at the time famously had very node-specific designs without industry-standard PDKs. The moment engineers were told to switch a design to 14nm, it basically reset the 5 year design-to-product pipeline. Management failed because they did not hedge the risk by starting a parallel 14nm design effort at first sign of 10nm troubles.
Buying Altera and forcing them to port their entire roadmap over to a broken 10nm process was made even more stupid by the fact that Intel didn't have a usable PDK that outsiders and acquisitions could work with.
AFAICT, this was a self-serving bit of reverse myth-making from Otellini. If there really had been a single binary decision Intel got wrong—saying no to Jobs when they might have said yes—then their collapse looks like bad luck: Nobody bats 1.000.
But the way Apple insiders tells this story, there was no way Intel was even being considered in the (short!) window when the original iPhone was being built. Intel was in the middle of selling Xscale, and even that design was too power-hungry. Jobs had two iPhone teams working in secret against each other, and was setting up things on the side. He likely approached Otellini before either team was far along.
AI is the most recent and buzziest form of HPC, but people have been buying AI-style clusters for a while. The Knights [] products were a great entry point into that market if they would just figure out the software of it. Knights [] essentially had Atom cores with huge SIMD units, and fast forward to today, an AMD GPU is essentially a small scalar core attached to a huge SIMD unit.
IMO it seems like they had the Google problem: if you can't sell a billion units it has to be canceled.
That's because Larrabee wasn't totally canceled, it was just pivoted from a gaming GPU to an HPC accelerator in the form of Knights Corner/Landing. The idea of having lots of x86 cores with wide SIMD units didn't change, but it was a lot more successful in the HPC world because anybody could compile an old MPI or OpenMP application with AVX512 and it just worked.
Larrabee itself was a complete disaster. Xeon Phi resurrected the Larrabee trash heap into something sellable. Sadly, it was never really that easy to program a Xeon Phi.