So what happened to intel that they’re now no longer way ahead in both performance and manufacturing technology? They’re being squeezed from all sides, and don’t seem to be pulling ahead...
They stumbled on 10nm is what happened. They expected their manufacturing prowess to continue to let them build massive dies.
So now they're stuck with CPU architecture designs that mandate huge dies to scale up the core counts, which they can't manufacture with anything close to reasonable yields on 14nm.
Rock meet hard place.
AMD by contrast designed with the expectation that they couldn't make big chips, so went with a "glue a bunch of small ones together" design. Which seems to have played out stunningly well for them. Now they can bin the golden cores, slap them together, and ramp the clocks.
It's not so much Intel were ahead in terms of architecture design, but rather AMD was way behind. Bulldozer was a disaster, and Zen is a ground-up redesign that means AMD have a decent core at the level of Intel's again.
But AMD have two advantages Intel don't. Their Zen architecture is designed for multi-chip module scalability, so they can deliver higher core counts at much better yields (especially important on new process nodes!) and thus manufacturing costs than Intel's monolithic designs. And AMD uses 3rd-party fabs that, unlike Intel, are already doing great on the new process node.
I dunno if it was as much a "disaster" as it was Intel's Sandy Bridge doing so well.
Intel's Sandy Bridge (2600k / 2700k) were HUGE improvements back in 2011. Bulldozer was a step-wider (the cores were roughly the same as K10 but you'd get 8-cores instead of 4), and Piledriver / Steamroller incrementally improved on the formula.
Bulldozer managed to increase core counts from ~4 (AMD Phenom) to 8 with Bulldozer. Sure, the 8 "cores" of Bulldozer shared a decoder and perhaps was more appropriately a 4-core with hyperthreading... but it was still a core-to-core improvement compared to K10.
But the incremental upgrades to AMD's K10 were just no match for the 20%+ boosts that Intel was doing with their Sandy Bridge architecture. Ultimately, Intel's Hyperthreads (4c/8t) were roughly the same as AMD's "8 core (4-decoders)" setup... because Sandy Bridge was just so far ahead of the game.
Bulldozer was less performant than its predecessor once you factor in the process node. That's why it was a disaster: AMD managed to design a worse processor and were stuck with it for years. They did make it less bad over time, but it still sucked.
This. Bulldozer was (ironically) AMD's Netburst moment. They made a speculative play ("modules" in the case of Bulldozer, long pipelines in the case of Netburst) to chase high core counts/clock speeds (respectively) and the technology didn't end up panning out like they expected on top of performing worse out-of-the-gate than the architectures they were meant to succeed.
The difference is that Intel had the cash and political clout to wait out Netburst and force the market to take it while Bulldozer nearly killed AMD (which only held on thanks to its GPU division, itself now in crisis due to lack of competitive products due to under-investment).
As others have pointed out, it's not entirely about Intel doing badly, but rather Intel putting out a mediocre performance and AMD finally really succeeding.
And while the various technical reasons for that are interesting, I do think that at least some of the credit/blame needs to go to the leadership. AMD had a string of bad leadership that led to the company to near bankruptcy, but then they hired Lisa Su, who is basically a superstar engineer-turned-manager who has excelled in everything she's touched, from doing very low-level transistor research to being the leader of large teams (and now the entire company). At the same time, Intel hasn't actually had a good CEO for a long while now, with the top leadership in the company chasing the latest fads and spending billions on weird acquisitions that only get written off later, while the key areas of the business are not doing nearly as well as they did under their predecessors.
Yeah, if AMD could run on fumes for the better part of a decade, when not even dominating the market previously, just having reasonable alternative choices to the market leader, I'm pretty sure the 40 year long market leader can survive a bad arch refresh cycle or two.
Intel isn't finished, not by a long shot. Its very likely they are working on a chiplet design of their own.
Next iteration is hard. Breakthroughs are even harder. When Intel (or whoever) does it, they will have years of advantage - unless the information leaks. (And it probably will. At the latest when the true next gen will hit the shelves.)
Other than that, it's business as usual. News are just shiny mirrors, spectacle. We still don't know how durable AMD's "luck" is, how sales and stock price and other relevant numbers will ebb and flow, and so on.
