Read more.Quote:
Firm apologises for PC CPU shipment delays. Is increasing use of third party foundries.
Printable View
Read more.Quote:
Firm apologises for PC CPU shipment delays. Is increasing use of third party foundries.
In today's news, intel struggles to remember how to compete and slowly begins to realise that it has been resting on its laurels for far too long, taking for granted its previous dominant position. Meanwhile AMD continues to kick it around the park.
Classic tortoise-hare situation, or an alpha male who's dominated the pack for so long it's forgotten what it's like to have to slug it out suddenly being confronted with a fresh challenger emerging from adolescence.
People keep saying that, but honestly I think this is the best they could have done. AFAICS at no point did they let up on their R&D spending, they just aren't a talented enough company to compete right now.
I notice they claim there that they have increased their 14nm capacity, I wonder if that is referring to how they shut down one of their failed 10nm lines and re-fitted it back to 14nm.
I'm not sure it's much to do with talent, they've probably got that in spades, it's more that someone took a gamble on cobalt interconnects and that didn't pan out, if it had they'd have been well ahead of the curve with a route forward all the way down to 2-3nm.
Taking a gamble goes against all the engineering methodologies of rapid evolution of product popular in the last 20 years. They didn't have to take a gamble on process, they didn't have to tie their cpu designs to that 10nm gamble. Really, they could have left out all sorts of things and still released a 10nm line, leaving the bits left out for a 10+++ process later.
So Intel pumped billions into R&D, and out came lacklustre improvements in their cores which are still crawling with security flaws which they can't manufacture because they stupidly bet the farm with no plan-b. This is on a background of regular employee sackings that can't be good for morale.
Intel have stayed at the top for decades through a combination of buying the best process tech, monopolistic practices and luck. Engineering talent has never been a big part of the equation.
No one said it was the engineers who took the gamble, based on nothing more than a hunch i suspect it was whatsisname, their CEO that stood down that over-rode the advise he was getting from the shop floor, so to speak.
OK, so perhaps there was a gamble taken by one person on whether to do a mad push for an impossible 10nm. I find that hard to believe, it would take quite the charismatic leader to try that and clearly neither of us can even remember the guy's (it was a bloke wasn't it?) name :D
To me it feels more like someone in the middle tried and for too long succeeded in covering up how development was failing, so the upper management thought that 10nm would be on time as usual. That can easily happen in a company where there is a climate of fear, and Intel is a company that apparently does regular performance appraisals where they then automatically fire the bottom 10% which is a recipe for backside covering if ever I saw it. If this was an intended gamble, there would have been close monitoring and there would have been a plan-b. There should have been a point where the CPU designers needed firm design rules and at that point any tech that is holding the process back gets pushed out for later iteration.
To be fair, 14nm came along nicely with evolving iterations coming along at a decent pace. That is to be expected, Intel always was a silicon foundry first so yeah they must have some people at that coal face that know what they are doing. But the 10nm farce shouldn't be expected, and whilst it smacks of management incompetence it just shouldn't happen at an engineering company.
But then this is the company that bought us the Pentium 4, and had all their hopes pinned on that abomination. Core was lucky serendipity from a small team in Israel that did an overly good job of the mobile Pentium design, it was never an intended product, the future vision was high GHz P4.
Edit: I suppose their greatest idiocy was the "x86 everywhere" push some years back that gave us a 486 style design for IoT and the fabled Larrabee. x86 is a poor choice for the desktop really, but has so much momentum that it continues. But to try and displace RISC and VLIW cores where their strengths are particularly effective showed a real lack of perspective.
Charlie makes one of those comments that once you see it seems blindingly obvious but I hadn't joined the dots: Intel can't make enough chips because, to keep up with AMD, the chips they are currently selling are twice the size they used to be.
I guess that explains why Intel have been keeping their mainstream on 4 cores for so long, if it meant they could avoid investing some of their cash in new fabs. The move to 10nm would have allowed them to cram a few more cores on a die without increasing the size they deem us worthy of had it worked. But hey, this is Intel, they can buy their way into greater fab capacity. Still not convinced they have anything in reserve in their basic core design though.
https://semiaccurate.com/2019/11/25/...update-letter/