Intel Arrow Lake to be made elsewhere as 20A process node canned Surprising news about Intel continues to emerge with the chipmaker vowing to use an external foundry in place of its own 20A process to make the upcoming Arrow Lake processors, amid talk that Broadcom has rejected Intel's 18A process as not ready for mass production. The Santa Clara giant posted an update on its website …
Concluding that the process is not ready yet for high volume production is either a frank assessment that the yield/reliability is insufficient or a very nuanced way of saying the performance doesn't meet its target.
I worked at Intel for many new process ramps and the yield was always an issue that got chipped away at fairly quickly. Admittedly the process now is much more involved and they've leaped over a number of generations with 18A and the learnings they would have gained. The yield improvement roadmap and their ability to execute to it is likely a huge question.
That was my thought. The "Not ready for production" and "We have working units" aren't mutually exclusive. But maybe only 1 in X chips on each wafer are actually "good", because Gremlins that need to be worked out in the process? Broadcom probably want to see a 90% yield, and some real world testing to show the chips don't over-volt and kill themselves. Once the process is reliable and proven, THEN they sign the contract.
Since 20A was only going to be used for a couple of Intel CPUs, and the real foundry push doesn't come until 18A, it isn't clear what purpose 20A had other than maybe as practice for 18A. But since 18A is going to sink or swim on its own merits, and Intel needs to tighten its belt where it can, spending half a billion to fully equipping a fab for a process that is only used for a couple products and has a short lifespan seems like something that can justifiably be cut.
What’s unusual to me: 20A as a testing ground was the financially prudent choice. Isn’t $500m of validation worth it to ensure 18A rolls out near-perfectly?
20A as a validation step was billed as critical. Intel getting cocky about 18A and trying to save costs sounds like a terrible 2010-era Intel mentality.
It is illogical, or at least inconsistent.
Cut costs in other places, not anything related to 18A’s delivery. The latest nodes are the wrong place to cut funding, especially in what is essentially a major validation test.
It isn't like they can't still test 18A, but they can do it without setting up a single use production line in one of their fabs and releasing products based upon it. That's the real "expense", tying up valuable fab capacity with 20A that could be built for 18A so that when it is ready to go they have more production capability than they'd otherwise have. That's their first real node as a foundry, they need it to hit the ground running and not waste a lot of time refitting a fab built for a process designed to be obsolete in six months.
You are spot on and Intel has done this before. The company had already ramped up the 120nm node on its 200mm production line in 2001 and also brought it's first 300mm line up on the same process node just be ensure the performance and yield were matched. This uncovered numerous startup issues and tool quirks that were resolved. The 120nm volume was very limited and they quickly moved to the 90nm node.
Intel is another company whose ability to execute on its core mission has been compromised by having MBAs replace engineers at the top of the company. Sure, they can lay folks off and close facilities and improve the next couple of quarters' results, but chipmaking is not a business for cheapskates, and anybody not deeply involved in it probably can't make the informed decisions necessary to stay on top of it. Accountants only see the cash flows, not what makes them happen. Boeing, boeing, gone...
>"one of the benefits of our early success on Intel 18A is that it enables us to shift engineering resources from Intel 20A earlier than expected as we near completion of our five-nodes-in-four-years plan."
Da, Komrade, Did I say 3x? I must have mis-read my notes, tractor production is actually up 10x over the previous 5 year plan.
... well, current, anyway. Twenty amps is a mighty flow of electronic goodness, I can see why they would want to drop it to eighteen amps.
Ohhh, right. You meant to say 20Å or better still 2nm, which someone will now point out sometimes means two nautical miles.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
