Posts: 470   +2
Staff
What just happened? Intel is canceling its 20A process node for Foundry customers and reallocating resources to the 18A node, which is scheduled for production in 2025. The company also stated that it will transfer the majority of Arrow Lake production to an "external partner," believed to be TSMC.

Intel noted that the lessons learned from the 20A node contributed to the success of 18A, which incorporates several techniques, materials, and transistor architectures that originally debuted with 20A. The 20A process was the first to integrate both RibbonFET gate-all-around transistor architecture and PowerVia backside power delivery technology.

RibbonFET represents Intel's first new transistor design since FinFET, while PowerVia enhances transistor performance and reduces power consumption.

Intel's 20A node, which was the foundation for Arrow Lake CPUs, was announced in 2021 by CEO Pat Gelsinger. It was intended to mark the beginning of the "Angstrom era" of semiconductors, where transistor sizes would be measured in angstroms (A) rather than nanometers (nm). Consequently, the company named its 2-nanometer chips 20A and its 1.8-nanometer chips 18A to reflect this new naming convention.

Team Blue also noted that 18A is already operational, successfully booting on operating systems and yielding positive results. According to the company, the success of 18A allowed it to shift engineering resources from 20A earlier than anticipated. Intel also claimed that despite the cancellation of 20A, it is still on track to meet its goal of delivering five nodes in four years.

Intel may be optimistic about its 18A node, but a recent Reuters report suggests that it is far from production-ready. Sources quoted by the agency indicated that Broadcom found the initial production run of 18A wafers to be highly unimpressive after extensive testing on its internal systems, including AI accelerators and networking switches.

In a statement to Reuters, an Intel spokesperson reiterated that 18A is delivering good yields and is on track to enter production next year. However, the spokesperson declined to comment on whether Broadcom's partnership with Intel remains intact despite the disappointing 18A tests or if the company is considering withdrawing from a potential manufacturing deal. For its part, Broadcom stated that it is still "evaluating the product and service offerings of Intel Foundry" and has not yet made a final decision.

Permalink to story:

 
Intel just keeps going from bad to worst.
 
Intel just keeps going from bad to worst.
I don't understand that logic. If Intel has a 2nm chip design that can successfully be transferred to a node that can create 1.8nm features then, theoretically, it means higher yields
 
I don't understand that logic. If Intel has a 2nm chip design that can successfully be transferred to a node that can create 1.8nm features then, theoretically, it means higher yields

This article missed some information. Such as that 18A defect rate is at D0 <0.40.
You also have to consider that most Intel CPU that are to be released in 2024 and early 2025 will be made outside of Intel.
 
Given overall Intel's shape, I find it very hard to believe in their positive future.
Or perhaps, this is a great opportunity to buy their stocks and wait till they do what AMD did?
 
I don't understand that logic. If Intel has a 2nm chip design that can successfully be transferred to a node that can create 1.8nm features then, theoretically, it means higher yields
Key word being "theoretically" here. This is the same old dance Intel has been doing for years. Their 18A process is doing so well that they decided to cancel 20A and outsource their chips to TSMC. Oh yeah, makes sense. Next we'll hear that their 16A process is doing so great they might as well cancel 18A process too. All the while actual products based on those processes are getting pushed further and further back.

When a big company like Intel has to outsource their products to their competitors and make promises about future success people need to take that news with a grain of salt. I'll believe anything Intel and their uber-dork CEO says when I actually see it listed with a price tag on Newegg or Amazon. Until then it's all just PR BS to appease investors.
 
Two stints with Intel as a green badge; one at Beaverton campus and one at Hillsboro campus. It isn't just Pat that is failing. The entire culture of that company is insane. If you dare point out how their competitor is doing better, even in an insignificant way (God forbid in a key area), you are labeled a heretic and quite literally bullied by the cult members. Not 'everyone' drinks the cool aide, but the key people sure do. And not in a "CYA" way, but full on 'let's eat a bowl full of tainted pudding and ride the comet" way.

We've all watched externally through the years. Even when they are on top, the amount of turmoil in the company is too rough to justify working there. Haven't been back since my last stint. Thus, I'm not surprised they are a mess; only surprised it took this long for the cards to collapse.

It's (probably) wrong, but I will enjoy the fall of Rome while eating a bag of popcorn. Well deserved IMO; and justice for all the other Blue and Green badge folks who spoke up to make the company better but were slapped down or fired.
 
Two stints with Intel as a green badge; one at Beaverton campus and one at Hillsboro campus. It isn't just Pat that is failing. The entire culture of that company is insane. If you dare point out how their competitor is doing better, even in an insignificant way (God forbid in a key area), you are labeled a heretic and quite literally bullied by the cult members. Not 'everyone' drinks the cool aide, but the key people sure do. And not in a "CYA" way, but full on 'let's eat a bowl full of tainted pudding and ride the comet" way.

We've all watched externally through the years. Even when they are on top, the amount of turmoil in the company is too rough to justify working there. Haven't been back since my last stint. Thus, I'm not surprised they are a mess; only surprised it took this long for the cards to collapse.

It's (probably) wrong, but I will enjoy the fall of Rome while eating a bag of popcorn. Well deserved IMO; and justice for all the other Blue and Green badge folks who spoke up to make the company better but were slapped down or fired.
The best thing Intel can do right now is purge their management and streamline heavily. The second best thing is get their fabs back on track. #1 will greatly help #2.
 
Key word being "theoretically" here. This is the same old dance Intel has been doing for years. Their 18A process is doing so well that they decided to cancel 20A and outsource their chips to TSMC. Oh yeah, makes sense. Next we'll hear that their 16A process is doing so great they might as well cancel 18A process too. All the while actual products based on those processes are getting pushed further and further back.

When a big company like Intel has to outsource their products to their competitors and make promises about future success people need to take that news with a grain of salt. I'll believe anything Intel and their uber-dork CEO says when I actually see it listed with a price tag on Newegg or Amazon. Until then it's all just PR BS to appease investors.
I believe 18A was a small increment to 20A and was scheduled to run on the same equipment and fabs. Also, if I remember correctly, 20A was only for intel and 18A was for external customers also. So, if they can skip 20A and bring forward 18A a quarter or two, it could make sense. Or maybe they lack the resources to bring both nodes to production. For sure, skipping a node will increase the risk of delaying the next one because of unforeseen issues. After all these years of missteps trying to leapfrog the competition intel needed to execute in small steps and on time. Now they seem to be forced into the same old failing pattern: big risky leaps. The thing is aggravated because this time they are paying TSMC billions to build their CPUs. So, probably 18A is do or die for the intel foundry.
 

Similar threads