Intel denies rumors of 3nm delays, says planned 2024 releases are on schedule

[Jimmy2x]

In a nutshell: Earlier this month, several outlets reported delays to Intel's upcoming 3nm-based technologies. CEO Pat Gelsinger squashed those rumors during Intel's Capital Allocation Update call, assuring attendees that products based on both Intel and TSMC's 3nm nodes are on track for their planned release dates. Gelsinger likens the rumors to those spread surrounding the Intel 4 process, which at the time were also proven false.

Intel's CEO did not shy away from addressing rumors and concerns surrounding the company's 3nm-based efforts. Gelsinger instead met the claims head on, attempting to put any doubts of Intel's progress to rest by assuring participants that the Granite Rapids, Sierra Forest, and TSMC-based 3nm graphics tile technologies are all still progressing on schedule.

The high-performance Granite Rapids lineup and energy efficient Sierra Forest lineup are being developed using Intel's in-house 3nm lithography process. The process is a refinement of the Intel 4 process, but with added performance and efficiencies that could extend beyond their data center-focused audience. The TSMC-based 3nm resources are specifically scheduled for use as a graphics tile on Intel's upcoming Meteor Lake and Arrow Lake client processors.

The subject of process nodes has become a rather muddy exercise in marketing over the last several years. While we often hear references to process node names as an indicator of size and technical superiority, the truth is that the elements originally driving process node naming conventions deviated from their original course more than 25 years ago. Today's node names are selected and adjusted by the companies introducing them to fit and compare to their current competition.

A good example of process node names not meaning much of anything is the previous Intel 10nm process and Samsung 7nm process. To the average customer, the 7nm process node was made to sound like a significant advancement and advantage over Intel's 10nm parts. But when reviewing the specifications of each node, it became apparent that the 10nm and 7nm process nodes had very similar characteristics.

Intel maintains its stance on their in-house foundry services, assuring customers and stakeholders that the internal 3nm process nodes will enter production and distribution in late 2023 into 2024. Recent reports also indicate that TSMC's 3nm support looks promising, reaching full production in late 2023 and resulting in yields as high as 80%.

Intel's efforts to regain a foothold in the consumer and data center hardware space continue to grow. According to Gelsinger, he feels confident that the company has turned a major corner regarding previous execution challenges and is adamant that "... these rumors, like many others, will be proven by our execution to be firmly false."

Permalink to story.

https://www.techspot.com/news/97731-intel-denies-rumors-3nm-delays-planned-2024-releases.html

 

[DSirius]

What Intel Pat Gelsinger CEO failed to mention is the yield percentage of their so called 3 (nm) node. The 4 (nm) node has a percentage quite low that still it is not financial feasible. Nowadays, it became a bad habit for chip manufacturers like Intel, TSMC, Samsung and others to announce that they achieved 4,3, 1 nm nodes fabrication right after they get few good samples. The problem is that after that, they often delay the mass production until they are able to refine the process to produce enough percentage yields without defects.

 

[Puiu]

It's not delayed, they have always planed it to be out a few years from now.

 

[rsmith222]

Can anyone create a scientific standard to replace this nm nonsense?

 

[Kashim]

If Intel is so great and they're doing so well, then why do they need to rely on TSMC for their 3nm graphics tile? Why can't they build the whole thing in-house? Their ultra-dork CEO talks like they're ahead of everyone and takes shots and AMD, but then turns around and buys chips from the same place AMD does. Makes you wonder.

 

[neeyik]

If Intel is so great and they're doing so well, then why do they need to rely on TSMC for their 3nm graphics tile? Why can't they build the whole thing in-house?

The fabs running with the Intel 4 and Intel 20 nodes are likely to be at high capacity making the CPU, SoC, and IO tiles. Graphics tiles are going to be of a more common format than, say, CPU tiles -- in other words, more SKUs will use the same Graphics tile than CPU ones. Intel deemed that it would be better to use TSMC for capacity and cost reasons.

 

[rsmith222]

Every Intel order is one its competitors can't make, as long as supply constraint exists at TSMC. Of course, every order also enriches TSMC.