Gelsinger 5 Nodes in 4 years; Success or Failure?

[fansink]

It’s been over 3 years since Pat Gelsinger made the somewhat suspect claim that Intel will have completed 5 nodes in 4 years (Intel 7, Intel 4, Intel 3, 20A, and 18A).

Intel is currently fabricating 2 of the 5 nodes, and the clock runs out in less than 1 year.

Are the 5 nodes really just 3 nodes (half nodes shouldn’t count)?

Do the nodes have to be in high volume production?

 

[nghanayem]

They have over a year to finish. Pat is 3yr old but "5N4Y" is only 2 years old.

Intel's Process Roadmap to 2025: with 4nm, 3nm, 20A and 18A?!

www.anandtech.com

If half nodes don't count it is 2N3Y (assuming 20A ARL launches this year) because 10nm has been working well since 2020.

If intel can launch 20A ARL by July 26th 2025, then intel has met their original goal for "5N4Y" and if 18A anything gets delivered before Dec 31st 2025, then they have also met their first post "5N4Y" goal. Of course intel pulled in 18A from ramping sometime in 2025 to ramping @EOY, and I think we can all agree that counting a subnode that had launched products a year before T=0 as being the first node in 5N4Y is disingenuous. So intel has wisely shifted the "5N4Y" goalpost up to drop 10SF and include 18A. Therefore, if CWF launches by July 26th 2025, then intel has met the goals for the newer more aggressive "5N4Y".

 

[lisadelreal]

They have over a year to finish. Pat is 3yr old but "5N4Y" is only 2 years old.

Intel's Process Roadmap to 2025: with 4nm, 3nm, 20A and 18A?!

www.anandtech.com

View attachment 1711

View attachment 1712

If half nodes don't count it is 2N3Y (assuming 20A ARL launches this year) because 10nm has been working well since 2020.

If intel can launch 20A ARL by July 26th 2025, then intel has met their original goal for "5N4Y" and if 18A anything gets delivered before Dec 31st 2025, then they have also met their first post "5N4Y" goal. Of course intel pulled in 18A from ramping sometime in 2025 to ramping @EOY, and I think we can all agree that counting a subnode that uno online had launched products a year before T=0 as being the first node in 5N4Y is disingenuous. So intel has wisely shifted the "5N4Y" goalpost up to drop 10SF and include 18A. Therefore, if CWF launches by July 26th 2025, then intel has met the goals for the newer more aggressive "5N4Y".

How does Intel's roadmap compare to TSMC's?

 

[Piefkee]

The open question is what does it really mean? Does it mean the manufacturing high volume or is it just that the development of the node is complete?
The startet the 5 node in 4 year with already producing the Intel 10 or (Intel 7) node.
As of now only Intel 4 is in production with launch of Intel 3 over the year 2024.

 

[Fred Chen]

Leading-edge DRAM is roughly a node a year, although it might be slowing down now.

 

[nghanayem]

Leading-edge DRAM is roughly a node a year, although it might be slowing down now.

I always thought that it was so cool that the NAND and DRAM guys did a new node every 12-18mo. Even if the uplift from say 1a to 1B isn't crazy huge, the fact the follow through is so fast still makes it crazy impressive.

 

[MKWVentures]

We have no idea on success yet. Delivery means shipping a product to end customers. Not "manufacturing ready". Intel has committed to shipping Intel3 by June 30th and 20A by Dec 31st. 18A by Mid 2025. Lets see what actually ships and not get hung up on Intel claiming success or haters saying "Intel never delivers". If they ship the 3 half node processes in the next year, it is a success from a nodes point of view.

Strategy and financials are a different story.

 

[Xebec]

We have no idea on success yet. Delivery means shipping a product to end customers. Not "manufacturing ready". Intel has committed to shipping Intel3 by June 30th and 20A by Dec 31st. 18A by Mid 2025. Lets see what actually ships and not get hung up on Intel claiming success or haters saying "Intel never delivers". If they ship the 3 half node processes in the next year, it is a success from a nodes point of view.

Strategy and financials are a different story.

Did they actually commit to 18A by Mid 2025? Everything I’m seeing indicates Intel products Q4 2025. (In line with their usual annual releases around Oct-Nov).

I am pretty optimistic though - Intel 3 and 20A sound really healthy..

 

[TBiggs]

If you listen to the Ian interview of Pat it was interesting how Pat was explicit on the company having bet the future product line and the company on the node execution.

Somehow everyone here is shocked, but is Intel’s bet any differerent than Apple or TSMC? All the eggs are there miss one cycle or in Intels 10nm case multiple and the whole thing is a disaster. Can you imagine a September without an iPhone launch ?

Intel seems like it learned from the past. One thing I have learned is there is no corner on tools, chemicals or secret thing only the Taiwan, Korean or Americans know. What kills a company or a process or recover is the management and culture. Sounds like Intel doing a lot of talking but behind that they also realized their errors. If I was a betting man I’d bet Pat will deliver the technology side. The product is more complicated.

 

[MKWVentures]

Did they actually commit to 18A by Mid 2025? Everything I’m seeing indicates Intel products Q4 2025. (In line with their usual annual releases around Oct-Nov).

I am pretty optimistic though - Intel 3 and 20A sound really healthy..

they haven't committed anything that I know of. Lead 18A product is datacenter product so 18A isnt done until it is done. Saying "process is ready but product has issues means nothing on lead product.

That said, I believe the Clearwater forest launch goal is Q2 2025 ...

 

[TBiggs]

they haven't committed anything that I know of. Lead 18A product is datacenter product so 18A isnt done until it is done. Saying "process is ready but product has issues means nothing on lead product.

That said, I believe the Clearwater forest launch goal is Q2 2025 ...

I have often wondered about Intel product struggles and what Intel product success at TSMC compare. Is it a product / design issue or is it an IP / PDK issue or a process issue.

The track record of TI at TSMC on leading edge is remarkable for so many companies, I wonder if Intel has matched it there too. If they have that would suggest the problem is the Intel product teams but the Fab side.

 

[Daniel Nenni]

If you listen to the Ian interview of Pat it was interesting how Pat was explicit on the company having bet the future product line and the company on the node execution.

Somehow everyone here is shocked, but is Intel’s bet any differerent than Apple or TSMC? All the eggs are there miss one cycle or in Intels 10nm case multiple and the whole thing is a disaster. Can you imagine a September without an iPhone launch ?

Intel seems like it learned from the past. One thing I have learned is there is no corner on tools, chemicals or secret thing only the Taiwan, Korean or Americans know. What kills a company or a process or recover is the management and culture. Sounds like Intel doing a lot of talking but behind that they also realized their errors. If I was a betting man I’d bet Pat will deliver the technology side. The product is more complicated.

Apple has repeated nodes before on Samsung and TSMC. They improve the node but it is still considered the same node. They did two SoCs at N5 (A14 and A15) and one at N4 (A16). Apple will be on N3 for a couple as well (A17 and A18 for sure and maybe A19 depending on N2).

 

[hskuo]

I would like to see intel IFS D0 chart like the old one which I have missed for a decade from intel. D0 chart is better index for success. It would be like the ones tsmc published in the past.

 

[cliff]

C'mon guys. How many of you actually have taped out on time... and you aren't fighting physics.

Ever hear of the Pi factor? I use this multiplier on my software developers. It is accurate to 1.7%.