Intel 2020s Process Technology Roadmap: 10nm+++, 3nm, 2nm, and 1.4nm for 2029 | Hacker News Reader

Intel 2020s Process Technology Roadmap: 10nm+++, 3nm, 2nm, and 1.4nm for 2029

fuse.wikichip.org

59 pointsrbanffy19 comments

Other discussion: https://news.ycombinator.com/item?id=21770622

Meanwhile, in reality, Intel rerelease 22nm chips.

https://www.tomshardware.com/news/intel-resuscitates-22nm-ha...

Revive your brand through wild slides and sell high profit old products. Interesting strategy.

"WikiChip is happy to see Intel remains committed to furthering Moore’s Law for the foreseeable future with an ambitious roadmap."

What is Wikichip? Some sort of Intel mouthpiece? How can anyone belive an Intel roadmap at this point. At best it's fiction, at worst security fraud (according to the Matt Levine definition)

Literally. What even is 10nm+++?

Only thing that needs to matter anyway is performance in a suite of benchmarks

don't forget powerdraw.

With Intel it is always important to ask if the powerdraw includes or excludes the fridge hiding under counter*

* https://www.tomshardware.com/news/intel-28-core-cpu-5ghz,372...

"relentless innovation", or so the slide says.

The first part of the roadmap is particularly interesting. Heading into 2021, you will notice that Intel has a new, unannounced, process “10nm+++” (note the three pluses). This is a further refined version of its 10nm node. Presumably, this will be used for Intel’s Sapphire Rapids CPU which is also planned for 2021. Intel’s 10nm, including 10nm+, has been rather underwhelming as far as the performance goes with only a single SKU exceeding 4 GHz.

Wikichip is neutral. I follow David Schor on Twitter and his coverage of the semiconductor industry is both informative and unbiased.

Btw, here’s the full paragraph from Wikichip: “All in all, WikiChip is happy to see Intel remains committed to furthering Moore’s Law for the foreseeable future with an ambitious roadmap. It remains to be seen whether Intel can execute on this roadmap.”

14, 10, 7, etc do not matter anymore [1], because there is no metric that could correspond to that number directly. It’s all marketing nanometers. I’m all intel, but how does one estimate bs like “1.4nm for 2029”? By dividing 14nm by years passed since 2019?

[1] https://en.wikichip.org/wiki/10_nm_lithography_process

"Did I ever tell you the definition of insanity?"

Isn't that just marketing and wishful thinking? Not that long ago they were "predicting" 10nm process for 2016 [1].

[1] https://mr-uploads.s3.amazonaws.com/uploads/2016/02/Series-6...

Thank you, based-AMD.

Once you can process high-definition video there is little need for further performance in mass-market processors. Progress thereafter would be mainly in lowering power consumption, reducing cost, and greater integration of more functions on chip to lower parts counts.

Therefore, the business drivers must be mainly server applications in the data center.

When we can process HD video, we come up with 4K. When we do 4K, we come up with 4K HDR. When we are done with that, we'll go for higher FPS. Or surround video. And then in stereo. And then do it with voxels and multiple depth channels from different PoV's.

And we didn't even start with intelligent assistants that don't require sharing your data with a cloud-based entity.

I thought once my 20" 1200x1600 8-bit color CRT was all I'd ever need.

> Once you can process high-definition video there is little need for further performance in mass-market processors.

In the grand scheme of things high-definition video is an extremely low bar for consuming and archiving content. It's totally flat, distorted, has a fixed viewpoint, a tiny field of view, super low dynamic range, etc. It doesn't match what you would see if you were there in the slightest. It's closer to looking at a photography than to being there. We should strive to do better than that. Decoding high-def video is mainly done on the GPU nowadays anyway.

Current video is a pretty good match for the needs of advertisers, who are the major source of funding for video. There is relatively little market for video that is not either free or heavily subsidized by advertisers.

TV manufacturers are counting on 8K, OLED, and "smart TV" functions to keep volumes flat by encouraging replacements of what is a durable consumer good. LCD TVs seem to be quite long-lived.

Production costs go up a lot when you increase the resolution because then you can see all the details they don't want you to see.

What about software developers, who want the fastest possible compiles? Too small of a market to care?

Reply on news.ycombinator.com