Hacker News Reader

Comment by Havoc | Hacker News Reader

Well that's annoying. How is anyone supposed to judge anything if they just make numbers up?

Smallest feature size no longer correlates with transistor density in recent years. So manufacturers just use the number to convey increases in transistor density.

See https://en.wikichip.org/wiki/technology_node#History

That's why they do it. The sizes of Intel's 10nm is similar to or slightly smaller than everyone else's 7nm. Compare https://en.wikichip.org/wiki/7_nm_lithography_process and https://en.wikichip.org/wiki/10_nm_lithography_process .

I think bits of SRAM per unit of area is a benchmark that some people use, as it corresponds to something that's meaningful and measurable.

That's a problematic metric as well because we don't know how much area the assist circuit takes up. Modern high-density SRAM cells cannot operate as is, they need an assist circuit to compensate for variations. For example for Intel's 10nm SRAM, they claim 77% area effiency (https://fuse.wikichip.org/news/525/iedm-2017-isscc-2018-inte...). But without those values, just bits/mm2 or so is problematic.

The problem is even that is flawed as things are going three dimensional.

So just start measuring in bits of SRAM per unit of volume?

Hopefully their customers are more sophisticated than designing their chip based on a single number.

Well, after a while or is sort of public knowledge that 10nm Intel is about the same as 12nm AMD, etc.

But yes, the industry needs standardized advertising practices badly.

>standardized advertising practices

Oh man, I wish this would happen in any industry...sadly, advertising doesn't seem to see any benefit. Why standardize when you can differentiate?

Uh, it's the other way around. 10nm Intel is 7nm global foundries. 22nm Intel is superior to 14nm global foundries.

Intel is the one that doesn't fudge their numbers.

Even Intel's numbers are just marketing, however.

Everyone is fudging the numbers...

SemiWiki has a good overview of Intel 10nm vs GF 7nm:

https://www.semiwiki.com/forum/content/7191-iedm-2017-intel-...

For a while I think it was the other way around: Intel's processes were, in practice, actually substantially less dense than their competitors at the same nominal process node size.

Intel still fudges their numbers. They just fudge their numbers less than everyone else.

How so?

The numbers don't seem to have much relation at all with all of the real process node numbers.

Reply on news.ycombinator.com