By: --- (---.delete@this.redheron.com), July 30, 2021 9:38 am
Room: Moderated Discussions
Heikki Kultala (heikki.kult.ala.delete@this.gmail.com) on July 29, 2021 11:18 pm wrote:
> Doug S (foo.delete@this.bar.bar) on July 29, 2021 5:44 pm wrote:
> > None of it really matters, since the process names have nothing to do with a physical dimension
> > anywhere in the design. It is just a placeholder for "2x the transistors in the next generation"
> > but we aren't even seeing that lately as TSMC only got 1.8x scaling on N5 and 1.7x on N3
> > - but TSMC wasn't calling those 5nm and 3nm, it is mostly outsiders doing so (maybe TSMC
> > does as well, but probably only because outsiders referred to them that way)
> >
> > Who knows what TSMC will call the stuff below N2, will it be N1.4 or P1400 or just
> > choose another letter at random, multiply by 10, so X14 then X10 and so on.
>
> TSMC did not get 1.8x scaling on N5. In reality (by synthesizing any
> reasonable piece of logic than does something) it's much worse.
>
> Or, lets say that TSMC might have gotten 1.8x for single best-case standard cell
> component type for their marketing materials, but TSMCs customers get MUCH LESS
> than 1,8x for their real-world designs that actually do something useful.
Can you provide evidence for this?
The only customers I know of for N5 right now are Apple and Kirin 9000.
Apple density is 134MTr/mm^2
Kirin density is 145MTr/mm^2
Kirin 980 (on 7nm) is 93MTr/mm^2
So Kirin scaling is ~1.56x. Not 1.8, but unclear how much of that is a consequence of 5G modem (both analog and digital IO scaling less well; and no time to POP optimize the 5G modem design)
Apple scaling is definitely worse, but that is again probably a scheduling issue, that Apple's priority was a drama-free transition to ARM Macs, and once again POP was sacrificed to meet that goal.
N5 scaling seems to have been a victim of the fact that
- the Huawei numbers never really became public because of the political shutdown
- the only other data point was Apple and
- people with an agenda were happy to convert a single data point (which is more about Apple's particular circumstances than about TSMC) into a generic claim
Coming up we have
Zen4 -- unlikely to be useful because AMD never pushes density
Xe -- ???
Dimensity 2000
Snapdragon 875 aka 888 was supposed to be, but moved to Samsung
A15
A15 seems likely to be the best datapoint.
> Doug S (foo.delete@this.bar.bar) on July 29, 2021 5:44 pm wrote:
> > None of it really matters, since the process names have nothing to do with a physical dimension
> > anywhere in the design. It is just a placeholder for "2x the transistors in the next generation"
> > but we aren't even seeing that lately as TSMC only got 1.8x scaling on N5 and 1.7x on N3
> > - but TSMC wasn't calling those 5nm and 3nm, it is mostly outsiders doing so (maybe TSMC
> > does as well, but probably only because outsiders referred to them that way)
> >
> > Who knows what TSMC will call the stuff below N2, will it be N1.4 or P1400 or just
> > choose another letter at random, multiply by 10, so X14 then X10 and so on.
>
> TSMC did not get 1.8x scaling on N5. In reality (by synthesizing any
> reasonable piece of logic than does something) it's much worse.
>
> Or, lets say that TSMC might have gotten 1.8x for single best-case standard cell
> component type for their marketing materials, but TSMCs customers get MUCH LESS
> than 1,8x for their real-world designs that actually do something useful.
Can you provide evidence for this?
The only customers I know of for N5 right now are Apple and Kirin 9000.
Apple density is 134MTr/mm^2
Kirin density is 145MTr/mm^2
Kirin 980 (on 7nm) is 93MTr/mm^2
So Kirin scaling is ~1.56x. Not 1.8, but unclear how much of that is a consequence of 5G modem (both analog and digital IO scaling less well; and no time to POP optimize the 5G modem design)
Apple scaling is definitely worse, but that is again probably a scheduling issue, that Apple's priority was a drama-free transition to ARM Macs, and once again POP was sacrificed to meet that goal.
N5 scaling seems to have been a victim of the fact that
- the Huawei numbers never really became public because of the political shutdown
- the only other data point was Apple and
- people with an agenda were happy to convert a single data point (which is more about Apple's particular circumstances than about TSMC) into a generic claim
Coming up we have
Zen4 -- unlikely to be useful because AMD never pushes density
Xe -- ???
Dimensity 2000
Snapdragon 875 aka 888 was supposed to be, but moved to Samsung
A15
A15 seems likely to be the best datapoint.
