Taiwan Semiconductor Manufacturing Co. (TSMC) will produce 3 nanometer chips at its second, still-under-construction plant in Kumamoto, Japan, the company announced Thursday.
Just a reminder for anyone that thinks 3nm chips means the transistors themselves are only 3nm, they are bigger than that. 3nm is the marketing name for the fab process they are using.
Close, except it’s not a marketing term. It’s part of a published IEEE standard.
The actual gate pitch and metal pitch vary by manufacturer and process type.
From Wikipedia:
The term “3 nanometer” has no direct relation to any actual physical feature (such as gate length, metal pitch, or gate pitch) of the transistors. According to the projections contained in the 2021 update of the International Roadmap for Devices and Systems published by IEEE Standards Association Industry Connection, a 3 nm node is expected to have a contacted gate pitch of 48 nanometers, and a tightest metal pitch of 24 nanometers.[12]
If I’m not entirely mistaken there is still some basis to the nanometre number, it just doesn’t refer to the actual smallest feature size or gate pitch anymore. Basically in the mid-2000 Dennard scaling stopped working and ever since the nanometre numbers are “made up”. Dennard scaling was how most progress was made by just shrinking transistors. But that doesn’t mean just because Dennard scaling doesn’t work anymore there is no progress, it’s just harder to achieve. So the semiconductor manufacturers just continued naming their fabrication methods as if Dennard scaling still worked. So basically a modern “3nm” process is equivalent in some way to what would theoretically be possible if you had an actual 3nm process.
They should count up by some benchmark. If x/mm^2 doesn’t capture the improvement anymore, and they aren’t shrinking things much anymore, benchmark some common output.
It’s not necessarily about transistors/mm^2, there is also power consumption and clock frequency. Back in the mid-2000’s clock frequencies stopped just under 4GHz and then went down for a few years before going back up to way past 4GHz in the last ten years or so.
Yup that’s exactly what I’m talking about. They need a benchmark for what it can do, not the size of a part or a made up size for marketing. Or just disconnect the specs from the marketing name entirely.
Its at least somewhat based on the transistor density increase they get from other techniques right? Like “3 nm” is the equivalent transistor size they’d need to get the same transistor density using 2005 chip design.
IEEE rebranded 802.11ax as wifi 6 because… Marketing
Minor correction: The standard is IEEE, but it was developed by the WiFi Alliance (who make their money by certifying devices as meeting the WiFi 6 standard). It’s a pretty fair marketing strategy though. Normal users aren’t going to notice 802.11ac vs 802.11ax
Just a reminder for anyone that thinks 3nm chips means the transistors themselves are only 3nm, they are bigger than that. 3nm is the marketing name for the fab process they are using.
Close, except it’s not a marketing term. It’s part of a published IEEE standard.
The actual gate pitch and metal pitch vary by manufacturer and process type.
From Wikipedia:
So, the IEEE has an actual norm for marketing speak.
Which, honestly, ought to happen more often.
If I’m not entirely mistaken there is still some basis to the nanometre number, it just doesn’t refer to the actual smallest feature size or gate pitch anymore. Basically in the mid-2000 Dennard scaling stopped working and ever since the nanometre numbers are “made up”. Dennard scaling was how most progress was made by just shrinking transistors. But that doesn’t mean just because Dennard scaling doesn’t work anymore there is no progress, it’s just harder to achieve. So the semiconductor manufacturers just continued naming their fabrication methods as if Dennard scaling still worked. So basically a modern “3nm” process is equivalent in some way to what would theoretically be possible if you had an actual 3nm process.
They should count up by some benchmark. If x/mm^2 doesn’t capture the improvement anymore, and they aren’t shrinking things much anymore, benchmark some common output.
It’s not necessarily about transistors/mm^2, there is also power consumption and clock frequency. Back in the mid-2000’s clock frequencies stopped just under 4GHz and then went down for a few years before going back up to way past 4GHz in the last ten years or so.
Yup that’s exactly what I’m talking about. They need a benchmark for what it can do, not the size of a part or a made up size for marketing. Or just disconnect the specs from the marketing name entirely.
Its at least somewhat based on the transistor density increase they get from other techniques right? Like “3 nm” is the equivalent transistor size they’d need to get the same transistor density using 2005 chip design.
And the number keeps going down because… That’s good marketing. IEEE rebranded 802.11ax as wifi 6 because… Marketing. They can do it too.
Minor correction: The standard is IEEE, but it was developed by the WiFi Alliance (who make their money by certifying devices as meeting the WiFi 6 standard). It’s a pretty fair marketing strategy though. Normal users aren’t going to notice 802.11ac vs 802.11ax