The industry needs a new naming scheme for these processes.. it's misleading at this point and basically tells us nothing about the density itself. For example Intel's 10nm process is nearly equivalent to Samsung's 5nm DUV process, Intel's 7nm process is 35% more dense than TSMC's 5nm DUV. So where exactly does TSMCs 6nm fall? Who knows because the number is completely arbitrary.

Ampere@6nm, bring it on !

Denial:

The industry needs a new naming scheme for these processes.. it's misleading at this point and basically tells us nothing about the density itself. For example Intel's 10nm process is nearly equivalent to Samsung's 5nm DUV process, Intel's 7nm process is 35% more dense than TSMC's 5nm DUV. So where exactly does TSMCs 6nm fall? Who knows because the number is completely arbitrary.

Completely agree about misleading Information about the chip production technology processes from TMSC. Intel has setup the standard here in chip production where numbers in "nm" process are real, everyone else are just cheating on this matter, just to "sound" better for the people and to sell more of the chips when they're out

Denial:

The industry needs a new naming scheme for these processes.. it's misleading at this point and basically tells us nothing about the density itself. For example Intel's 10nm process is nearly equivalent to Samsung's 5nm DUV process, Intel's 7nm process is 35% more dense than TSMC's 5nm DUV. So where exactly does TSMCs 6nm fall? Who knows because the number is completely arbitrary.

What matters is price per transistor of small/medium/large/jumbo chips. Power efficiency. Max clock. Density itself is not really important if you are not building 20B transistor chips. That's why I like way Samsung described their new process... "20 percent lower power consumption or 10 percent higher performance" They included: "up to a 25 percent increase in logic area efficiency" for "density" That means you take old design and it will eat only 0.8x of original power (no Clock bump). In same area, you can place 1.25x more transistors => 1.25 x 0.8 = 1 (No increase in power consumption.) But since chip now has 25% more transistors, it potentially has 25% higher performance at same power. Those values can clearly tell what new process can do without knowing exact density. = = = = I do not care about nm written on the product box, I care about what it does.10/7/6/5nm, who cares about intel/TSMC/Samsung numerical scheme when 7 proves better than 10 and is cheaper. A: "Intel's 7nm transistors are smaller than Samsung's!!!" B: "But Samsung's 7nm transistors work."

So now they are gonna spend next 50 years like this (7nm,6nm,6.5nm,6nm,5.5nm)...Show us next revolution

do u think we reach 1000pm or picometer?? we not to far away from it

devastator:

do u think we reach 1000pm or picometer?? we not to far away from it

If it is achieved, it will probably be called 0.x nm cos most people are just too stupid. Just like phone batteries where people say 3thousand milliamp hour instead of 3amp hour.

Denial:

The industry needs a new naming scheme for these processes.. it's misleading at this point and basically tells us nothing about the density itself. For example Intel's 10nm process is nearly equivalent to Samsung's 5nm DUV process, Intel's 7nm process is 35% more dense than TSMC's 5nm DUV. So where exactly does TSMCs 6nm fall? Who knows because the number is completely arbitrary.

Surely it simply tells you the EUV process and how that compares to that company's previous EUV process density as TSMC have done here, how is that misleading? You mention Intel 7nm process but that is irrelevant at the moment as nobody yet knows whether that will be running anytime in 2021 and at what capacity.

sunnyp_343:

So now they are gonna spend next 50 years like this (7nm,6nm,6.5nm,6nm,5.5nm)...Show us next revolution

It's called qunatum computing, and they are already working on / with it.

Fox2232:

What matters is price per transistor of small/medium/large/jumbo chips. Power efficiency. Max clock. Density itself is not really important if you are not building 20B transistor chips. That's why I like way Samsung described their new process... "20 percent lower power consumption or 10 percent higher performance" They included: "up to a 25 percent increase in logic area efficiency" for "density" That means you take old design and it will eat only 0.8x of original power (no Clock bump). In same area, you can place 1.25x more transistors => 1.25 x 0.8 = 1 (No increase in power consumption.) But since chip now has 25% more transistors, it potentially has 25% higher performance at same power. Those values can clearly tell what new process can do without knowing exact density. = = = = I do not care about nm written on the product box, I care about what it does.10/7/6/5nm, who cares about intel/TSMC/Samsung numerical scheme when 7 proves better than 10 and is cheaper. A: "Intel's 7nm transistors are smaller than Samsung's!!!" B: "But Samsung's 7nm transistors work."

That's kind of why I said its misleading? Not only does the nm number not tell us anything about the density of the process but it doesn't even tell us how it compares to previous processes or a competitors process in any way shape or form. My argument is is merely that the "x nm" label is pointless and needs to go. I don't really care what you replace it with - want to factor in switching speed, fine, want to factor in power consumption - go ahead. Don't care. Stop using x nm. It basically means nothing and yet I constantly see people compare the x nm numbers directly to other vendors.

Is it just me or are the foundries starting to have a reverse e-peen contest? Mine is smaller than yours its says right in the name, 6nm!

fantaskarsef:

It's called qunatum computing, and they are already working on / with it.

We are a long ways off from quantum computing becoming mainstream and frankly it's not even considered a good replacement due to inherent error rates but I wouldn't go as far to say never. There will be a lot of things happen before you and I are typing on quantum computers if that ever occurs. We should eventually see pure silicon replaced by some other substance say silicon nanosheets, graphene, gallium nitride or whatever they come up with but I suspect that may only take us as far as 1nm. I expect us to hit a wall on shrinking CPU process nodes around 1-3nm. We may get smaller for things like flash and DRAM but logic circuits that need to run at high frequencies will hit some real physical limits.

JamesSneed:

We are a long ways off from quantum computing becoming mainstream and frankly it's not even considered a good replacement due to inherent error rates but I wouldn't go as far to say never. There will be a lot of things happen before you and I are typing on quantum computers if that ever occurs. We should eventually see pure silicon replaced by some other substance say silicon nanosheets, graphene, gallium nitride or whatever they come up with but I suspect that may only take us as far as 1nm. I expect us to hit a wall on shrinking CPU process nodes around 1-3nm. We may get smaller for things like flash and DRAM but logic circuits that need to run at high frequencies will hit some real physical limits.

Sunny was asking about revolution. I thought that quantum computing, with it's above binary logic would be such a revolution, changing interposers or shrinking down to 1nm essentially is "only" evolution. You are right though, we won't put our hands on a quantum computer anytime soon, except if one of us works with science / IBM / Intel / whoever works on it first hand. Although, 50 years in the future was meant... think back 50 years and see what we got now, it's a revolution by itself if we look at that.