Most ICs will be produced using DUV processes for the indefinite future until the costs of EUV can be drastically reduced and throughput increased.
Unless we have developed Star Wars holograms for common use, computers are pretty much topping out on anything that we need to do with them.
At my first job in semiconductors we used 2” wafers.
Once chiplets are worked out for gaming GPUs the prices will stabilize and quit going through the roof.
Quantum computing might not double the number of transistors, but at some point it might allow Moore’s law to catch up on power.
Curious if this takes into consideration the cost of inflation and other cost increases due to global demand for materials, supply chain issues, and the whoke host of orger global issues?
I can’t even figure out how to speed up my computer’s download rate.
Asian guy who thinks he’s Michael Douglas circa 1998 looks past his nose at IBM Eagle.
I’d say the biggest problem we have with computers these days is outrageously bloated and buggy code. MS-Windows has been a notorious example of that for a long time, but it’s also getting really bad in the Linux world as well for the past decade. I’m darn close to abandoning KDE for my desktop, since it has really become a bloated mess. I have an i9 CPU with 16 cores and 32GB of ram, and find it hard to believe how bad things have gotten.
I agree with the premise of the article that Moore’s Law is pretty much dead, until a major change is made to the architecture. The incremental improvements in hardware are nothing like we saw in years past. It would also be nice if programmers started taking SMP into acount when writing software. Seems like most of the time my CPUs are idle. Might have 1 or 2 threads actually using the CPU, but not as much as you’d think.
I mean, it's like the R&D has come to a crashing halt!
/lame attempt at humor
The time has come for a paradigm-shift.
Quantum dots?
Neutrino-based computing?
Regards,
Buy them! Trade them! Swap them! Steal them!
You can even wipe them! Like with a cloth!
Correct!
Without EUV AND careful planning of high speed circuits, Moore’s Law died after 14nm.
The size of individual silicon atoms (around 0.2nm) would be a hard physical limit (with circuits one atom wide), but its behaviour becomes unstable way before that because of electrons going down one wire and showing up on another at scales way above that. Electrons are weird.
Alternative semiconductors such as gallium nitride and silicon carbide cope much better at higher temperature so clock speeds can be pumped.
RISC chips also show promise to at least continue Moore for a while longer.
NVIDEA is close to the bleeding edge of chip fab (4nm) but given current tech and materials I am afraid they may be right. TMSC claims to be very close on a 1nm process and IBM has a 2nm process. These is just an order of magnitude from physical limits for silicon being 0.2nm and spooky behavior will show up way above that size, and implies the Moores law curve kinda like this: (just the shape, ignore the labels)
Heard that the ‘times of smaller semiconductors were done at 20nm’.
The 3d printing and so forth at 1nm will expand a lot more.
JENSEN HUANG! LMAO! No wonder nVidia tried to produce chips in chinkland. My GOD, what are they thinking!! Get that a$$hole outta control!