Comments by "kazedcat" (@kazedcat) on "Asianometry" channel.

@awesomewav2419 AMD will not buy from Intel fabs for leading edge. AMD is heavily invested in TSMC's technology stack. Switching everything to Intel's Technology stack would take a decade of work.
1
No SRAM is far larger. It will not reach DRAM level of cell density anytime soon.
1
Microgravity could ruin the crystal structure.
1
 @M4nusky Not exactly a conductor but without a gap a small amount of energy like the energy release when you clap your hands can flip the state of a transistor. So it is a very unstable switch.
1
@RooMan93 Security code is much more expensive in man hours. A developer who understand discrete elliptic curve is very rare.
1
@RooMan93 The industry already have a solution for that a security library. This way underpaid developer will not be touching any security codes and just use API calls to access them.
1
Real capitalist only talk in terms of profit. Those other ideology likes to attach themselves to capitalism to stay relevant. He like to frame things in terms of cost, marketability, and returns. That is the capitalist way of talking.
1
It's not necessarily true. In fluid dynamics if you scale everything down the Reynolds number changes. So it needs to be proven that there is no similar factor happening in microelectronics.
1
Search YouTube for EUV Scanner
1
Short answer yes. The long answer leakage is an issue but there are a lot of tricks that can be applied to mitigate the problem. One of this tricks is switching transistor architecture from FINFET into GAAFET
1
No ASML makes the equipment's(oven) they install it on foundries fab(kitchen) the foundries figure out the new recipe using the new oven then they cook the new recipe for the fabless chip maker like Nvidia and AMD. Intel cook for themselves because they have their own kitchen(fab)
1
@Erik-rp1hi Yes TSMC figures out the design rules then industry partners use this design rules to design chips.
1
Sadly it is not that simple. Signal propagation is still bound to RC delay. So even if charge mobility is instantaneous. RC delay will still limit the frequency of a processor.
1
 @ralanham76 We are already taking advantage of quantum tunneling. The power wall is mostly a classical physics problem not a quantum physics problem.
1
Atomic deposition has been around for a long time. With it you can deposit a layer of material 1 atom thick. Mirrors also needs a perfect shape so it is highly likely that the aberrations specs are derive from the mirrors geometric shape.
1
@petergerdes1094 First 2nm node process actually has a 20nm wide transistor, silicon latice are 0.2 nm apart so a 2nm transistor has 100 silicon atoms across. Second we are not using bulk silicon property, Field Effect gets stronger the smaller the features are and the only thing that limits things getting smaller are leakage current and quantum tunneling. Both of this limitation has tons of tricks we can use to mitigate the problems. In conclusion we are nowhere near a hard wall that stop shrinking transistor but we are already on a very steep hill and climbing it gets more and more expensive.
1
@petergerdes1094 CMOS only has either N or P type. They work because the field effect of the metal gate directly creates holes or electrons. The doping is done to manipulate the threshold voltage but with enough gate voltage it is possible to force undoped semiconductor to have the same channel effect. Field effect transistor does not require bulk material mechanics. We have laboratory devices that work using a single layer of 2d lattice material. In fact the field effect works much better using 2d material than bulk silicon semiconductor. This is the direction we are going after moving to the nano ribbon silicon gate all around FET The next step is replacing the silicon channel with 2d ribbon material. This is a field effect transistor using a single layer of 2d molecular lattice.
1
Yes this machines has been planned for decades. The industry itself have schedules and roadmaps that everyone agrees on and they have regular meetings that updates the schedules and roadmaps With any unexpected factors
1
nzoomed There are some research on using multiple e beam to speed up scanning but I don't think it can compete with the speed of photolithography. Using it to create mask is okay because mask is reuse several hundred times so output is amplified.
1
@nzoomed Photoresist is driven by chemical reaction. Faster reaction means less precision which is what you want by using e beam.
1
No a transistor that does not turn off is not efficient at all.
1
They probably have IP licensing revenue but that is peanuts compared to their other businesses. Their gaming console revenue alone would dwarf their IP revenue and that console chip business is the lowest margin segment of their income. AMD has a lot of businesses.
1
@lucasfernandesgrotto6279 AMD likes to bet on technology sometimes it works sometimes it doesn't. Bulldozer was failure but they are trying something new with that architecture. Chiplet on CPU works and it is now giving them advantage over Intel. Chiplet on gaming GPU is kind of not working but maybe they can make them work later. HBM was one of those bet and the technology works but the financial side was not working for them then.
1
Dilute the brine with sea water 100:1
1
@volkhen0 There is not enough ice to cover the entire earth with additional water 70 meters high. Ice is less dense than water so if your projection is true then the equator should be covered right now with more than 70meters of ice. Just look at the surface area of the earth covered in ice that is a tiny fraction compared to the entire surface of the earth. Also yes a significant portion of the polar ice cap has melted already. And don't expect the sea level to continue its rise indefinitely the volume of ice on earth is not infinite.
1
You copy pasted your comments. Did you pay for the IP?
1
@paulmichaelfreedman8334 Laser is still photolithography unless you are using laser to directly burn spots in the paper to print an image.
1
@homelessrobot But planck scale is not digital yes they are quantize but quantum object does not behave like classical object. So instead of bits you are dealing with qubits. And when dealing with qubits entanglement and decoherence becomes a problem which is a lot harder than analog.
1
@homelessrobot Think of it this way. digital means integer (Z). Analog means real numbers (R). Quantum objects are complex integers (Z×Z). Yes they are countable but they are a ZxZ set instead of just Z.
1
The Internet is content theft. Ban the internet.
1
Even with recycling there is still waste water. Reverse osmosis has two output stream. One output water with less impurities than input and the other output water that has more impurities than the input. You cannot feed back the dirty water back at the input because that will reduce the purity of clean output. In multiple stage purification you can use the dirty output of later stages as input for the early stages but the chain ends on the first stage where you have raw water as input and you have dirty water output that is less pure than raw water and ends up as waste water.
1
In addition to this they are manipulating the shape of the light. They need a dipole light source for increase resolution.
1
@tylerdurden3722 They use a mirror to focus the light into a circular beam. A second mirror to convert a circular beam into an annular beam. Then a third mirror to convert annular beam into a dipole beam. This is before the beam hit the mask . Then after reflecting from the mask there are a set of mirrors for magnification and aberrations correction.
1
They could not buy the machines to reverse engineer.
1