I saw this today and thought it was a very interesting article. It is about placing 100 billion transistors onto a computer chip.
4 Likes
In 1965, Mr. Gordon Moore made a prediction about the pace of technological progress that proved to be quite prescient. He noticed that the number of transistors in integrated circuits roughly doubled every two years.
I try to keep up with technology like you all do, but size speed capacité is developing so fast as one comes out something replaces it
This morning looking at an apple mac chips m2 and then read about the next générations
The M6 chip will be the first built on a 2-nanometer process instead of the 3-nanometer process that Apple has used for the last several chip generations. Rumors suggest Apple will use TSMC’s N2 process. Compared to the 3nm process, the 2nm process cuts down on transistor size so more can be packaged on a chip. Decreases in node size typically bring improved processor speeds and better power efficiency.
Not so long ago we were raving over the move to never saw the 186 (went away too quick to buy any) 286, then 386, (i had just set up a classroom equipped with 12) then 486 came so quickly to replace them
If asked tomorrow what would I buy not sure I am capable of answering any more.
4 Likes
We have reached some limits. The 3-4Ghz barrier is still with us . We are circumventing it with multiple processors and cacheing tricks.
4 Likes
It makes me think of two philosophical questions : How many angels can dance on the head of a pin? and Isn’t it amazing that the faster the computers get it doesn’t make us smarter, it just increases the speed at which we can make mistakes?
5 Likes
Also reminds me of the moral issue
"There are two ways to get enough: One is to accumulate more, and the other is to desire less.”
Bishop Fulton Sheen
5 Likes
amen brother
1 Like
Moore’s law was originally intended as a joke. Apparently they took it seriously.
But… we’re at such a point now, why do we still need more? Is it that important to fit more on a single die?
Wouldn’t it be better to invest in cooling technology, so the chips last longer and can be made faster?
2 Likes
It seems humans will do anything and everything that can be done … regardless of whether it is useful or a priority.
It is a sort of evolutionary strategy to explore every avenue.
We may have reached the point where a different strategy would be advantageous… but you cant direct people what to do .
Every time a door opens, we all run through it.
2 Likes
The old jokes are the best, dont need explain but it was true for a long time in computing anything over 2 days is so last year !
1 Like
If this is true, it sounds like good progress.
“50% better performance and consuming 70% less energy than current generation.”
2 Likes
Do you mean a cpu will run with more Ghz … that is real progress… we have been at that 3-4 Ghz limit for more than 10 years.
1 Like
And based on what I read we well might be stuck there a some time to come.
Apparently, increasing the GHz beyond the current limit causes too much heat.
“The physics wall is where the chip melts or misfires if you push clocks higher.”
The 50% performance boost came from;
" The 50% performance jump comes from architecture, not frequency:"
Disclaimer: Info posted obtained from interface with an AI.
2 Likes
I dont see how architecture can make it crunch numbers faster.
but I could use a 50% performance jump, especially if it came with more memory
1 Like
How about this meaning better functions?
1 Like
Yes, I guess you can improve the design of an ALU, or improve caching, or speed up memoryaccess. … but the raw speed of the cpu has reached a limit.
That is why my 15 year old computer is still a respectable speed.
1 Like
Info from AI.
" The 50% performance jump comes from architecture, not frequency:
- Backside power delivery → lower resistance, lower voltage drop, faster switching.
- Nanosheet transistors → higher drive current, less leakage, more IPC.
- Better materials → cooler operation → more stable high‑performance switching.
- Denser EUV layout → shorter wire paths → faster logic.
Blunt version: They made each clock cycle do more work while using less power. GHz stays the same — the physics wall doesn’t move."
So I asked, “what is the physics wall?”
AI response.
" Short mode: The physics wall = the point where silicon can’t switch faster without breaking.
Blunt version: Past ~4–5 GHz, transistors need more voltage, leak more current, and make too much heat. Those three things explode exponentially, not linearly — so you hit a hard ceiling.
The four pieces of the wall:
- Switching speed — silicon can’t flip states much faster.
- Voltage wall — higher GHz needs higher voltage → runaway heat.
- Leakage — tiny transistors leak current at high frequency → instability.
- Thermal density — big CPU cores generate too much heat in too small an area.
One‑liner: The physics wall is where the chip melts or misfires if you push clocks higher."
1 Like
Sounds like the wall is a property of Silicon.
Maybe new materials could change it.
1 Like
Good question. I thought maybe if the silicon got too hot maybe it would start to soften or melt. So I was curious and wanted to learn more so I ask AI,
Is it because if the silicon gets too hot it starts to melt? And the response was;
" No — it’s not about melting.
Silicon won’t literally melt at 4–5 GHz. The problem hits way before melting:
The transistors stop switching correctly because voltage, leakage, and heat spike too fast. They misfire, flip bits wrong, and the CPU becomes unstable long before any physical melting.
Blunt version: It’s an electrical limit, not a melting limit."
2 Likes
I found this
“No, Gallium Arsenide (GaAs) does not have a 4 GHz limit. In fact, it easily operates at frequencies ranging from 30 MHz well into the microwave and millimeter-wave bands, often reaching up to 250 GHz. [1, 2, 3]”
So it is an issue confined to Silicone
Why dont they use something else for cpu chips?
1 Like
Very good questions Neville. I am learning a lot about CPUs and the structure of building progress and the limits of the architecture of of them. I know you dis-trust AI and I know way too little to be of any help. I don’t mind using AI so I will give what AI displayed back to me. I can give longer explanations if you desire it.
"Silicon is used because nothing else works as well for mass‑produced, tiny, fast, low‑power transistors. Alternatives exist, but each one fails in some critical way.
- Gallium nitride — great for power electronics, terrible for dense logic. You can’t pack billions of fast switching transistors into it.
- Graphene — amazing mobility, but no bandgap → can’t make proper on/off switches → useless for CPUs.
- Carbon nanotubes — promising, but manufacturing defects ruin yields. You can’t build a reliable 100‑billion‑transistor chip with random broken tubes.
- III–V semiconductors — fast, but impossible to fabricate at silicon scale and cost.
Silicon is the only material we can manufacture with insane precision, low defects, and microscopic features at global scale. Everything else either can’t switch cleanly, can’t be mass‑produced, or can’t survive the heat."
Edited by me to remove some of the boldness.
2 Likes