Monday, June 1, 2026

transistor pattern storage

it doesn't matter how fast the computing part of a chip is 
if it has to wait for the memory part of the chip to transfer data 
and those wait times seem short at a human time scale 
but are long in a transistor time scale 
and those wait times occur often as the chip works 
which is what holds computing performance back 

cpu stores data in its cache memory 
so that one part of the chip is computing transistors 
and the other part is memory transistors 

in a many core cpu 
each core could have its own cache 
but that also means many small pieces of cache 
which could be too little cache for some workloads 
such as ai, gaming or supercomputing 

so data transfer wait times that occur often and repeatedly 
and too many transistors that are used for cache memory 
while fewer transistors do computing 
make chips slow 
which holds back every company that rely on computing devices 
and that slows down the economy in every country 

but if each transistor that does the computing in the cpu 
together forms patterns 
and those patterns store data 
then that would replace the cache 
which would mean more computing power 
and more memory 

the transistors that do the computing in a cpu 
constantly switch between on and off 
so the memory patterns would be shimmering 
from transistors switching between on and off while they compute 

so it wouldn't matter if the transistors in those patterns 
are sometimes on or off 
since the pattern is the storage unit 
and later that data is stored on the hard drive 
so that the data remains if the computer is turned off 

so researchers don't need to develop some memristor cpu combo 
instead normal transistors could be used 
but they would have to be used differently 
possibly with some software 
controlling the memory patterns of the transistors 

transistor pattern storage would make transfer of data 
from the memory part of the chip 
to the computing part of the chip 
nearly instantaneous 
because the memory is stored as memory patterns 
on the transistors that do the computing 
which would make those chips extremely fast 
without needing to develop a new chip technology 

so data transfer that is nearly instantaneous within the chip 
and almost all the transistors of the cpu would do computing 
while patterns of those transistors store data 
then that would be truly next gen 

No comments:

Post a Comment

deniable operator training programme

I'm willing and capable  of starting and/or running  a deniable operator training programme  that trains deniable operators  in every co...