How does a cpu represent bits physically ?

Thread Starter

Oz V

Joined Sep 7, 2019
may be my question was asked earlier but i want info away from logical abstractions, in different contrast. START: a transistor can represent two states on and off means 1 and 0 means 1bit. so why we dont use only 32 transistors to represent a 32bit data. why there are logic gates which contains normally 2-8 transistors in each to represent 1 bit only in each gate, which commonly input 2 bits and output 1 bit. what type of input bits they are and why only one output then how we use that one output furthur, why we degrade 2 bit to one. END summary: I need a very lowest possible physical structure of cpu starts from a transistor to assembly, without any abstract terminologies. I read many tutorials on the web but i could nt understand. that why iam here.


Joined Jun 5, 2013
Bits are stored in various ways within a CPU.

A single transistor is indeed used in dynamic RAM. In this case, a charge is placed on the gate of the MOSFET to represent a 1 and no charge to represent a zero. But this charge degrades over time, so a very complex circuit is used to read and refresh the data periodically. Also, the access to this data is relatively slow.

For memory that needs to be accessed quickly and without refresh needed, we use static RAM. This is essentially a flip flop, a circuit using multiple transistors that can maintain a one of two states until it is explicitly changed.

And for bits that are not stored, but are part of a computation, the bits are simply the outputs of gates. The outputs depend on the inputs, so they do not store anything. If the inputs change, the output changes.



Joined Feb 24, 2006
The smart alecy answer is that bits are represented by a voltage levels, but I don't think you will get what you want until you adjust your expectations. There are basic circuits which are fundamental in the construction of digital circuits. Their physical implementation has changed over the last century, but the abstractions have remained relatively constant in form and appearance. This is only a suggestion but the circuits you need to be familiar with are:
  1. Astable multivibrator
  2. Monostable multivibrator (aka One-Shot)
  3. Bistable multivibrator (aks Flip-Flop)
When I was an undergraduate (ca. 1967) we built each of those circuits with vacuum tubes (12AT7's) and transistors (2N3904's), because, you know, transistors might just be a passing fad. Once you see the level of detail required of an actual circuit, you will greatly appreciate the abstraction, that gates and other functional blocks like multiplexers, decoders, counters, shift registers and ALUs(Arithmetic Logic Units) provide. I doubt that you could look at an ALU in circuit form and have even a clue about what it was doing.
Last edited:


Joined Oct 2, 2009
In a computer's CPU a bit is stored in a bistable circuit, i.e. a circuit that can be held in one of two states, LOW or HIGH.

The basic bistable logic circuit is a flip-flop.

Here is one implementation of a J-K flip-flop.



Joined Mar 14, 2008
a transistor can represent two states on and off means 1 and 0 means 1bit. so why we dont use only 32 transistors to represent a 32bit data.
It depends upon why the bit is being stored.
If it's in the CPU then it's stored in flip-flops so it can be rapidly changed and manipulated at the CPU speed.
If its in temporary storage (RAM) then it's stored on the gate charge of a signal transistor that can be rapidly changed (but not as fast as a flip-flop).
If its in semi-permanent storage (EEPROM or Flash) then its stored on the capacitance of a floating gate transistor, which can be changed only relatively slowly (and for a limited amount of cycles).
If its in permanent storage (ROM) then there is a physical link that is modified to store the bit on one transistor.


Joined Mar 14, 2008
*some of* Todays ICs contain Billions of transistors. I can't even wrap my head around that one.
I know what you mean.
I remember, way back when, that the announcement of a 1k-bit RAM with over a thousand transistors seemed amazing. ;)


Joined Jun 5, 2013
Hence the name “core memory”.

And it was not just the PDP-11, pretty much all computers of that era used it.

I worked for Digital at the time the VAX was released. It had an option for semiconductor memory. The price, I kid you not, was $1M for 1Mbyte! And that was 1979 dollars!



Joined Mar 19, 2019
That's why an 8Mhz 8088 w/ 16k of memory for $1000 was a "Good Deal" even if there were no real applications to run on it. Then came Visicalc and WordStar... And here we are today playing on the Internet.


Joined Apr 5, 2008

My first PC was an 10 Mhz 8086 "turbo" with 1 Mbyte of memory and an 40 MegaByte harddisk.
It was faster as the IBM pc's we had in the office.
It costed me Fl 5000 in 1986.
The color monitor, a NEC multisync II costed an other Fl 1500.