How Computer Memory Works - Computerphile 

10/10 pun

Oscar Miranda not a pun.

That reminds me working on my apple 2e and I was taking all the expansion cards out and a capacitor on the printer card shocked me then. Right after wards the system had issues booting and I remember lying on the floor hating life thinking I messed up the CPU..... Luckily it was because my ram was bad and before hand the system was running off the expansion card ram which I just removed lol.

I did find that quite validating, myself.

@@OM-el6oy why are you doing this

Holy, thanks so much. This is exactly the kind of book I always wanted to read, starting from small components such as the transistors, going through machine and low-level languages, and reaching up to high-level languages. I don't know why, but I always wanted to understand how it all works, and this looks like a very promising book to help me achieve that.

How are you liking it@@kif-zallrhat1870

its badass you know something that only a few people in the world know!

Not very useful…

It's basic set logic.

It's basic set logic.

Scott Blacklock .

And not a not.

MOST IMPORTANT POINT EVER!! (what you said, telling students to expect difficulties)

@@theultimatereductionist7592 Digital logic labs were (are) HORRIBLE

Absolutely. In the first few lessons of my first electronics course I was afraid my program director had forgotten to enroll me in some introductory classes, because I felt like I had missed something I should have known already.

What major? EE or ECE?

The cutaways to his face and his hands getting in the way and the wires eclipsing each other...no! Ben Eater, save us!!

@Archer True

Bits for representing information and Boolean Algebra for electronic circuits are both the work of Clauden Shannon

@Archer It's a simple idea based on simple question: How to store information in electricty. Answer is simple but is a bit tricky.

It's weird but all the technology and logic was around for a while before someone connected all the dots. The book Code by Charles Petzold is amazing for outlining a lot of the history.

Archer, above, explained the process well. But if you want one liner slogan: No one and everyone. Different era and different people played different roles. Looking at what's called the 'end result' is misleading. Here the example for this misleading 'end result' is what you called 'this'. What is 'this'? How can you isolate 'this' from the other complementary knowledge?

SSD isn't persistent. If you leave it off long enough it'll loose its data.

Jeroen Bollen and this "long enough time" being multiple years?

***** Indeed.

it's memory... but is it --Random-- Access Memory? I think there used to be a thing before RAM that was more determined, is this not it?

emilnyb This is Random Access Memory. It allows you to access any memory cell at the same speed, regardless of the previous cell you tried to access. Technically storage like SSD is also *Random Access*.

+Joshua Whiteley Really? I thought I had an understanding as to how they worked, but this just seemed like a really confusing explanation. Whatever works for you I suppose, though.

Yeah!

Well minecraft physics are a bit non-realistic Still i like to play mcpe

8:28 a t-flip flop

@Baymax15: It is the S-R Latch. T-flip-flop toggles the output with each clock tick, and is more complex ;)

Dr. Steve, are you Mumbo Jumbo's father?

PC Systems developed slowly bit by bit over the course of 3 decades if not more and you're trying to learn system that was created in span of such a long time in low amount of time. As you learn new stuff it will fall in like a puzzle little by little. Give it time and you'll see your understanding grow

Then just buy a mac.

@@c25789 True true,buy a mac,break it cause its garbage,and u can see the chips and stuff and u can learn how memory works for yourself

You take your ram and put it into your socket... Turn on the computer and it works.

Even if you remove the input, the LED stayed ON

realityscorned yes, obviously coming at it from the CS side. UW-Madison ECE grad here. Did you notice he never turned off the power while pluging/unplugging the ICs? And the foil in the packaging used to align the pins into the breadboard? Oi!

@@bjornolson6527 even as 1st year graduates we had marks from Lab deducted if the invigilator caught us plugging/unplugging ICs without turning the power off.

@@EvilSapphireR I can't think of a way of how can that harm the IC. Especially just a simple TTL chip

I just wanted to say: didn't he just build a flip flop?!?!

@@andinomm thats technically a sr latch if im right

The reasoning behind this is the logic provided by the NOR gate. The basis of this idea starts with the fact that we have a 00 bit input into SR. Q is basically a way of describing the output. If we don't know any state of the circuit then saying the lower gates output is Q' is kind of unintuitive but if you assume you start with the R gate and say R is 0 then 0 NOR'd with a unknown state must be the opposite of whatever came in from the lower gate. With this reasoning, if we don't care what we were given and just call the output Q the output from the lower gate must be Q' (because both S and R is 0). if you instead change S to 1 the gates will assume a definite value regardless of what Q and Q' was.

Thank you, blockhet. This clarifies everything.

@@mahmudfasihulazam737 it does?

It's either a capacitor or a condenser, no such thing as a condensator.

KonstantinUb Capacitor was the word I was looking for. English is not my native language and haven't used the word in a very long time.

It's used in mine as well, although that's still no reason not to know the proper English word (especially if you're into electronics and CS!)

to my knowledge MOSFET has nothing to do with capacitors, MOSFET is a kind of transistor technology

3amsleep MOSFET is not a capacitor but a transistor. The MO is for metal oxide, and the FE stands for Field Effect which explains how the transistor(but other than it acts just like a normal transistor). The circuit basically just ties the leads of the capacitor to the transistor source and drain. and until the collector/gate is given a voltage the cap will almost maintain it's charge. I say almost because every transistor will leak a bit of voltage, and so the voltage has to be reapplied from time to time.This is what tabularasa0606 means by refreshing, and this why it's called dynamic ram. However it basically works the same way,the underlying details are different, with the source being the set and the gate being reset. The reason it looks different is because the q0(q not) is actually an output used in the circuit, the bit is q, q0 is just used to help with calculation. This is really how memory works in abstract sense, while dynamic and static ram are really implementations of this abstraction, if the make any sense.

that would be way back in the tube days when a transistor was the size of a honeydew mellon ....

Are there any resources that one can read up more on this?

JinGwee Unfortunately there aren't really any good consolidated sources online that I know of. Plus there's the fact that memory optimization is so fundamental that it runs up and down the entire abstraction stack of computers and causes massive headaches for anyone trying to write parallel programs. I'll give you some things you search for if you're interested. Logic-Level: SR Latch (what was shown in the video) D Flip Flop (more useful in digital logic but shows how clocks are used) Layout/Transistor-level: (mostly ripped from CMOS VLSI Design; A Circuits and Systems Perspective by Neil Weste and David Harris) Elmore Delay Estimation Domino Logic (touched on in a previous video) and precharge SRAM/6T (6 transistor) SRAM Cell DRAM (as well as the capacitors in it and refresh) Differential Sense Amplifier/Small-Signal Sensing Address line predecoding Hierarchical wordlines Domino Address decoder Hierarchical bitlines Column multiplexing Double Data Rate bus (DDR) Memory Subsystem Level: Direct-map Cache Fully associative/2-way set associative Cache Cache replacement policy Translation Lookaside Buffer (you probably need to understand virtual/physical memory and process address space for this) Cache/TLB miss/fault Memory/Cache hierarchy Burst Transfer/Burst Mode Application: Spatial-Temporal data locality Polytope Model and optimization Pointer Aliasing Operating System/Kernel: Page table and page faults Page replacement policy Cache coloring Parallelism: Race conditions Locking (or if you're adventurous, transactional memory) atomicity and mutual exclusion Non-Uniform Memory Access Architectures Cache coherency protocols Relaxed Memory model False cache-line sharing (usually done at the application level) I apologize if the list is too short.

Ferroneoboron san That's quite a formidable list. Thanks for the reply. I really appreciate it!

TTL data book ... is similar but has more info on every 74 series and by extension all 54 series chipsets ... btw when you can use the 54 series as they are the better chips (mil standard spec) the 74xx series chips are tighter parameters but the 54 set have a large operating environment and so last better

Came to the comments looking for this! Thanks!

Thank You!!

I'm sure that if you tried drawing the diagram and working it out yourself, you'd understand it just fine. It can be hard to follow an explanation just by listening.

You're right. It does loop forever or "flip-fop". When two gates output into each other's inputs, they are called flip-flops which can be used as a type of memory. Because they keep flip-flopping, they can keep their current state until they receive more input and that's why they're useful for computer memory.

Darek Amadeus I don't know, I would guess these videos are in fact made for people who don't study it as a profession, because students and professionals sure already know all this stuff. Thanks for the encouraging though, and I don't actually think I'm dumb, but it certainly feels so in such circumstances XD

It's not explained very easily, it assumes that the viewer has some knowledge of computer science. If you'd like to learn more I'd recommend a book called 'Nand to Tetris'.

palmomki The difference is, these videos don't give you the full backstory, I mean they make assumptions that you have a very basic level of understanding of both how a computer works, what RAM is used for and how electronic circuits work. If you have never studied circuits, or atleast not in enough detail to have heard of the concept of logic gates, or even resistors for that matter, then it can make it challenging. For an example, its like expecting someone to understand how to use fractions, before being taught what a fraction was. It doesn't make you dumb for not knowing something, it just means that the entire concept is something new. The same is very applicable to other things such as chemistry, engineering etc. The smartest computer scientist in the world might have never studied chemistry, and therefore if he watched a basic video talking about reactions, and didn't understand it, simply because he didn't know the very basics of chemical interactions, it wouldn't make him Dumb, just not experienced in a specific field.

Ben Eater

Bio-engineering? My sister is bio-eng. I suppose biology is alot more like studying medicine. Alot of reading and remembering. Both require understandment, but I believe computer engineering and computer science is learned in a bit different way. You don't learn programming by reading for example. You have to physically do it until you do it automatically.

I think Ben Eater made a video of that when building a "GPU". But the video is too long. I watched it, but it's too technical (but well explained)

+Jeff Cram Yikes!! I worked as an electrician and automation programmer. Did you get burned very bad? I once shorted the 9200 volt side of a small 9200/480 transformer and it was like a stick of dynamite going off, but the current was low and I didn't get burned, just scared.

sounds fun

Bones Jones No the metals were the path of the short, I just stood there stunned, not moving a muscle from the explosion and flash of light. People from across the street came over and asked if I was alright. The worst part was when I had to call my boss and tell him we had to replace a service panel, it would cost about $1,000.

It helps that with high voltage the current is not so big, especially if it it is AC

@@portaadonai did you get fired lol

+Arnolds Kļavenieks ahhahaha

+rommelfcc Not really.

+Terry Wilson (featherwinglove) I've always wondered how Minecraft people started calling it an RS NOR latch when everyone else seems to call it an SR NOR latch. Also iirc it's more common to use an SR NAND latch in memory irl.

+notanimposter I think RS was an accident derived from the initials of "RedStone"

+rommelfcc I came here to understand complex redstone circuits better XD

Correct. They are 'and' and 'or' gates with a 'not' gate after them.

He opened up the 30-yr old Mac

Q touched my pecker.

Q has upgraded to LGBTQ. Very modern. All the cool kids are doing it.

Maybe look up Charles Babbage's Difference Engine.

They should make a video of a simple CPU with programmable "ROM", a small instruction set and registers and half adder etc. It is really not a big project. Maybe this video is a first in a series of logic circuits towards such a project?

thanks!

+TheiLame I'm sure someone answered the question already but the comment seems to be gone, so I'll quickly answer again for anyone else wondering. The chip has 8 pins. Each NOR gate has 2 inputs and 1 output. There are two gates, totalling 6 pins. That leaves 2 pins. In the case of these chips, those two pins are connected to +5V and ground, high and low voltage, respectively. This is where the power comes from when no input voltage is applied to the gate. If you inspect the transistor diagram for a two input NOR gate, you will find that each gate actually has 4 inputs: A, B, high, and low. The gate-level depiction is just an abstraction to make diagrams easier to read.

+Andrew Robertson more generally the high input is called Vcc and the low is called Vss. And meaning: (Vcc) Collector Common Voltage and Vss Substrate Voltage

Benjamin Buzbee Notice how the chip has 14 pins, and only 12 of them are for logic circuits input-output. In schematic with the gates, pins no7 and no14 are empty, that's because its the logic gate schematic, not the circuit schematic. In the circuit, pin no7 is the chips ground, and pin no14 is the chips power line (in this case +5v). You can see one very small red wire connecting the 5v rail with the row no1 in which pin no14 is plugged into (rows are inter connected in each 6 holes chunk, columns are not connected), and a big brown wire connecting the ground to the row 7 (where the pin 7 is). Basically, the whole chip is constantly under voltage, and if there was no purple wires, since the gates are NOR and all inputs would be 0, all 4 output pins would also be on +5V (aka 1) constantly (in other words, if you were to remove both purple wires, both LED's would be ON).

One thing on the logic gates are the inputs and outputs but another thing is the power supply of the chip, so in this case you have on a single logic gate two inputs, one output and two extra ends, one for 5v and one for 0v

+Abdulrahman Mahdaly SRAM tends to be used in CPU caches, or any application where quicker access is required. DRAM (CMOS RAM) is slower and requires a refresh cycle to stop the charges decaying.

Salafrance Thanks :)

+Abdulrahman Mahdaly The circuit he designed is basically called an RS Latch. It has two inputs (R and S) and an Output (Q). When S is high and R is low, Q is set to high. When R is high and S is low, Q is set to low. When BOTH R and S are set to low, Q stays the value it was last time. The RS latch is the most basic component of memory in a circuit. It can further be extended to build a D flip-flop which is vital for memory in microprocessors. An Array of D flip-flops is called a Register which can hold a small bit of memory. That memory is used for many things including pipelining and register blocks.

DRAM is not CMOS RAM, it is capacitor and MOSFET based, and analog circuit. It would be more accurate to call SRAM CMOS RAM.