Addressing Memory (Pt1) - Computerphile

Each BIT in memory doesn't have it's own unique wiring, they share connections - Dr 'Heartbleed' Bagley explains how we address them.

Memory Recall - COMING SOON
Professor Steve Furber on the ARM chip: COMING SOON, Subscribe for Updates!

Correction: In the graphic at 7min:30secs, a-0 & a-1 are incorrectly labelled a-1 & a-2

This video was filmed and edited by Sean Riley.

[xaostek]

Sometimes, whenever i'm watching these videos, i'm normally at least mildly interested in the subject matter. Then I suddenly remember that in order to watch this video, this thing that he's described in 8 minutes is happening about a foot away from me, millions of times a second, near the speed of light, and it just blows my mind. This channel just does these things to me sometimes.

[NikiHerl]

It'd be cool to have all the "previous videos on how memory works" in a play list. I think I remember watching (most? of) them, but it's easier to wrap ones head around the various topic when seeing the entire video, not just a 15 second recap =)

[kirkpatricklucas]

This guy spent 8 minutes explaining a simple demultiplexer without ever having uttered the word, "demultiplexer"

[jndominica]

The animated schematics are neatly done. :)

[ProfessorGreenington]

Well, tomorrow I'm going to write an exam in "Technischer Informatik" which is German for technical computer science. Good thing I checked my subscriptions on YouTube before starting to study because the content of this video will also be part of the exam :-)

[ericsbuds]

this is so interesting to me. I love learning this type of stuff! thanks guys

[weetabixharry]

This is okay for a 2-bit memory, but for a large (deep) memory I'm pretty sure you wouldn't want to implement independent addressing logic for every memory location. Presumably, there's a smart way to share logic to minimize the number of transistors used. This must be a solved problem... and I bet the answer is interesting, but so far my googling skills are failing me.

[STDrepository]

When he finished with the 0 and 1 flipflop I was like, "but those only correspond with high and low on the A1". So how are you going to do more than 2 flipflops!! you can't have more than high and low state on the A1! And then he was just like "so we add A2"... And suddenly it all clicked for me and now I understand memory addressing.

[MarkSinclair]

My interest has been peaked too many times without action. I want to create these circuits for myself. Before diving into breadboards and transistors is there any software out there that will let me create these kinds of circuits that is easy for  beginner? If it was online that would be a bonus!!

0:38 A common mistake, it's not "between nought and fifteen" but must be "from nought to fifteen", the first contains fourteen values (n>0 ; n<15), the latter includes sixteen values (n>=0 ; n<=15) !

[PhilHibbs]

Are we going to see a video on the "rowhammer" problem?

[fogglee]

That second And gate I think is unnecessary, the first one is ensure the data only changes when the write line is asserted so no need to do it again, because theres only two flip-flops, just invert the output of the first And gate.
Also I'm pretty sure real systems don't do this. The memory locations are literally stacked on top of each (into a physical rectangle) other so it does X/Y decoding instead.

[TheStevenWhiting]

Try out prison architect. They coded logic into that for controlling doors etc. Someone ended up making an 8bit I think it was virtual computer with it :)

[waynesmith]

Should have been an OR gate. A HIGH on the WR line holds the Clock input HIGH until the Address line is stabilized, then the single address line should be inverted to decode the address "1". 

[MagnusTheUltramarine]

So, for each clock cycle you have to rewrite all the values that are 1, right?

[FerroNeoBoron]

Wouldn't this implementation have bad clock skew for reasonalbe size memories? I'm also pretty sure you'd have issues with the clock input glitching. It seems like if you wanted to go with a purely digital example of memory that it would be less fraught to use 2-to-1 multiplexer feedback.

[power-max]

3:50: Why not connect both gates to the clock, and just invert one them? like use a common emitter NPN transistor to invert?

Unless having both clock pins at 0 is also a requirement, why bother with that 6:35 logic gate stuff?

[WinrichNaujoks]

So this is part 1 but is a follow up to an earlier video. Part 0? Where is it?

[ykl]

This is getting ridiculous. You are going thru my last year's coursework.

[danielprovder]

Brady, can you please do a video on assembler?