Computer hardware

What is inside a computer

Hardware means everything in a computer you can physically touch. There are five main parts, and the first three do the actual work.

The processor, or CPU, calculates. Everything a computer does, from adding two numbers to drawing this page, ends up as billions of tiny operations, and the processor performs them one after another, very fast. A modern processor has several cores, which is to say several accountants working side by side. The server this site runs on has four.

Memory, or RAM, is the desk. Every open program, every open file and every browser tab you left alone is spread out on that desk. The processor only works with what is on the desk. A bigger desk means more things can stay open at once.

Storage is the filing cabinet. It is slower, but its contents survive when the power goes. Two kinds are common today: a mechanical hard drive with a spinning metal platter, and an SSD with no moving parts at all, which is why it answers far faster.

Two more parts get no thought until they fail. The motherboard is the big board everything plugs into; the part that matters to you is that it decides which components physically fit in this computer at all. The power supply and the cooling deliver electricity and carry heat away, and a machine that shuts down by itself or runs its fan at a maddening pitch usually has a problem right there, not in the processor.

A good metaphor breaks somewhere, and it is better to say where yourself. Put the desk and the cabinet side by side and the difference looks like nothing more than distance. The real difference is much larger than that, and the third part of this lesson gives the number.

How is RAM different from storage

This is the most common misunderstanding about computers, and it is worth settling once. Memory and storage are both places to put things, but they do two entirely different jobs.

Memory lives on electricity. The moment the computer switches off, whatever was in memory is gone, with no appeal. That is why a power cut destroys the file you had not saved: that file was still on the desk and had never made it to the cabinet. Storage is the opposite; it needs no power and holds its contents for years.

So "my computer has a terabyte, why is it slow" is not the right question. A terabyte is the size of the cabinet and says nothing about the size of the desk. The right question is how much of the desk is still free once everything you use is open.

And here is the part that gets said less often, and it is exactly why more RAM speeds a machine up even when you open nothing new: an operating system refuses to leave free memory idle. Every file read from disk leaves a copy in memory so the next read does not have to go to the disk at all. The Linux kernel documentation puts it in one line: whenever a file is read, the data is put into the page cache to avoid expensive disk access on subsequent reads, and those copies are handed back the moment memory runs short. Memory that looks free is in fact working.

The practical consequence: when memory fills up, those copies are the first thing the operating system sacrifices. From that moment on, work that was being answered out of memory yesterday has to go back to the disk. A computer does not get slow all at once; it gets slow past a point.

What actually makes a computer feel slow

Our position is plain: for everyday work, the processor is almost never the culprit. Too little memory and a slow disk are what make a computer feel slow, and the reason is a number we can put here.

We measured it on the very server that serves rgb.ir. We made a 128 MB file and read 2000 random 4 KB chunks out of it. The first time we made sure those reads genuinely reached the SSD: each one took about 0.22 ms. Then we repeated the same 2000 reads with the data now sitting in memory: about 0.0013 ms each. That is roughly 160 times faster, and that is against a modern SSD.

Now the sentence from the previous part means something. While memory has room, a large share of what a computer does happens at the small number. When memory fills, the same work happens at the big one. The gap is wide enough that the fastest processor on earth does not make it up, because the processor spends the whole time sitting still, waiting.

On a mechanical hard drive the gap is far wider, because there a physical arm has to move. We do not quote a figure for it, because this machine has no mechanical drive and we did not measure one; we only state the direction. That single fact is what makes swapping an old laptop hard drive for an SSD the most noticeable upgrade available.

Two more causes have nothing to do with parts and cost nothing. First, the programs that start with the machine and hold memory all day. Second, a disk that is nearly full and leaves the operating system no room for its temporary files; we watch the same thing on our own servers, and this machine sits at 85 percent, which is the last point we are comfortable with. When a site we host gets sluggish, those two are the first things we look at, not processor power.

And the boundary of the claim: if the slow thing is exporting video, rendering in 3D or compiling code, then the processor really is the bottleneck and more memory changes nothing. That is not an exception, it is a separate rule.

Who actually needs a graphics card

A graphics card is a processor too, of a different kind. The main processor does complicated things one after another; a graphics card does one simple thing to millions of points at the same time. For colouring in the pixels of a 3D scene that is exactly what is needed.

Our position, and we know a salesperson would not enjoy it: for most people a separate graphics card is the most common way to waste money. Web browsing, office work, video playback and even a good deal of web design work run well enough on the graphics built into the processor. A graphics card makes none of them faster.

Four groups genuinely need one: gaming, video editing and colour work, 3D modelling and rendering, and, for a few years now, running AI models on your own computer. That fourth group has an important difference: the deciding number there is not the card speed but the size of its dedicated memory, because a model either fits in it or does not. That subject has a lesson of its own.

One small point that is often stated wrongly: most laptops and every phone have their graphics built into the processor and take a share of the main memory for it. On such a device, the memory that goes to graphics is taken off your desk.

What to check before buying or upgrading

Start by finding out what you already have. On Windows, Task Manager, Performance tab; on macOS, Activity Monitor; on Linux, the two commands free -h and lscpu. You need three numbers: total memory, how much is in use at an ordinary working moment, and the type and free space of the disk.

Then check compatibility, and here is a fact people learn too late: on many of the thin laptops of the last few years the memory is soldered to the board and cannot be added to at all. Which means "I will add RAM later" is not a plan for that machine. Storage, on many of the same machines, is still replaceable. The answer has to come from the manufacturer own page for your exact model number, not for the model family.

And the last thing, which usually costs nothing: before you pay for a part, empty the list of programs that start automatically and get the disk under eighty percent. If it is still slow after those two, then go shopping. That is the order.

The natural next step is to see what the operating system does with these same parts, and why two computers with identical hardware can behave completely differently.