# PC Build Guide–The things they don’t tell you about RAM

There are lots of PC build guides out on the web. There are lots of things they do not tell you, too.

So what are these things?

Let’s break it down to the main one.

## How to choose fast RAM.

Your CPU is the foundation of everything else. If it is slow the rest is slow. Get the fastest CPU you can afford. Less cores is okay unless you are using well optimised programs with extensible mult-threading.
What does that mean?
Can you software use all available threads? If yes, then lots of cores and threads will be fine. If no, then it won’t help much having a 16 core 32 thread CPU.

## The first bottle-neck!

Between your CPU and RAM is the first bottleneck. This is overcome with "True Latency".

True latency is measure in nanoseconds for RAM. We want to get this under 10ns.

"But nanoseconds don’t matter. You won’t notice them!"

"We will when the difference is multiplied by a billion – which is the number of CPU cycles per second!"

The rule-of-thumb with RAM Is getting the best true latency for your dollar. Ideally this is under 10, or under 9.

How to figure out true latency.

The back-of-napkin calculation will be given but first we have to define some terms.

RAM speed is often referred to in MHz.

RAM CAS latency, or CL, is given in number of cycles. This is usually a value between 14 and 20.

So if we pick some ram it might have these attributes.

RAM_maker_X

3200Mhz
CAS (or CL) 16.

…and a bunch of other things we will ignore.

## The steps

3200Mhz is the speed.

16 is the latency.

To figure true latency:

1) Divide speed by 2.

1600

2) Divide that by 100.

16

3) Divide CL by the figure at number 2.

4) If it’s equal to 1 then you get 10 nanoseconds.

Otherwise use this formula.

`frequency/200 = a`
`cas = c`
`1/(a/c) = true latency`

So here’s some examples:

3200 C14.

16/14 = ~1.14

1 / ~1/14

8.75 nanoseconds.

3600 C16

18/16 = 1.125

1/1.125

8.8888* nanoseconds

4200 C20

21/20 =  1.05

9.5 nanoseconds