Section 3: POST (Power-On Self Test)

The Big Picture: The Pre-Flight Check

Before a pilot takes off, they run a checklist: engines, fuel, controls. If something is wrong, they stop. A PC does the same. POST (Power-On Self-Test) is the firmware's checklist. It tests the most important parts, sets them up, and only then looks for an operating system. If something vital is broken, POST stops and tells you, often with beeps or lights, because the screen may not work yet.

3.1 First Steps After Power

What it is: A lot happens in the first fraction of a second, before any code from the BIOS runs.

Press power button
     │
     ▼
1. PSU turns on and checks its voltages
     │  sends the "Power Good" signal (Chapter 4)
     ▼
2. Chipset holds the CPU in RESET until power is stable
     │
     ▼
3. Reset released ─► CPU starts in real mode
     │  jumps to 0xFFFFFFF0 (reset vector)
     ▼
4. Firmware code begins (first code runs straight from the ROM chip)

Why wait for "Power Good"? If the CPU started while voltages were still rising, it could run wrong and crash or be damaged. The reset line keeps it frozen until everything is steady.

The very first code is special. There is no RAM yet, so the code cannot use normal memory. It runs from the ROM chip directly and uses the CPU's own registers (or CPU cache, set up as a tiny RAM) to work. Once the real RAM is ready, the firmware moves on.

3.2 Hardware Initialization Sequence

What it is: POST follows a fixed order. Each step prepares the next. The exact order varies by maker, but the shape is always like this:

Power ► CPU ► ROM ► Chipset ► RAM ► Video ► Keyboard ► Drives/PCIe ► Boot
        (earlier steps need less; each step may need the one before it)

Why this order? You cannot show an error on screen before the graphics card works, and you cannot start the graphics card before RAM works. So the earliest errors cannot be shown as text. They are shown by beeps or LEDs (see 3.6).

Figure 5.4 — The parts POST has to bring up, on an old board: CPU socket (1), memory slots (2), chipset chips (3, 4), the graphics slot (5), PCI slots (6), the BIOS chip (8) and the battery (9). Photo: Arnulf zu Linden, CC BY-SA 3.0, numbers added, via Wikimedia Commons.

3.3 CPU Register Initialization

What it is: Registers are the CPU's tiny, super-fast storage boxes (you met them in Chapter 2). At reset, the CPU sets them to known starting values, so the firmware always begins from the same clean state.

Key facts on x86 at reset:

The other cores (called APs, Application Processors) wait. Later, the firmware (and then the OS) wakes them with special signals. That is why, in early boot, even a 16-core CPU runs just one core.

What the firmware sets up early:

• A tiny stack (a scratch area in memory for function calls). Before RAM works, it uses the cache as RAM.
• Caching rules (called MTRRs/PAT) for different memory areas.
• The descriptor tables that allow a switch to 32/64-bit mode (in UEFI).
• Microcode: CPU makers sometimes ship fixes for the CPU. The firmware can load the newest microcode update (a patch for the CPU's own internal program) at this stage.

3.4 Memory Testing

What it is: Before using RAM, the firmware must check that it works. Bad RAM causes crashes and corrupted files, so it is better to catch it here.

Two parts:

1. Memory training. Modern DDR RAM runs at billions of transfers a second. Signals are so fast that the memory controller must calibrate itself: it sends test patterns and adjusts delays until reads and writes are clean. Each stick tells the controller its speed and timings through a small chip called the SPD (Serial Presence Detect). This training is why a PC with a lot of RAM, or after a RAM change, can take a long time on first boot.
2. Memory test. The firmware writes patterns to RAM and reads them back. Old BIOSes counted up the RAM on screen ("640K... 1024K... 32768K OK"). Today a quick test is common (a thorough test of a whole 32 GB would take many minutes).

Windows tip: Windows has its own tool. Press Win, type Windows Memory Diagnostic, run it and restart. A more thorough free tool is MemTest86 (a third party, from its official site).

Symptoms of a RAM problem during POST: long or repeating beeps, a memory LED on the board, a black screen, or the PC restarting a few times before it starts ("training again").

3.5 Device Discovery

What it is: After RAM and video work, POST looks for everything else and gives each device resources.

What it looks for:

For each device, the firmware gives it memory addresses, I/O ports and an interrupt line, so the OS can find them later. This is called resource assignment. Old PCs used Plug and Play (PnP) for this. On the Award screen in Figure 5.2 you can see a menu called PnP/PCI Configurations.

What you may see on a classic screen: A summary table at the end of POST listing the CPU, RAM size, disks and PCI devices. Many modern PCs hide this behind the maker's logo (press Esc or Tab to peek).

Windows tip: To see what the firmware told Windows, run:

Get-CimInstance Win32_ComputerSystem | Select-Object Manufacturer, Model, TotalPhysicalMemory
Get-CimInstance Win32_BIOS | Select-Object SMBIOSBIOSVersion, ReleaseDate

3.6 Error Detection and Beep Codes

What it is: If POST finds a serious fault before the screen works, it cannot print text. So it uses sound (a beep from a little speaker) or lights.

Why a speaker? A buzzer needs almost nothing to work, no RAM, no graphics. It is the simplest way to say "help".

Classic beep codes. The patterns depend on the maker. Always check your motherboard manual. These are examples to show the idea:

Modern error signals:

Common POST messages in plain words:

A quick troubleshooting order for "no display" at power-on:

1. Check the power switch on the PSU and the 24-pin and CPU power cables.
2. Re-seat the RAM (push until both clips click).
3. Unplug extra cards and drives, try with the minimum parts.
4. Look at the debug LEDs and listen for beeps.
5. Clear the CMOS (Section 7).

Summary