Two machines can run the same operating system, accept the same USB stick and still disagree completely about how to start. The difference lives below the operating system, in the firmware — and it is the single most common reason a carefully prepared USB drive is met with a blinking cursor instead of an installer.
The firmware is the first thing that runs
Before any operating system exists, a small program stored on the motherboard wakes the hardware up, decides which device to start from, and hands control over. For decades that program was the BIOS. Since roughly 2012 it has been UEFI on virtually every consumer machine, though many boards still offer a compatibility mode.
The change is not cosmetic. The two designs look for completely different things on a drive.
How legacy BIOS boot works
A BIOS machine reads the first 512-byte sector of the drive — the Master Boot Record. That sector holds both a partition table and a scrap of executable code, which is far too small to load an operating system by itself. Its only job is to locate and start the next stage, which then starts the one after that.
This chain works, and it worked for a very long time, but it carries the constraints of its era:
- The MBR partition table addresses at most 2 TB and allows four primary partitions.
- The boot code runs in 16-bit real mode, with almost no services to build on.
- Nothing verifies what is being executed — whatever sits in that sector runs.
How UEFI boot works
UEFI firmware can read a FAT filesystem on its own. Drives set up for it carry a dedicated EFI System Partition, and inside it a boot loader stored as an ordinary file — on removable media the fallback path is \EFI\BOOT\BOOTX64.EFI for 64-bit systems. The firmware opens that file and runs it.
Because the firmware understands files rather than sectors, several things become possible: partition tables use GPT and handle drives far beyond 2 TB, boot entries can be listed and reordered by name, and — where Secure Boot is enabled — the loader's cryptographic signature is checked before it is allowed to run.
| Legacy BIOS | UEFI | |
|---|---|---|
| Partition scheme | MBR | GPT (usually) |
| What is read first | Boot sector, 512 bytes | A file on the EFI partition |
| Filesystem for boot | None required | FAT32 |
| Drive size ceiling | About 2 TB | Effectively no practical limit |
| Signature checking | None | Optional, via Secure Boot |
Why this matters when preparing a USB drive
A stick prepared for one mode will not necessarily start on the other. This is where most confusion begins: the files copied across are fine, the image is not corrupt, and the machine still refuses to boot — because it is looking for something the drive does not have.
Partition scheme has to match the target
If the machine boots in UEFI mode, it needs a partition it can read as FAT and a loader file in the expected place. If it boots in legacy mode, it needs a valid boot sector. Utilities that write boot media ask for the target scheme up front for exactly this reason, and generally default to whatever the supplied image suggests.
FAT32 has a 4 GB file limit
Some installation images contain a single file larger than 4 GB, which FAT32 cannot store. Tooling works around this by splitting the file or by using an additional partition, but it explains why a drive sometimes ends up with a layout that looks more complicated than expected.
Secure Boot rejects unsigned loaders
With Secure Boot active, a loader without a recognised signature will simply not run. Major distributions ship signed loaders and start normally; smaller or self-built images may need Secure Boot switched off in firmware settings first.
Finding out which mode a machine uses
On Windows, System Information reports BIOS mode as either Legacy or UEFI. On most Linux systems, the presence of the /sys/firmware/efi directory means the system was started in UEFI mode. Firmware settings themselves usually expose the choice as a Boot Mode option, sometimes alongside a Compatibility Support Module — CSM — which emulates the older behaviour.
Note that the mode is decided at boot time, not by the hardware alone: a UEFI-capable machine started through CSM behaves like a legacy one, and an operating system installed in one mode generally cannot be started in the other without being repaired.
Frequently asked questions
Is UEFI simply a newer BIOS?
It replaces the same layer, but it is a different design rather than an updated version. UEFI is a specification for a firmware interface, with its own boot process, partition expectations and services.
Can one USB drive boot both ways?
Yes. Media can be laid out with both a valid boot sector and an EFI partition, so either firmware finds what it needs. Many install images are built this way.
Does GPT require UEFI?
The two are usually paired, and Windows requires GPT to boot in UEFI mode, but the partition scheme and the firmware mode are separate concepts. Some legacy setups can read GPT for data drives.
Should Secure Boot be turned off?
Generally no. It is a protection against unsigned code running before the operating system. Disabling it is a targeted step for a specific unsigned image, best reversed afterwards.