UINT8 - 8-Bit Unsigned Integer | Integer Format Guide

Bit Layout

Value

V

8 bits

An 8-bit unsigned integer uses all 8 bits for magnitude, giving a range of 0 to 255 (2⁸ - 1). It is the unsigned char type in C/C++, byte in C#/Java, and u8 in Rust.

Overview

UINT8 is the fundamental unit of byte-addressable memory. Its 256 possible values are sufficient for color channels, character encodings, and raw binary protocols. In machine learning, UINT8 is the standard input format for image data before normalization to float.

Common Uses

Image pixel data: UINT8 is the standard for 8-bit-per-channel color (0–255 per channel in 24-bit RGB and 32-bit RGBA). Most image formats (JPEG, PNG) store pixel data as UINT8.
Character encodings: ASCII characters, UTF-8 code units, and raw binary protocols are all byte-oriented UINT8 data.
ML image preprocessing: Raw images are loaded as UINT8 tensors, then normalized to FP32 or FP16 (e.g., dividing by 255.0) before model inference.
Embedded systems: 8-bit microcontrollers (AVR/Arduino, PIC) use UINT8 as their native register width.

See also: INT8 For the signed variant using two's complement (range -128 to 127), which is the primary format for ML inference quantization.

Range & Properties

Property	Value
Total Bits	8
Minimum Value	0
Maximum Value	255
Representable Values	256

Key Bit Patterns

Value	Bit Pattern	Signed	Unsigned
Zero	00000000	0	0
One	00000001	1	1
Max Unsigned	11111111	-1	255
Max Signed	01111111	127	127
All Ones (-1)	11111111	-1	255
Min Signed (-128)	10000000	-128	128

Interactive Bit Visualizer

Click any of the 8 bits to flip them. With only 256 possible values, you can explore the entire format.

Decimal:

Hex:

Format Comparison

Property	UINT8	UINT32	UINT16	UINT4
Total Bits	8	32	16	4
Signed Min	-128	-2,147,483,648	-32,768	-8
Signed Max	127	2,147,483,647	32,767	7
Unsigned Max	255	4,294,967,295	65,535	15
Step (ULP)	1	1	1	1
Values	256	4,294,967,296	65,536	16

Where UINT8 Is Used

Image and video pipelines: UINT8 is the universal pixel format. Virtually all image decoders (JPEG, PNG, WebP) output UINT8, and GPU texture sampling reads UINT8 channels natively.
ML frameworks: PyTorch provides torch.uint8 for image tensor storage. The ONNX specification defines UINT8 = 2 as a core data type.
GPU hardware: The PTX ISA documents .u8x4 packed types for unsigned byte operations. cuDNN supports UINT8 in INT8x4 packed tensor formats for inference.
Embedded and byte-level data: 8-bit microcontrollers (AVR/Arduino, PIC) use UINT8 as their native data width. ASCII, UTF-8, and raw binary protocols are all UINT8-oriented.