GPU

NVIDIA GeForce RTX 4080

Edit@1 months ago

Intergrated Memory(VRAM)
Capacity

16 GB

(GDDR6X 256-bit)

Bandwidth

716 GB/s

102 Token/s

Vector Compute
FP64
761.50 G
FP32
48.74 T
FP16
48.74 T
BF16
48.74 T
INT32
24.37 T
INT8
X

NVIDIA GeForce RTX 4080 General-Purpose Float-Point performance (Vector Performance / Scalar Performance)

FP64: 761.50 GFLOPS

FP32: 48.74 TFLOPS

FP16: 48.74 TFLOPS

BF16: 48.74 TFLOPS

INT32: 24.37 TOPS

Matirx Compute
FP64
X
FP32
X
FP16
97.47 T
194.95 T
FP8
194.95 T
389.90 T
TF32
48.74 T
97.47 T
BF16
97.47 T
194.95 T
INT16
X
INT8
389.90 T
779.80 T
INT4
779.80 T
1559.59 T

NVIDIA GeForce RTX 4080 AI performance (Tensor Performance / Matrix Performance)

FP16: 97.47 TFLOPS, with sparsity: 194.95 TFLOPS

FP8: 194.95 TFLOPS, with sparsity: 389.90 TFLOPS

TF32: 48.74 TFLOPS, with sparsity: 97.47 TFLOPS

BF16: 97.47 TFLOPS, with sparsity: 194.95 TFLOPS

INT8: 389.90 TOPS, with sparsity: 779.80 TOPS

INT4: 779.80 TOPS, with sparsity: 1559.59 TOPS

Hardware Specs
NVIDIA GeForce RTX 4080 is a 5nm chip, has 45900 million transistors, launched by NVIDIA at 2022. It has 16 GB built-in(On-Board/On-Chip) memory with bandwidth up to 716 GB/s. It has 9728 general-purpose ALUs(CUDA cores/Shader cores) and 304 matrix cores(Tensor cores) .
Process Node
5 nm
Launch Year
2022
Vector(CUDA) Cores
9728
Matrix(Tensor) Cores
304
Core Frequency
2205 ~ 2505 MHz
Cache
64MB