GPU

NVIDIA GeForce RTX 3060

Edit@1 months ago

Intergrated Memory(VRAM)
Capacity

12 GB

(GDDR6 192-bit)

Bandwidth

360 GB/s

51 Token/s

Vector Compute
FP64
199 G
FP32
12.74 T
FP16
12.74 T
BF16
12.74 T
INT32
6.37 T
INT8
X

NVIDIA GeForce RTX 3060 General-Purpose Float-Point performance (Vector Performance / Scalar Performance)

FP64: 199 GFLOPS

FP32: 12.74 TFLOPS

FP16: 12.74 TFLOPS

BF16: 12.74 TFLOPS

INT32: 6.37 TOPS

Matirx Compute
FP64
X
FP32
X
FP16
25.48 T
50.95 T
FP8
X
TF32
12.74 T
25.48 T
BF16
25.48 T
50.95 T
INT16
X
INT8
101.90 T
203.80 T
INT4
203.80 T
407.60 T

NVIDIA GeForce RTX 3060 AI performance (Tensor Performance / Matrix Performance)

FP16: 25.48 TFLOPS, with sparsity: 50.95 TFLOPS

TF32: 12.74 TFLOPS, with sparsity: 25.48 TFLOPS

BF16: 25.48 TFLOPS, with sparsity: 50.95 TFLOPS

INT8: 101.90 TOPS, with sparsity: 203.80 TOPS

INT4: 203.80 TOPS, with sparsity: 407.60 TOPS

Hardware Specs
NVIDIA GeForce RTX 3060 is a 8nm chip, has 12000 million transistors, launched by NVIDIA at 2021. It has 12 GB built-in(On-Board/On-Chip) memory with bandwidth up to 360 GB/s. It has 3584 general-purpose ALUs(CUDA cores/Shader cores) and 112 matrix cores(Tensor cores) .
Process Node
8 nm
Launch Year
2021
Vector(CUDA) Cores
3584
Matrix(Tensor) Cores
112
Core Frequency
1320 ~ 1777 MHz
Cache
3MB