GPU

NVIDIA GeForce RTX 4070 Ti SUPER

Edit@1 months ago

Intergrated Memory(VRAM)
Capacity

16 GB

(GDDR6X 256-bit)

Bandwidth

672 GB/s

96 Token/s

Vector Compute
FP64
689 G
FP32
44.10 T
FP16
44.10 T
BF16
44.10 T
INT32
22.05 T
INT8
X

NVIDIA GeForce RTX 4070 Ti SUPER General-Purpose Float-Point performance (Vector Performance / Scalar Performance)

FP64: 689 GFLOPS

FP32: 44.10 TFLOPS

FP16: 44.10 TFLOPS

BF16: 44.10 TFLOPS

INT32: 22.05 TOPS

Matirx Compute
FP64
X
FP32
X
FP16
88.20 T
176.39 T
FP8
176.39 T
352.79 T
TF32
44.10 T
88.20 T
BF16
88.20 T
176.39 T
INT16
X
INT8
352.79 T
705.58 T
INT4
705.58 T
1411.15 T

NVIDIA GeForce RTX 4070 Ti SUPER AI performance (Tensor Performance / Matrix Performance)

FP16: 88.20 TFLOPS, with sparsity: 176.39 TFLOPS

FP8: 176.39 TFLOPS, with sparsity: 352.79 TFLOPS

TF32: 44.10 TFLOPS, with sparsity: 88.20 TFLOPS

BF16: 88.20 TFLOPS, with sparsity: 176.39 TFLOPS

INT8: 352.79 TOPS, with sparsity: 705.58 TOPS

INT4: 705.58 TOPS, with sparsity: 1411.15 TOPS

Hardware Specs
NVIDIA GeForce RTX 4070 Ti SUPER is a 5nm chip, has 45900 million transistors, launched by NVIDIA at 2024. It has 16 GB built-in(On-Board/On-Chip) memory with bandwidth up to 672 GB/s. It has 8448 general-purpose ALUs(CUDA cores/Shader cores) and 264 matrix cores(Tensor cores) .
Process Node
5 nm
Launch Year
2024
Vector(CUDA) Cores
8448
Matrix(Tensor) Cores
264
Core Frequency
2340 ~ 2610 MHz
Cache
48MB