GPU

NVIDIA CMP 40HX

Edit@11 months ago

Intergrated Memory(VRAM)
Capacity

8 GB

(GDDR6 256-bit)

Bandwidth

448 GB/s

64 Token/s

Vector Compute
FP64
237.60 G
FP32
7.60 T
FP16
15.21 T
BF16
X
INT32
7.60 T
INT8
X

NVIDIA CMP 40HX General-Purpose Float-Point performance (Vector Performance / Scalar Performance)

FP64: 237.60 GFLOPS

FP32: 7.60 TFLOPS

FP16: 15.21 TFLOPS

INT32: 7.60 TOPS

Matirx Compute
FP64
X
FP32
X
FP16
X
FP8
X
TF32
X
BF16
X
INT16
X
INT8
X
INT4
X

NVIDIA CMP 40HX AI performance (Tensor Performance / Matrix Performance)

Hardware Specs
NVIDIA CMP 40HX is a 12nm chip, has 10800 million transistors, launched by NVIDIA at 2021. It has 8 GB built-in(On-Board/On-Chip) memory with bandwidth up to 448 GB/s. It has 2304 general-purpose ALUs(CUDA cores/Shader cores) and 288 matrix cores(Tensor cores) .
Process Node
12 nm
Launch Year
2021
Vector(CUDA) Cores
2304
Matrix(Tensor) Cores
288
Core Frequency
1470 ~ 1650 MHz
Cache
4MB