GPU

NVIDIA RTX PRO 6000 Blackwell

Edit@1 months ago

Intergrated Memory(VRAM)
Capacity

96 GB

(GDDR7 512-bit)

Bandwidth

1792 GB/s

256 Token/s

Vector Compute
FP64
1.97 T
FP32
126 T
FP16
126 T
BF16
X
INT32
X
INT8
X

NVIDIA RTX PRO 6000 Blackwell General-Purpose Float-Point performance (Vector Performance / Scalar Performance)

FP64: 1.97 TFLOPS

FP32: 126 TFLOPS

FP16: 126 TFLOPS

Matirx Compute
FP64
X
FP32
X
FP16
251.90 T
503.80 T
FP8
503.80 T
1007.61 T
TF32
125.95 T
251.90 T
BF16
251.90 T
503.80 T
INT16
X
INT8
1007.61 T
2015.22 T
INT4
X
Theoretical peak performance: RTX-Pro-Blackwell

NVIDIA RTX PRO 6000 Blackwell AI performance (Tensor Performance / Matrix Performance)

FP16: 251.90 TFLOPS, with sparsity: 503.80 TFLOPS

FP8: 503.80 TFLOPS, with sparsity: 1007.61 TFLOPS

TF32: 125.95 TFLOPS, with sparsity: 251.90 TFLOPS

BF16: 251.90 TFLOPS, with sparsity: 503.80 TFLOPS

INT8: 1007.61 TOPS, with sparsity: 2015.22 TOPS

Hardware Specs
NVIDIA RTX PRO 6000 Blackwell is a 5nm chip, has 92200 million transistors, launched by NVIDIA at 2025. It has 96 GB built-in(On-Board/On-Chip) memory with bandwidth up to 1792 GB/s. It has 24064 general-purpose ALUs(CUDA cores/Shader cores) and 752 matrix cores(Tensor cores) .
Process Node
5 nm
Launch Year
2025
Vector(CUDA) Cores
24064
Matrix(Tensor) Cores
752
Core Frequency
1590 ~ 2617 MHz
Cache
128MB