NVIDIA
GPUDatacenterReleased Mar 18, 2025

NVIDIA RTX PRO 6000 Blackwell Server Edition

96 GB GDDR7 ECC · 512-bit · 24,064 cores

The data-centre version of the RTX PRO 6000 Blackwell, bringing the same core configuration and 96 GB of GDDR7 into servers for AI inference, rendering, video and virtual workstation workloads. The air-cooled dual-slot, full-height full-length card is passively cooled by the host chassis, a single-slot liquid-cooled version is also offered, and board power is configurable up to 600 W. It has four encode and four decode engines, and MIG can partition it into up to four isolated instances.

General

Manufacturer
NVIDIA
Model
RTX PRO 6000 Blackwell Server Edition
Segment
Datacenter
GPU chip
GB202
Architecture
Blackwell
Release date
Mar 18, 2025
Price
Search on Amazon

Memory

Memory size
96 GB
Memory type
GDDR7 ECC
Bus width
512-bit
Bandwidth
1,597 GB/s
Memory clock
—

Clocks & silicon

GPU clock
—
Boost clock
—
Shader units
24,064 CUDA Cores
Tensor / AI cores
752
RT cores
188
Transistors
92.2 billion
Technology (nm)
5 nm
TSMC 4N (5 nm-class)
TDP / TBP
600 W

Interface & APIs

Bus interface
PCIe 5.0 x16
SLI / multi-GPU
None (no SLI / NVLink)
DirectX
12 Ultimate (12_2)
OpenGL
4.6
Vulkan
1.4
Other APIs
CUDA 12.0, OpenCL 3.0, DLSS 4

AI support

Precisions
FP8FP16BF16INT8INT4TF32FP64
AI accelerator
5th Gen Tensor Cores (752) + 4th Gen RT Cores (188)
Peak AI TOPS
4,000 TOPS
Notes
Fifth-generation Tensor Cores add FP4 to the FP8/BF16/FP16/INT8/INT4/TF32 set. NVIDIA publishes 4 PFLOPS of FP4 Tensor throughput (4,000 AI TOPS), 2 PFLOPS FP8, 1 PFLOP FP16/BF16 and 234 TFLOPS TF32. Compute throughput is 120 TFLOPS FP32 and 355 TFLOPS for ray tracing.

Related NVIDIA GPUs

All NVIDIA GPUs →

Datacenter gpus from NVIDIA announced around the same time as the NVIDIA RTX PRO 6000 Blackwell Server Edition. Add any of them to your compare list to see the specs side by side.

NVIDIAgpu

NVIDIA H200 NVL

141 GB HBM3e · 6144-bit · 16,896 cores

Released
Nov 18, 2024