Blackwell Architecture · 4,352 CUDA Cores · 136 Tensor Cores (5th Gen) · 34 RT Cores (4th Gen) · PCIe 5.0 x16
RTX PRO 2000 Blackwell
NVIDIA Blackwell (GB206)
4,352
136 (5th Gen)
Up to 3× prev gen perf
34 (4th Gen)
RTX Mega Geometry
545 AI TOPS
(FP4 with sparsity)
17 TFLOPS
52 TFLOPS
16 GB GDDR7 ECC
128-bit
288 GB/s
18 Gbps effective
PCIe 5.0 x16
Full-length connector
HHHL Dual-Slot
2.7"H × 6.6"L
~0.29 kg
70W
Active blower fan
4× Mini DisplayPort 2.1b
Up to 8K @ 165 Hz
1× 9th Gen NVENC
1× 6th Gen NVDEC
4:2:2 H.264/HEVC
The NVIDIA RTX PRO 2000 Blackwell brings the Blackwell architecture to entry-level and mid-range professional workstations. With 4,352 CUDA cores, 136 fifth-generation Tensor Cores delivering 545 AI TOPS, and 16 GB of GDDR7 ECC memory, this 70W half-height, half-length (HHHL) card punches well above its power class. It is the go-to choice for AI-assisted CAD, real-time ray-traced rendering, multi-application content creation, and edge AI inference — all from compact desktop systems.
Built on the GB206 GPU, the RTX PRO 2000 provides PCIe 5.0 x16 connectivity for double the bandwidth of the previous generation, plus 9th-gen NVENC and 6th-gen NVDEC engines with 4:2:2 H.264/HEVC hardware encode/decode. Four Mini DisplayPort 2.1b outputs drive up to 8K @ 165 Hz displays. The 16 GB GDDR7 frame buffer with ECC ensures data integrity for professional applications, while NVIDIA enterprise drivers and ISV certifications guarantee reliability across CAD, DCC, and engineering software stacks.
QS Compute supplies NVIDIA RTX PRO 2000 Blackwell GPUs from partners including Leadtek and PNY across the APAC region — ideal for AI developers prototyping models locally, designers running real-time ray tracing, and enterprises deploying compact multi-GPU inference nodes.