Product information
The NVIDIA RTX PRO 5000 Blackwell (900‑5G153‑2250‑000) is a high-performance professional workstation graphics card based on the Blackwell architecture for AI development, LLM inference, generative AI workflows, high-resolution video production, complex 3D modeling, and scientific computing. It features 48 GB of GDDR7 memory with ECC, a memory bandwidth of 1,344 GB/s, and supports PCIe 5.0 x16 for maximum CPU-GPU data transfer.
Based on the NVIDIA Blackwell architecture with 14,080 CUDA cores, 440 5th-generation Tensor cores, and 110 4th-generation Ray Tracing cores, the card enables ultra-fast AI calculations (up to 2,064 AI TOPS), real-time ray tracing, and DLSS 4 Multi-Frame Generation. With 3× 9th-gen NVENC and 3× 6th-gen NVDEC, it is optimized for professional video encoding/decoding workflows and offers 4× DisplayPort 2.1 for up to 8K resolution at 100 Hz.
Product Series: NVIDIA RTX PRO 5000 Blackwell Series
Model: RTX PRO 5000 Blackwell 48GB
Order Number: 900-5G153-2250-000
GTIN/EAN: 8592978700263
Product Type: Professional Workstation GPU, PCIe 5.0 x16, Dual-Slot, Active Cooling
GPU:
- Architecture: NVIDIA Blackwell (GB202)
- CUDA Cores: 14,080
- Tensor Cores: 440 (5th Generation)
- Ray Tracing Cores: 110 (4th Generation)
- AI TOPS: 2,064 (FP4, 5th-Gen Tensor Cores)
- Base Clock: 1,740 MHz
- Boost Clock: 2,377 MHz
- Manufacturing Process: TSMC 4NP (customized for Blackwell)
Memory:
- Memory Capacity: 48 GB GDDR7 with ECC
- Memory Interface: 384-bit
- Memory Bandwidth: 1,344 GB/s
- Memory Clock: 1,750 MHz (28 Gbps effective)
- ECC Support: Yes (for maximum data integrity)
Display Outputs:
- 4 × DisplayPort 2.1 (UHBR20)
- Maximum Resolution: 7680 × 4320 (8K) at 100 Hz
- Multi-Monitor Support: Yes (up to 4 displays)
- HDR Support: Yes (DisplayPort 2.1)
Performance:
- FP32 Performance: up to 67.5 TFLOPS (theoretical)
- FP16 Performance: up to 135 TFLOPS (with Tensor Cores)
- INT8 Performance: up to 270 TOPS (with Tensor Cores)
- FP4 Performance: up to 2,064 AI TOPS (for LLM inference)
- Ray Tracing Performance: up to 165 GRays/s
- NVENC: 3 × 9th-Gen (AV1, H.265, H.264)
- NVDEC: 3 × 6th-Gen (AV1, H.265, H.264)
Interfaces:
- Host Interface: PCIe 5.0 x16 (32 GT/s, backward compatible with PCIe 4.0/3.0)
- SLI/NVLink: No (not supported)
- Multi-Instance GPU (MIG): Yes (up to 2 isolated instances)
Cooling & Power Supply:
- Cooling: Active (Dual-Slot, Axial Fan)
- TDP (Total Board Power): 300 W
- Power Connector: 1 × 12VHPWR (PCIe 5.0, 16-pin)
- Recommended PSU: 750 W (depending on the system)
Physical Specifications:
- Form Factor: Dual-Slot (4.4" H × 10.5" L / 111.76 mm × 266.7 mm)
- Weight: approx. 1.2 kg (depending on configuration)
- Bracket: Standard (Low-Profile optional, available separately)
- Hot-Swap: No (not supported)
Software & Drivers:
- Operating Systems: Windows 10/11, Linux (Ubuntu, RHEL, etc.)
- Drivers: NVIDIA RTX Enterprise / Studio Drivers
- API Support: DirectX 12.2, OpenGL 4.6, Vulkan 1.4, OpenCL 3.0, CUDA 12.x
- Virtualization: Yes (vGPU, GRID, VMware, Hyper-V)
- Certifications: ISV certified for Adobe, Autodesk, Dassault, etc.
Environment:
- Operating Temperature: 0 °C to 50 °C
- Storage Temperature: -40 °C to 70 °C
- Relative Humidity: 10% to 90%, non-condensing
- Maximum Operating Altitude: up to 3,000 m (with derating)
Certifications:
- CE
- FCC
- RoHS compliant
- WEEE compliant
- ENERGY STAR (depending on configuration)
Manufacturer Warranty:
- 3-year hardware warranty (depending on sales region)
- Extended support options available via NVIDIA Services
The NVIDIA RTX PRO 5000 Blackwell (900‑5G153‑2250‑000) is a high-performance professional workstation graphics card based on the Blackwell architecture for AI development, LLM inference, generative AI workflows, high-resolution video production, complex 3D modeling, and scientific computing. It features 48 GB of GDDR7 memory with ECC, a memory bandwidth of 1,344 GB/s, and supports PCIe 5.0 x16 for maximum CPU-GPU data transfer.
Based on the NVIDIA Blackwell architecture with 14,080 CUDA cores, 440 5th-generation Tensor cores, and 110 4th-generation Ray Tracing cores, the card enables ultra-fast AI calculations (up to 2,064 AI TOPS), real-time ray tracing, and DLSS 4 Multi-Frame Generation. With 3× 9th-gen NVENC and 3× 6th-gen NVDEC, it is optimized for professional video encoding/decoding workflows and offers 4× DisplayPort 2.1 for up to 8K resolution at 100 Hz.
Product Series: NVIDIA RTX PRO 5000 Blackwell Series
Model: RTX PRO 5000 Blackwell 48GB
Order Number: 900-5G153-2250-000
GTIN/EAN: 8592978700263
Product Type: Professional Workstation GPU, PCIe 5.0 x16, Dual-Slot, Active Cooling
GPU:
- Architecture: NVIDIA Blackwell (GB202)
- CUDA Cores: 14,080
- Tensor Cores: 440 (5th Generation)
- Ray Tracing Cores: 110 (4th Generation)
- AI TOPS: 2,064 (FP4, 5th-Gen Tensor Cores)
- Base Clock: 1,740 MHz
- Boost Clock: 2,377 MHz
- Manufacturing Process: TSMC 4NP (customized for Blackwell)
Memory:
- Memory Capacity: 48 GB GDDR7 with ECC
- Memory Interface: 384-bit
- Memory Bandwidth: 1,344 GB/s
- Memory Clock: 1,750 MHz (28 Gbps effective)
- ECC Support: Yes (for maximum data integrity)
Display Outputs:
- 4 × DisplayPort 2.1 (UHBR20)
- Maximum Resolution: 7680 × 4320 (8K) at 100 Hz
- Multi-Monitor Support: Yes (up to 4 displays)
- HDR Support: Yes (DisplayPort 2.1)
Performance:
- FP32 Performance: up to 67.5 TFLOPS (theoretical)
- FP16 Performance: up to 135 TFLOPS (with Tensor Cores)
- INT8 Performance: up to 270 TOPS (with Tensor Cores)
- FP4 Performance: up to 2,064 AI TOPS (for LLM inference)
- Ray Tracing Performance: up to 165 GRays/s
- NVENC: 3 × 9th-Gen (AV1, H.265, H.264)
- NVDEC: 3 × 6th-Gen (AV1, H.265, H.264)
Interfaces:
- Host Interface: PCIe 5.0 x16 (32 GT/s, backward compatible with PCIe 4.0/3.0)
- SLI/NVLink: No (not supported)
- Multi-Instance GPU (MIG): Yes (up to 2 isolated instances)
Cooling & Power Supply:
- Cooling: Active (Dual-Slot, Axial Fan)
- TDP (Total Board Power): 300 W
- Power Connector: 1 × 12VHPWR (PCIe 5.0, 16-pin)
- Recommended PSU: 750 W (depending on the system)
Physical Specifications:
- Form Factor: Dual-Slot (4.4" H × 10.5" L / 111.76 mm × 266.7 mm)
- Weight: approx. 1.2 kg (depending on configuration)
- Bracket: Standard (Low-Profile optional, available separately)
- Hot-Swap: No (not supported)
Software & Drivers:
- Operating Systems: Windows 10/11, Linux (Ubuntu, RHEL, etc.)
- Drivers: NVIDIA RTX Enterprise / Studio Drivers
- API Support: DirectX 12.2, OpenGL 4.6, Vulkan 1.4, OpenCL 3.0, CUDA 12.x
- Virtualization: Yes (vGPU, GRID, VMware, Hyper-V)
- Certifications: ISV certified for Adobe, Autodesk, Dassault, etc.
Environment:
- Operating Temperature: 0 °C to 50 °C
- Storage Temperature: -40 °C to 70 °C
- Relative Humidity: 10% to 90%, non-condensing
- Maximum Operating Altitude: up to 3,000 m (with derating)
Certifications:
- CE
- FCC
- RoHS compliant
- WEEE compliant
- ENERGY STAR (depending on configuration)
Manufacturer Warranty:
- 3-year hardware warranty (depending on sales region)
- Extended support options available via NVIDIA Services
NVIDIA RTX PRO 5000 Blackwell 900‑5G153‑2250‑000 – 48 GB GDDR7 ECC Workstation‑GPU, PCIe 5.0 x16, 4× DisplayPort 2.1, 300 W, Dual‑Slot
- Regular price
- 9.550,99 €
- Sale price
- 9.550,99 €
- Regular price
-
Couldn't load pickup availability