BFKK
BFKK 3U Rackmount Server Chassis EATX/ATX/MATX 4x5.25 6x3.5 Top/Side Cooling
- Rackmount
US$214.99
0.00265886BTC
現貨
NVIDIA
NVIDIA RTX PRO 6000 Blackwell 96 GB GDDR7 900-5G144-2200-000-VGA workstation graphics card
Workstation GPU with 96 GB GDDR7 memory and 24,064 CUDA cores for AI and graphics workloads
- InterfacePCI Express 5.0 x16
- Chipset ManufacturerNVIDIA
- GPURTX PRO 6000
- CUDA Cores24064
- Memory Size96GB
- Memory Interface512-bit
- 96 GB GDDR7 memory with ECC delivers 1.8 TB/s bandwidth for massive datasets
- 24,064 CUDA cores provide 125 TFLOPS single-precision compute for complex simulations
- Fifth-gen Tensor cores enable 4,000 AI TOPS with FP4 precision for accelerated inference
- Four DisplayPort 2.1b outputs drive multiple high-resolution displays simultaneously
- PCIe 5.0 x16 interface and 600 W max power support full-bandwidth data transfer
Professional AI workstation GPU
The NVIDIA RTX PRO 6000 Blackwell is a workstation graphics card built for AI researchers, data scientists and visual effects studios. It delivers 24,064 CUDA cores and fifth-generation Tensor cores with FP4 precision for large-model training and inference. Four DisplayPort 2.1b outputs drive multi-monitor configurations at high resolutions.
96 GB GDDR7 memory with ECC
The 96 GB GDDR7 frame buffer runs on a 512-bit interface at 1.8 TB/s bandwidth, providing headroom for massive datasets, complex scenes and multi-instance GPU partitions up to 4x 24 GB or 2x 48 GB. Workloads exceeding 96 GB must use system memory paging, which reduces throughput for the largest language models or simulation volumes.
Dual-slot PCIe 5.0 x16 card with 600 W power
The 305 mm dual-slot cooler uses a double-flow-through design and requires a single PCIe CEM5 16-pin connector rated for 600 W maximum consumption. A motherboard with a PCI Express 5.0 x16 slot and a power supply providing the 16-pin connector are mandatory; the card frees expansion space by occupying only two slots despite its length.
商品特色
- 96 GB GDDR7 memory with ECC delivers 1.8 TB/s bandwidth for massive datasets
- 24,064 CUDA cores provide 125 TFLOPS single-precision compute for complex simulations
- Fifth-gen Tensor cores enable 4,000 AI TOPS with FP4 precision for accelerated inference
- Four DisplayPort 2.1b outputs drive multiple high-resolution displays simultaneously
- PCIe 5.0 x16 interface and 600 W max power support full-bandwidth data transfer
商品規格
| Brand | NVIDIA |
|---|---|
| Series | Blackwell |
| Part Number | 900-5G144-2200-000 |
| Interface | PCI Express 5.0 x16 |
| Chipset Manufacturer | NVIDIA |
| GPU | RTX PRO 6000 |
| CUDA Cores | 24064 |
| Memory Size | 96GB |
| Memory Interface | 512-bit |
| Memory Type | GDDR7 with ECC |
| DirectX | DirectX 12 |
| OpenGL | OpenGL 4.6 |
| DisplayPort | 4x DisplayPort 2.1b |
| AI Features | AI TOPS: 4000 Tensor Cores: 5th Gen Ray Tracing Cores: 4th Gen NVIDIA Encoder (NVENC): 4x 9th Gen NVIDIA Decoder (NVDEC): 4x 6th Gen CUDA 12.8 OpenCL 3.0 DirectCompute |
| Cooler | Double-flow-through |
| Auxiliary Power Connector | 1x PCIe CEM5 16-pin |
| Power Consumption | Max Power Consumption: 600W |
| Features | MIG Support: >Up to 4x 24 GB >Up to 2x 48 GB >Up to 1x 96 GB |
| Max GPU Length | 305 mm |
| Slot Width | Dual Slots |
| Dimensions (L x H) | 5.4” H x 12” L |
| 出貨重量 | 2.36 kg |
| 包裝尺寸 | 387 × 191 × 89 mm |
商品問答
Will the PCIe 5.0 x16 interface work with my current PCIe 4.0 motherboard?
Yes, the card is backward compatible with PCIe 4.0 slots, though bandwidth will be limited to the motherboard's generation.
How can I verify the 24064 CUDA cores and 96 GB GDDR7 memory are recognised after installation?
Open NVIDIA Control Panel or GPU-Z to confirm the CUDA core count and 96 GB GDDR7 with ECC are reported correctly.
Does the box include a power adapter for the 16-pin CEM5 connector or must I buy one separately?
The card requires a 16-pin PCIe CEM5 power connector; an adapter is not included and must be sourced from your power supply or purchased separately.
同類商品
顧客也比較過
同一分類，同一套結帳流程 — 支援八種加密貨幣，匯率鎖定 30 分鐘。
