Powering Trillion-Parameter Models
KRS8000V3 is an L11 AI rack based on NVIDIA GB200 NVL72, integrating 36 Grace CPUs and 72 Blackwell GPUs in a rack-scale, liquid-cooled architecture, achieving breakthrough performance in real-time trillion-parameter large language model (LLM) inference and training.
KRS8000V3 with GB200 NVL72 is poised to redefine performance benchmarks for AI, HPC, and data analytics, making it a pivotal component in next-generation computing infrastructure.
LLM Inferencevs H100
LLM Trainingvs H100
Energy Efficiencyvs H100
Data Processingvs H100
 |
Blackwell Rack-Scale ArchitectureConnects 72 Blackwell GPUs via NVIDIA® NVLink™ Delivers 130 TB/s of low-latency communication bandwidth Acts as a single massive GPU for efficient processing |
|
|
|
 |
Performance EnhancementsAchieves 30X faster real-time trillion-parameter LLM inference compared to previous generations 4X faster training for large language models using FP8 precision |
|
|
|
 |
Data ProcessingIncludes a hardware decompression engine supporting LZ4, Deflate, and Snappy formats Provides up to 800 GB/s decompression throughput |
|
|
|
 |
Memory and BandwidthOffers 8 TB/s high memory bandwidth Grace CPU NVLink-C2C interconnect ensures high-speed data transfer |
KRS8000V3 Specifications
| NVIDIA® GB200 NVL72 | |
|---|---|
| Configuration | 36 Grace CPU and 72 Blackwell GPUs |
| NVLink Bandwidth | 130TB/s |
| GPU Memory | Bandwidth | Up to 13.39 TB HBM3e | 576 TB/s |
| CPU Memory | Bandwidth | Up to 17.28 TB LPDDR5X | Up to 18.4 TB/s |
| CPU Core Count | 2,592 Arm® Neoverse V2 cores |
| FP4 Tensor Core | 1,440 PFLOPS |
| FP8/16 Tensor Core | 720 PFLOPS |
| INT8 Tensor Core | 720 POPS |
| FP16/BF16 Tensor Core | 360 PFLOPS |
| TF32 Tensor Core | 180 PFLOPS |
| FP32 | 6,480 TFLOPS |
| FP64 / FP64 Tensor Core | 3,240 TFLOPS |
| Rack Specifications | |
|---|---|
| Dimensions | 600mm (23.6″) W x 2236mm (88″) H x 1200mm (47.2″) L |
| NVL Config | 72x 1 |
| NV OOB Switch | Option 1: 3 x SN2201 DC Option 2: 4 x SN2201 DC |
| NVL Cartridge | 4 |
| Rack Type (Per Rack) | 9x 1U NVlink Switch Trays 18x 1U Compute Trays 8x 1U Powershelf |
| Power-Shelf | 8x 33kW |
| Busbar | 1,400A |
| Rack Manifold | Option 1: 44RU, BF Option 2: 44RU, TF |
| CDU | Option 1: L2L In-Row Option 2: L2L In-Rack Option 3: L2A Sidecar |
| Compute Tray | |
|---|---|
| CPU/GPU | 2x Grace CPUs + 4x Blackwell GPUs |
| Count | 18 per Rack |
| Cooling | 1U liquid cooled |
| Storage | 8x E1.S |
| M.2 | 1x Onboard NVMe / SATA M.2 |
| Front I/O | 1x USB 3.0, 1x Mgmt I/O , 1x RJ45, 1x mini display port |
| N-S Networking | Support 2x FHFL PCIe 5.0 x16 (BF3 or NIC Card) |
| E-W Networking | 2x Mezzanine card on board 4x HHHL PCIe 5.0 x16 with 400G bandwidth |
| Fan | CPU region: 8x 12V 4056 hot-swap fans with N+1 redundancy |
| Management | DC-SCM BMC management module |
| TPM | Supports TPM 2.0 |
| Switch Tray | |
|---|---|
| Type | 2x NVLink X-800 Switch |
| Count | 9 per Rack |
| Bandwidth | 14.4TB/s |
| Cooling | 1U liquid cooled |
| Front IO | 2x RJ45, 1x USB, 1x UART |
