| Product | AMD Instinct™ MI100 Accelerator - 32GB HBM2 - PCIe 4.0 x16 - Passive Cooling | NVIDIA® A40 GPU Computing Accelerator - 48GB GDDR6 - PCIe 4.0 x16 - Passive Cooling | NVIDIA® L40S ADA GPU Computing Accelerator - 48GB GDDR6 - PCIe 4.0 x16 - Passive Cooling | NVIDIA® RTX A2000 - 12GB GDDR6 - PCIe 4.0 x16 - Active Cooling (4x mDP) | NVIDIA® RTX A6000 - 48GB GDDR6 - PCIe 4.0 x16 - Active Cooling (4xDP) |
| Price Change | |||||
| Action | Select | Select | Select | Select | Select |
| Main Specifications | |||||
| Product Series | AMD Instinct | Nvidia A40 | Nvidia L40S | ||
| Core Type | NVIDIA TENSOR | NVIDIA TENSOR | |||
| Core Clock Speed | 1502 MHz | ||||
| Host Interface | PCI Express 4.0 x16 | PCI Express 4.0 x16 | PCI Express 4.0 x16 | PCI Express 4.0 x16 | PCI Express 4.0 x16 |
| GPU Architecture | CDNA | Ampere | Ada Lovelace | ||
| Product Type | Workstation | Workstation | |||
| Product Line | NVIDIA Professional Graphics | NVIDIA Professional Graphics | |||
| Memory Technology | GDDR6 with ECC | GDDR6 | |||
| Memory Capacity | 12 GB GDDR6 with ECC | 48 GB | |||
| Max Displays | 4 Displays | ||||
| Detailed Specifications | |||||
| Streaming Processor Cores | 7,680 | 10752 CUDA Cores | 18,176 | 3328 CUDA Cores | 10752 Shading Units |
| Compute Units | 120 | ||||
| NVIDIA Tensor Cores | 336 Tensor Cores | 568 | Gen 4 | 104 | 336 | |
| NVIDIA RT Cores | 84 RT Cores | 142 | Gen 3 | 26 | 84 | |
| Memory Clock Speed | 1.2 GHz | 6001 MHz | 2000 MHz 16 Gbps effective | ||
| Memory Interface | 4096-bit | 384-bit | 192-bit | 384-bit | |
| Memory Speeds (GT/s) | 14.5Gbps GDDR6 | ||||
| Max Memory Size | 32 GB HBM2 | 48 GB GDDR6 with error-correcting code (ECC) | 48GB GDDR6 with ECC | ||
| Max Memory Bandwidth | Up to 1228.8 GB/s | 696 GB/s | 864 GB/s | ||
| Infinity Fabric™ Links | 3 | ||||
| Peak Infinity Fabric™ Link Bandwidth | 92 GB/s | ||||
| INT8 Tensor Core | 733 teraFLOPS | ||||
| TF32 Tensor Core | 183 teraFLOPS | ||||
| FP32 | 91.6 teraFLOPS | ||||
| Peak BFLOAT16 Tensor Core | 362.05 teraFLOPS | ||||
| Peak FP16 Tensor Core | 362.05 teraFLOPS | ||||
| Peak FP8 Tensor Core | 733 teraFLOPS | ||||
| Peak INT4 Tensor Core | 733 teraFLOPS | ||||
| Total NVLink Bandwidth | NVIDIA NVLink 112.5 GB/s (bidirectional) PCIe Gen4 16 GB/s | Not supported | |||
| Multi-Instance GPUs | No | ||||
| Tensor Performance | 63.9 TFLOPS | ||||
| NVIDIA CUDA™ Technology | Yes | ||||
| NVENC | NVDEC | 3x l 3x (includes AV1 encode and decode) | ||||
| Secure Boot with Root of Trust | Yes | ||||
| NEBS Ready | Level 3 | ||||
| Peak Half Precision (FP16) Performance | 184.6 TFLOPs | ||||
| Peak Single Precision Matrix (FP32) Performance | 46.1 TFLOPs | ||||
| Peak Single Precision (FP32) Performance | 23.1 TFLOPs | ||||
| Peak Double Precision (FP64) Performance | 11.5 TFLOPs | ||||
| Peak INT4 Performance | 184.6 TOPs | ||||
| Peak INT8 Performance | 184.6 TOPs | ||||
| Peak bfloat16 | 92.3 TFLOPs | ||||
| Transistor Count | 13.25 Billion | 28.3 Billion | |||
| DisplayPort Connectors | 3x DisplayPort 1.4 A40 is configured for virtualization by default with physical display connectors disabled. The display outputs can be enabled via management software tools. | 4x DisplayPort 1.4a | |||
| OS Support | Linux x86_64 | ||||
| Cooling | Passive | Passive | Passive | ||
| Dual Slot | yes | 2-slot Low-profile | |||
| Dimensions | 10.5" (267 mm) Board Length | 4.4" (H) x 10.5" (L) | 4.4" (H) x 10.5" (L) | 2.713" H x 6.6" L | 4.4" (H) x 10.5" (L) |
| Lithography | TSMC 7nm FinFET | Samsung 8nm | 8nm | Samsung 8nm | |
| Supplementary Power Connectors | 2x PCIe 8-pin connectors | 1x 8-pin CPU (EPS12V) | 1x 16-pin | 1x 8-pin EPS | |
| Max Graphics Card Power (W) | 300W | 300W | 350W | 70W | 300W |
| Processor | Ampere | Ampere (GA102) | |||
| Memory Bandwidth | 288 GB/sec | 768 GB/s | |||
| Core Clock Speed | 1455 MHz Base Clock 1860 MHz Boost Clock | ||||
| L2 Cache Size | 6 MB | ||||
| API Support | CUDA 8.5, OpenCL 2.0 Shader Model 6.5, OpenGL 4.6, DirectX 12 Ultimate (12_2), Vulkan 1.2 | ||||
| Texture Fill Rate | 625 GTexel/s | ||||
| Graphics Resolution | 2x 7680 x 4320 at 60 Hz | 7680 x 4320 x36 bpp at 60 Hz | |||
| Peak Double Precision FP64 Performance | 1,250 GFLOPS (1:32) | ||||
| Peak Single Precision FP32 Performance | 8.0 TFLOPS | 38.7 TFLOPS | |||
| Peak Half Precision FP16 Performance | 40.00 TFLOPS (1:1) | ||||
| Multi-GPU Scalability | NVLINK 2-way low profile (2-slot and 3-slot bridges) connects 2x NVIDIA RTX A6000 | ||||
| NVLink Interconnect | 112.5 GB/s (bidirectional) | ||||
| RT Core Performance | 15.6 TLOPS | ||||
| VR Ready | Yes | ||||
| Vulkan API | 1.2 | ||||
| DisplayPort Output | 4x DisplayPort 1.4a | ||||
| Mini DisplayPort Output | 4x mDP Latching | ||||
| Minimum Recommended Power, Single Card (W) | 700W | ||||
| Minimum Recommended Power, 2-Way (W) | 850 | ||||
| Minimum Recommended Power, 3-Way (W) | 1000 | ||||
| Minimum Recommended Power, 4-Way (W) | 1200 | ||||
| Thermal Solution | Active Heatsink | Active Heatsink | |||
| Slot Height | 2-Slot | 2-Slot | |||
| Action | Select | Select | Select | Select | Select |