Instance Details
| Compute | Value |
|---|---|
| vCPUs | 192 |
| Memory (GiB) | 768 |
| Memory per vCPU (GiB) | 4 |
| Physical Processor | Intel Xeon Scalable (Granite Rapids) |
| Clock Speed (GHz) | 3.9 GHz |
| CPU Architecture | x86_64 |
| GPU | 8 |
| GPU Average Wattage | 0 W |
| GPU Architecture | NVIDIA RTX PRO 4500 Blackwell |
| Video Memory (GiB) | 256 |
| GPU Compute Capability ? | 12 |
| FPGA | 0 |
| ffmpeg FPS | 301 |
| CoreMark iterations/Second | 5701982.329794 |
| NUMA Architecture | Value |
|---|---|
| Uses NUMA Architecture ? | Yes |
| NUMA Node Count | 2 |
| Max NUMA Distance | 21 |
| Cores per NUMA Node (Avg) | 96.0 |
| Threads per NUMA Node (Avg) | 96.0 |
| Memory per NUMA Node (Avg MB) | 371836 MB |
| L3 Cache per NUMA Node (Avg MB) | 480.0 MB |
| L3 Cache Shared | Yes |
| Networking | Value |
|---|---|
| Network Performance (Gibps) | 700 |
| Enhanced Networking | true |
| IPv6 | true |
| Placement Group ? |
| Storage | Value |
|---|---|
| EBS Optimized | true |
| Max Bandwidth (Mbps) on EBS | 80000 |
| Max Throughput (MB/s) on EBS | 10000 |
| Max I/O operations/second IOPS | 320000 |
| Baseline Bandwidth (Mbps) on EBS | 80000 |
| Baseline Throughput (MB/s) on EBS | 10000 |
| Baseline I/O operations/second IOPS | 320000 |
| Devices | 2 |
| Swap Partition | false |
| NVME Drive ? | true |
| Disk Space (GiB) | 3800 |
| SSD | true |
| Initialize Storage | false |
| Instance Store Read IOPS | 2,200,000 |
| Instance Store Write IOPS | 1,700,000 |
| Amazon | Value |
|---|---|
| Generation | current |
| Date Introduced | 2026-06-18 |
| Instance Type | g7.48xlarge |
| Family | GPU instance |
| Name | G7 48xlarge |
| Elastic Map Reduce EMR | false |
