AMD Instinct MI350X vs NVIDIA GH200 Superchip — Comparaison GPU (Apr 2026)
AMD Instinct MI350X (288GB HBM3e, 1,800 TFLOPS FP16, CDNA 4) vs NVIDIA GH200 Superchip (96GB HBM3, 989 TFLOPS FP16, Hopper)
|
AMD Instinct MI350X
288GB HBM3e · CDNA 4
|
NVIDIA GH200 Superchip
96GB HBM3 · Hopper
|
||
|---|---|---|---|
| Spécifications | |||
| Fabricant | AMD | NVIDIA | |
| Architecture | CDNA 4 | Hopper | |
| VRAM | 288 GB HBM3e | 96 GB HBM3 | |
| Bande passante | 8,000 GB/s | 4,000 GB/s | |
| FP16 (Tensor) | 1800.0 TFLOPS | 989.0 TFLOPS | |
| FP32 | 72.0 TFLOPS | 494.5 TFLOPS | |
| TDP | 1000W | 700W | |
| Année de sortie | 2025 | 2023 | |
| Segment | Data center | Data center | |
| Idéal pour | Next-gen AI training inference at scale | Large-scale AI training HPC | |
| Tarification Cloud | |||
| Le moins cher à la demande | — | — | |
| Le moins cher en Spot | — | — | |
| Fournisseurs | 1 | 0 | |
AMD Instinct MI350X
NVIDIA GH200 Superchip