…
Cloud rental price, compute per dollar and specs, side by side.
Open in the interactive comparer →| Spec / price | Instinct MI300X | Instinct MI325X |
|---|---|---|
| Memory | 192 GB HBM3 | 256 GB HBM3ebetter |
| Memory bandwidth | 5,300 GB/s | 6,000 GB/sbetter |
| FP16 / BF16 (dense) | 1.31 PF | 1.31 PF |
| FP8 (dense) | 2.61 PF | 2.61 PF |
| FP4 (dense) | — | — |
| INT8 (dense) | 2.61 PF | 2.61 PF |
| TF32 (dense) | 0.654 PF | 0.654 PF |
| FP32 (dense) | 0.163 PF | 0.163 PF |
| FP64 (dense) | 0.163 PF | 0.163 PF |
| Power | 750 W | 1000 W |
| Architecture | CDNA 3 (2023) | CDNA 3 (2024) |
| Lowest on-demand | $1.71better | $2.25 |
| Median on-demand | $2.99better | $3.80 |
| Lowest spot | $1.11 | — |
| FP16/BF16 $ per PFLOP·h | $1.31better | $1.72 |
| Providers | 10 | 3 |
The Instinct MI300X and Instinct MI325X have the same dense FP16/BF16 compute. The Instinct MI325X has 1.1× more memory bandwidth than the Instinct MI300X. Real-world speedups depend on the workload: training and prefill scale with compute, while LLM decoding scales mostly with memory bandwidth.
Right now the Instinct MI300X is cheaper per unit of compute: $1.31 vs $1.72 per PFLOP-hour at the lowest on-demand prices.
Running one GPU around the clock at the lowest on-demand price costs about $1,248 per month for the Instinct MI300X and $1,643 for the Instinct MI325X.
The Instinct MI325X has more memory: 256 GB vs 192 GB.