MI325X vs MI300X
Spec-sheet and pricing comparison of AMD Instinct MI325X and AMD Instinct MI300X with links to continuously measured LLM inference benchmarks on identical workloads.
Spec-sheet comparison
| MI325X | MI300X | MI325X / MI300X | |
|---|---|---|---|
| Memory per chip | 256 GB HBM3e | 192 GB HBM3 | 1.33x |
| Memory bandwidth | 6 TB/s | 5.3 TB/s | 1.13x |
| Dense FP8 compute | 2,615 TFLOP/s | 2,615 TFLOP/s | 1.0x |
| Dense FP4 compute | Not supported | Not supported | n/a |
| TDP | 1,000 W | 750 W | 1.33x |
| Hourly rate (neocloud tier) | $1.32/hr | $1.16/hr | 1.14x |
| Scale-up world size | 8 chips | 8 chips | 1.0x |
Ratios are spec-sheet values; see the live compare pages for measured deltas.
Frequently asked questions
- Which has more memory, MI325X or MI300X?
- MI325X offers 256 GB HBM3e per chip versus 192 GB HBM3 on MI300X (1.33x the capacity).
- How do MI325X and MI300X prices compare?
- At the neocloud tier the SemiAnalysis TCO model rates MI325X at $1.32/hr versus $1.16/hr for MI300X. Hourly price alone is misleading; the per-dollar compare pages divide measured throughput by these rates.
- Is MI325X faster than MI300X for LLM inference?
- On paper MI325X has 1.0x the dense FP8 compute of MI300X, but delivered tokens per second depend on the model, framework, precision and interactivity target. InferenceX measures both chips daily on identical workloads; see the live compare pages for current results.
See live benchmark results
Every number above is static hardware data. Delivered tokens per second, cost per million tokens and energy per token are measured continuously on the dashboard:
Go deeper with the SemiAnalysis models
InferenceX measures delivered inference performance. The SemiAnalysis institutional models cover the market behind these chips: who ships them, who buys them, and what they cost to own.
SemiAnalysis Accelerator & HBM Model
SKU-level AI accelerator shipments, pricing and specifications, from foundry wafer starts and HBM supply through customer-level installed base, quarterly with multi-year forecasts.
SemiAnalysis AI Cloud TCO Model
The source of the hourly rates on this page: all-in GPU cost of ownership built up from server capex, power, colocation and cost of capital, with rental price scenarios and a full cluster finance suite.