All AI inference chips

MI355X vs MI325X

Spec-sheet and pricing comparison of AMD Instinct MI355X and AMD Instinct MI325X with links to continuously measured LLM inference benchmarks on identical workloads.

Spec-sheet comparison

MI355XMI325XMI355X / MI325X
Memory per chip288 GB HBM3e256 GB HBM3e1.13x
Memory bandwidth8 TB/s6 TB/s1.33x
Dense FP8 compute5,033 TFLOP/s2,615 TFLOP/s1.92x
Dense FP4 compute10,066 TFLOP/sNot supportedn/a
TDP1,400 W1,000 W1.4x
Hourly rate (neocloud tier)$2.09/hr$1.32/hr1.58x
Scale-up world size8 chips8 chips1.0x

Ratios are spec-sheet values; see the live compare pages for measured deltas.

Frequently asked questions

Which has more memory, MI355X or MI325X?
MI355X offers 288 GB HBM3e per chip versus 256 GB HBM3e on MI325X (1.13x the capacity).
How do MI355X and MI325X prices compare?
At the neocloud tier the SemiAnalysis TCO model rates MI355X at $2.09/hr versus $1.32/hr for MI325X. Hourly price alone is misleading; the per-dollar compare pages divide measured throughput by these rates.
Is MI355X faster than MI325X for LLM inference?
On paper MI355X has 1.92x the dense FP8 compute of MI325X, 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.