Layer 2 · head to head
AMD Instinct MI350X vs AMD Instinct MI355X
Same 288 GB, same generation. The question is whether the flagship's extra compute changes your cost per token at all.
| AMD Instinct MI350X | AMD Instinct MI355X | |
|---|---|---|
| Can you get it | Rentable by the hour from multiple clouds; purchasable at scale. | Rentable by the hour from multiple clouds; purchasable at scale. |
| Memory | 288 GB HBM3E | 288 GB HBM3E |
| Bandwidth | 8 TB/s | 8 TB/s |
| Compute | FP4/FP8; below MI355X | ~10 PFLOPS sparse FP4 |
| Power | Not consistently published | Not consistently published |
| Interconnect | Infinity Fabric | Infinity Fabric |
| Software | ROCm | ROCm |
| Workload | both | both |
AMD Instinct MI350X
For: The volume part of the MI350 generation, with the same 288 GB memory advantage.
The catch: Same ROCm question as the MI355X. Check that your serving stack — vLLM, SGLang, TensorRT-LLM — actually supports your model on ROCm before committing.
Economics: AMD claims 2.6x the inference throughput of an H100 on Llama 3.1 405B.
AMD Instinct MI355X
For: Memory-bound inference. 288 GB against Blackwell's 192 GB is the cleanest technical argument AMD has made in a decade.
The catch: ROCm, not CUDA. The gap has narrowed sharply but it is still the reason procurement teams say no, and porting cost is real.
Economics: AMD positions this as the best price-per-token part for inference at scale, and claims 2.6x H100 inference throughput on Llama 3.1 405B for the MI350X.
Before either — can you power it?
Not consistently published against Not consistently published. In 2026 that comparison usually matters more than the FLOPS one: the US interconnection queue exceeds 2,600 GW with waits approaching five years, and roughly 80% of projects withdraw before energising.
Want this whole layer as one table?
The accelerators on these pages as a single CSV — memory, bandwidth, compute, power, interconnect, software stack, availability, who each part is for and the catch — ready to sort, filter or load into a notebook.
Get the silicon table — free for research and evaluation
Four short lines and the download starts at once (CC-BY-NC-4.0, built from these pages the moment you ask). What you type is used only to answer the request and is never sold. Prefer Hugging Face? Request it there.
Using it inside a product, a client report or anything you sell? That needs a commercial licence.