# AMD Instinct MI355X

14 Hacker News commenters recommended AMD Instinct MI355X and 5 criticized it, November 2025 to November 2026. It was recommended in 7 of 13 months. It meets the Hacker Approved badge rule.

**Verdict:** Strong memory and performance per dollar for large AI inference and training; critics say its software stack lags and it trails Nvidia's B300 on some benchmarks.

## Why people recommend it

- Capable hardware, competitive raw performance (6 comments): "AMD compute cards are sold under the Insinct line and are vastly more powerfull than their GPUs." ([source](https://news.ycombinator.com/item?id=46196248)) "Our hardware is arguably better (spec for spec) apart from critical areas like memory bandwidth" ([source](https://news.ycombinator.com/item?id=45934057)) "their Instinct parts are actually, at a hardware level, very very capable pieces of kit" ([source](https://news.ycombinator.com/item?id=45933833))
- Lots of VRAM for the price (4 comments): "The metric where AMD usually comes out on top is perf/$. Or with their instinct cards VRAM/$" ([source](https://news.ycombinator.com/item?id=47750471)) "With 288GB per GPU the MI355X can fit the model within one 8 GPU node." ([source](https://news.ycombinator.com/item?id=49200959)) "leads in VRAM and throughput but trails in accuracy (slightly)" ([source](https://news.ycombinator.com/item?id=46081689))
- Better performance per dollar and watt (3 comments): "The operating cost is so much lower it's paid for itself plus an additional $40k in the first year." ([source](https://news.ycombinator.com/item?id=49531584)) "The metric where AMD usually comes out on top is perf/$. Or with their instinct cards VRAM/$" ([source](https://news.ycombinator.com/item?id=47750471)) "they are better per dollar, better per watt, and out competing on raw performance" ([source](https://news.ycombinator.com/item?id=48910329))
- Good for inference and training (3 comments): "AMD has been doing inference chips for many years and are the leader for HPC." ([source](https://news.ycombinator.com/item?id=47863448)) "running our training workloads on the beefy Mi355x is awesome" ([source](https://news.ycombinator.com/item?id=49406601)) "MI355 are quite good for inference." ([source](https://news.ycombinator.com/item?id=48044797))

## What people warn about

- Trails B300 on benchmarks (2 comments): "leads in VRAM and throughput but trails in accuracy (slightly)" ([source](https://news.ycombinator.com/item?id=46081689)) "On every row the B300 beat the MI355X" ([source](https://news.ycombinator.com/item?id=49141794))
- Weak developer tools and ROCm software (2 comments): "Benchmarks are unreproducible, power costs are missing, ROCm was patched" ([source](https://news.ycombinator.com/item?id=49146725)) "Developer ux on amd is terrible." ([source](https://news.ycombinator.com/item?id=48910341))
- Power costs and draw unclear (2 comments): "Benchmarks are unreproducible, power costs are missing, ROCm was patched" ([source](https://news.ycombinator.com/item?id=49146725)) "A single MI355 is ~230W, just idle." ([source](https://news.ycombinator.com/item?id=46385986))

## Alternatives hackers recommend

- [NVIDIA H100](https://hackerapproved.com/p/nvidia-h100.md): 20 recommend
- [NVIDIA Tesla V100](https://hackerapproved.com/p/nvidia-tesla-v100.md): 16 recommend
- [Nvidia A100](https://hackerapproved.com/p/nvidia-a100-gpu.md): 10 recommend
- [Intel D270 GPU AI Accelerator](https://hackerapproved.com/p/intel-d270-gpu-ai-accelerator.md): 10 recommend
- [NVIDIA B200](https://hackerapproved.com/p/nvidia-b200.md): 9 recommend

## Where to get it

- [Visit the official site](https://www.amd.com/en/products/accelerators/instinct.html)

## Cite as

Hacker Approved, "AMD Instinct MI355X", https://hackerapproved.com/p/amd-instinct/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).