Name: MI325X
Brand: AMD

Question 1

What AI models can run on MI325X?

Accepted Answer

The MI325X (256 GB VRAM) can run models including Llama-3.2-3B-Instruct-SpinQuant_INT4_EO8, Llama-3.2-1B-Instruct-SpinQuant_INT4_EO8, Llama-3.2-3B-Instruct-QLORA_INT4_EO8, Llama-2-13b, HARC-Llama-3.1-8B-Instruct and more — 10 compatible models indexed so far. Compatible models are determined by whether the model's memory footprint fits within the GPU's 256 GB VRAM, potentially using quantization (Q8, Q6, or Q4) to reduce memory usage.

Question 2

What is the cheapest cloud option for MI325X?

Accepted Answer

We don't currently have cloud pricing data for the MI325X. This GPU may not be widely available from major cloud providers, or pricing data hasn't been synced yet.

Question 3

Is MI325X good for fine-tuning?

Accepted Answer

Yes, the MI325X is excellent for fine-tuning. With 256 GB of VRAM, it can handle full fine-tuning of models up to ~30B parameters and LoRA/QLoRA fine-tuning of much larger models. It's a top-tier choice for training workloads.

Question 4

How much VRAM does MI325X have?

Accepted Answer

The MI325X has 256 GB of VRAM (HBM3e). It also offers 6000 GB/s memory bandwidth, 1,307 TFLOPS of FP16 compute, FP8 support for efficient inference.

Question 5

What quantization formats work on MI325X?

Accepted Answer

The MI325X supports the following quantization formats: FP16 (full precision), FP8, Q8 (8-bit), Q6 (6-bit), Q4 (4-bit). FP16 works for models up to ~40B parameters. FP8 halves memory vs FP16 with minimal quality loss. Q4 quantization lets models up to ~870B parameters (by VRAM estimate) fit on the MI325X.

MI325X

Specifications

Cloud pricing

Compatible models

Compare this GPU

Frequently asked questions

VRAM	256 GB
Memory type	HBM3e
Bandwidth	6000 GB/s
FP16 TFLOPS	1,307
FP8 support	Yes