Question 1

What AI models can run on MI300X?

Accepted Answer

The MI300X (192 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 192 GB VRAM, potentially using quantization (Q8, Q6, or Q4) to reduce memory usage.

Question 2

What is the cheapest cloud option for MI300X?

Accepted Answer

The cheapest on-demand option for the MI300X is RunPod at $2.19/hr (1x_MI300X_SECURE). Spot/preemptible pricing is available at $1.99/hr. We track pricing across 14 providers to help you find the best deal.

Question 3

Is MI300X good for fine-tuning?

Accepted Answer

Yes, the MI300X is excellent for fine-tuning. With 192 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 MI300X have?

Accepted Answer

The MI300X has 192 GB of VRAM (HBM3). It also offers 5325 GB/s memory bandwidth, 1,307 TFLOPS of FP16 compute, FP8 support for efficient inference.

Question 5

What quantization formats work on MI300X?

Accepted Answer

The MI300X 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 ~652B parameters (by VRAM estimate) fit on the MI300X.

MI300X

Specifications

Cloud pricing

Compatible models

Compare this GPU

Frequently asked questions

VRAM	192 GB
Memory type	HBM3
Bandwidth	5325 GB/s
FP16 TFLOPS	1,307
FP8 support	Yes