MI325X
MI325X is an AMD GPU based on the CDNA 3 chiplet architecture, first released in 2025.
Built from BenchGecko data as of April 14, 2026 · updates when the data changes
MI325X is an AMD GPU based on the CDNA 3 chiplet architecture, first released in 2025.
Basic
The AMD Instinct MI325X is an AI GPU made by AMD, codenamed Aqua Vanjaram+. Peak throughput is 1,307 TFLOPS at FP16/BF16 and 2,614 TFLOPS at FP8. It carries 256 GB of HBM3e with 6 TB/s of memory bandwidth. Rated power (TDP) is 1,000 W. It is built on N5+N6 at TSMC.
Deep
In the BenchGecko dataset it sits in the frontier tier: current-generation flagship silicon shipping at scale to hyperscalers. It follows the MI300X and is followed by the MI355X. Specs come from manufacturer datasheets; the spec card on this page shows the dataset values and their as-of date.
Expert
That is 1.31 FP16 TFLOPS per watt. Memory bandwidth per unit of compute is 4.59 GB/s per FP16 TFLOP: a higher ratio helps memory-bound inference, while peak TFLOPS matter more for compute-bound training. Fabrication: Taiwan. Real throughput depends on the model, precision, batch size and software stack, so compare tokens per second per dollar for your own workload.
Depending on why you're here
- ·MI325X: 1,307 TFLOPS FP16/BF16 · 2,614 TFLOPS FP8
- ·Process: N5+N6 at TSMC
- ·Tier: frontier
- ·256 GB of HBM3e per chip
- ·Cloud availability in the dataset: TensorWave
- ·Compare it with other chips on /hardware/mi325x
- ·AMD · fabbed at TSMC
- ·Supply tightness is tracked on /hardware
- ·MI325X is a chip built to run AI, made by AMD
- ·It has 256 GB of fast memory so large models fit
- ·Companies buy or rent these to train and serve AI models
Frequently Asked Questions
Read the primary sources
- MI325X announcementwww.amd.com