Chips · accelerators · AMD
AMD Instinct MI300A · Instinct MI300
1,290 gigaflops of dense BF16 compute per watt, at 760 W
Shipped 6 Dec 2023; shipped by the first date promised. Epoch AI names it at no AI data center and 4 GPU clusters.
Embed
The frame credits AMD, Epoch AI, AI data centers and Epoch AI, Data on GPU clusters and links back to its page. Preview it.
Cite
Each figure in AMD’s own words
one chip, dense compute; memory is high-bandwidth memory- BF16, dense
- 981 TFLOPS
Peak bfloat16 980.6 TFLOPs
AMD Instinct MI300A product page - FP8, dense
- 1,960 TFLOPS
Peak Eight-bit Precision (FP8) Performance (E5M2, E4M3) 1.96 PFLOPs
AMD Instinct MI300A product page (1.96 PFLOPS on the page) - Memory
- 128 GB
Dedicated Memory Size 128 GB
AMD Instinct MI300A product page - Memory bandwidth
- 5.3 TB/s
Peak Memory Bandwidth 5.3 TB/s
AMD Instinct MI300A product page - Maximum power
- 760 W
Thermal Design Power (TDP) 550W | 760W
AMD Instinct MI300A product page - Compute per watt
- 1,290 GFLOPS/Wdense BF16 over maximum power; 2,579 GFLOPS/W at FP8
The 128 GB of HBM3 is shared by the CPU and GPU cores. AMD gives two TDP figures (550W and 760W); the record uses the higher one. The June 2022 promise names the first CDNA 3 products; the same page names MI300 as the first data center APU.
4 dated promises and the shipment
each date links to the page it is quoted from- 9 Jun 2022 Promised first CDNA 3 products (the MI300 family) in 2023.
The first AMD CDNA 3 architecture-based products are planned for 2023
AMD Financial Analyst Day 2022 press release - 13 Jun 2023 Promised launch in 2023.
led by the launch of our Instinct MI300 accelerators planned for later this year
AMD Data Center and AI Technology Premiere press release - 1 Aug 2023 Promised launch and production ramp in the fourth quarter of 2023.
on-track to launch and ramp production of MI300 accelerators in the fourth quarter
AMD second quarter 2023 financial results, 8-K exhibit 99.1 - 31 Oct 2023 Promised volume production in the fourth quarter of 2023.
AMD InstinctTM MI300A and MI300X GPUs are on track for volume production in the fourth quarter
AMD third quarter 2023 financial results, 8-K exhibit 99.1 - 6 Dec 2023 Shipped: AMD announced availability.
as well as the AMD InstinctTM MI300A accelerated processing unit (APU)
AMD Instinct MI300 Series launch press release
Named at no AI data center Epoch AI tracks
H100-equivalents and power are the whole site’s today, with every chip it holds51,184 chips in 4 GPU clusters
Epoch AI’s list of AI supercomputers, which stops in July 2025| Cluster | Owner | Country | Chips | First operational |
|---|---|---|---|---|
| Lawrence Livermore NL El Capitan Phase 2 | US Department of Energy | United States of America | 44,544 | 2024-11-18 |
| Lawrence Livermore NL Tuolumne | US Department of Energy | United States of America | 4,608 | 2024-11-18 |
| Sandia NL El Dorado | US Department of Energy | United States of America | 1,520 | 2024-11-18 |
| Lawrence Livermore NL RZAdams | US Department of Energy | United States of America | 512 | 2024-06-01 |
Epoch AI’s names for this chip: AMD Instinct MI300A.
Other AMD chips
- Instinct MI250X 684 gigaflops of dense BF16 compute per watt, at 560 W
- Instinct MI300X 1,733 gigaflops of dense BF16 compute per watt, at 750 W
- Instinct MI325X 1,300 gigaflops of dense BF16 compute per watt, at 1,000 W
- Instinct MI355X 1,786 gigaflops of dense BF16 compute per watt, at 1,400 W
- AMD’s own pages, each linked beside the figure or date it gives. Each page was read and each quote checked on 23 Oct 2026.
- Where it runs: Epoch AI, AI data centers and Epoch AI, GPU clusters, CC BY 4.0.