AMD had some big reveals at Computex 2024 including its next-gen Zen 5 architecture, but new AI accelerators were also teased. This includes the Instinct MI325X “CDNA 3”, the MI350X “CDNA 4” and the next-gen MI400 “CDNA Next” AI accelerators for data centers, and the cloud.
5VIEW GALLERY – 5 IMAGES
AMD’s new Instinct MI325X AI accelerator will use the same CDNA 3 architecture as the just-launched Instinct MI300 series AI accelerators, with a monster 288GB of HBM3E memory with up to 6TB/sec of memory bandwidth. We’ll see 1.3 PFLOPs of FP16, and up to 2.6 PFLOPs of FP8 compute performance, with up to 1 trillion parameters per server.
AMD compared its upcoming Instinct MI325X AI accelerator against NVIDIA’s new beefed-up Hopper H200 AI GPU:
- 2x Memory
- 1.3x Memory Bandwidth
- 1.3x Peak Theoretical FP16
- 1.3x Peak Theoretical FP8
- 2x Model Size per Server
5AMD plans to launch the refreshed Instinct MI325X AI accelerator in Q4 2024, featuring the same chiplet housing structure as the current MI300X, so we should see 12-Hi HBM3E stacks, which opens the door to that huge 288GB of HBM3E memory.
5Moving onto the next-gen AI accelerator announcements, is the new AMD Instinct MI350X which will be released in 2025. The new Instinct MI350 series will be based on TSMC’s newer 3nm process node, also offering up to a huge 288GB of HBM3E memory, and support for FP4/FP6 data types that NVIDIA’s new Blackwell AI GPUs support.
5AMD’s new Instinct MI350X AI accelerator will be based on the next-gen CDNA 4 architecture, and will arrive with OAM compatibility.
AMD also teased its future Instinct AI accelerators, where in 2026 we’ll be introduced to the next-gen Instinct MI400 series which is based on a future-gen CDNA architecture that the company has dubbed “CDNA Next”.
- The new AMD Instinct MI325X accelerator, which will bring 288GB of HBM3E memory and 6 terabytes per second of memory bandwidth, use the same industry standard Universal Baseboard server design used by the AMD Instinct MI300 series, and be generally available in Q4 2024. The accelerator will have industry leading memory capacity and bandwidth, 2x and 1.3x better than the competition respectively4, and 1.3x better5 compute performance than competition.
- The first product in the AMD Instinct MI350 Series, the AMD Instinct MI350X accelerator, is based on the AMD CDNA 4 architecture and is expected to be available in 2025. It will use the same industry standard Universal Baseboard server design as other MI300 Series accelerators and will be built using advanced 3nm process technology, support the FP4 and FP6 AI datatypes and have up to 288 GB of HBM3E memory.
- AMD CDNA “Next” architecture, which will power the AMD Instinct MI400 Series accelerators, is expected to be available in 2026 providing the latest features and capabilities that will help unlock additional performance and efficiency for inference and large-scale AI training.