AMD Instinct MI400 Leaked in Recent Updates: Expected to Have Up to 8 Chiplets on Dual Interposer Dies

AMD’s upcoming Instinct MI400 accelerators are set to introduce a transformative architecture featuring dedicated interposer dies that can support up to 8 chiplets. This groundbreaking innovation was recently highlighted in various development patches.

Introducing the AMD Instinct MI400 Accelerators: Key Features Unveiled

The next generation of AMD’s Instinct accelerators, the MI400, promises a significant shift in design that incorporates advanced tiles. While AMD continues to finalize the MI350, with a planned release this year, preliminary insights into the MI400, scheduled for launch in 2026, are emerging.

Recent data leaks have revealed intriguing details regarding the MI400’s architecture. Although AMD has yet to officially disclose comprehensive specifications, recent patches provide a tantalizing glimpse into what the new accelerators will entail.

Notably reviewed by Coelacanth Dream, updates posted on the Free Desktop indicate that the MI400 will now include up to Four XCDs (Accelerated Compute Dies), a substantial upgrade from the previous model’s two XCDs per Active Interposer Die (AID).Additionally, the MI400 will implement two AIDs along with distinct Multimedia and I/O dies.

MI400 Patch — Image Source: FreeDesktop.org

Each AID will feature a dedicated MID tile designed to optimize communication between compute units and I/O interfaces, marking a noteworthy advancement over earlier models. AMD’s MI350 already utilizes Infinity Fabric for inter-die communication; however, the MI400’s enhancements are aimed specifically at meeting the demands of large-scale AI training and inference applications. This new architecture, likely to be rebranded as UDNA under AMD’s strategy to unify RDNA and CDNA technologies, showcases AMD’s commitment to innovation.

In parallel, AMD is preparing to unveil the MI350 accelerators employing the CDNA 4 architecture this year. These accelerators are set to deliver remarkable improvements over the MI300 series, featuring a cutting-edge 3nm process node and enhanced energy efficiency. Expectations include up to a 35-fold increase in AI inference capabilities when compared to the MI300. Details regarding the performance of the MI400 have yet to be disclosed, making the community eagerly anticipate upcoming announcements.

Specifications Overview: AMD Instinct AI Accelerators

Accelerator Name	AMD Instinct MI400	AMD Instinct MI350X	AMD Instinct MI325X	AMD Instinct MI300X	AMD Instinct MI250X
GPU Architecture	CDNA Next / UDNA	CDNA 4	Aqua Vanjaram (CDNA 3)	Aqua Vanjaram (CDNA 3)	Aldebaran (CDNA 2)
GPU Process Node	TBD	3nm	5nm+6nm	5nm+6nm	6 nm
XCDs (Chiplets)	8 (MCM)	8 (MCM)	8 (MCM)	8 (MCM)	2 (MCM) 1 (Per Die)
GPU Cores	TBD	TBD	19, 456	19, 456	14, 080
GPU Clock Speed	TBD	TBD	2100 MHz	2100 MHz	1700 MHz
INT8 Compute	TBD	TBD	2614 TOPS	2614 TOPS	383 TOPs
FP6/FP4 Compute	TBD	9.2 PFLOPs	N/A	N/A	N/A
FP8 Compute	TBD	4.6 PFLOPs	2.6 PFLOPs	2.6 PFLOPs	N/A
FP16 Compute	TBD	2.3 PFLOPs	1.3 PFLOPs	1.3 PFLOPs	383 TFLOPs
FP32 Compute	TBD	TBD	163.4 TFLOPs	163.4 TFLOPs	95.7 TFLOPs
FP64 Compute	TBD	TBD	81.7 TFLOPs	81.7 TFLOPs	47.9 TFLOPs
VRAM	TBD	288 HBM3e	256 GB HBM3e	192GB HBM3	128 GB HBM2e
Infinity Cache	TBD	TBD	256 MB	256 MB	N/A
Memory Clock	TBD	8.0 Gbps?	5.9 Gbps	5.2 Gbps	3.2 Gbps
Memory Bus	TBD	8192-bit	8192-bit	8192-bit	8192-bit
Memory Bandwidth	TBD	8TB/s	6.0 TB/s	5.3 TB/s	3.2 TB/s
Form Factor	TBD	OAM	OAM	OAM	OAM
Cooling	TBD	Passive Cooling	Passive Cooling	Passive Cooling	Passive Cooling
TDP (Max)	TBD	TBD	1000W	750W	560W