Q9 Versal Computer-on-Module
Xiphos Technologies
High-performance AMD Versal AI-edge compute module with 152 AI Engine tiles, 2 TB NVMe storage, and 30-50 W power budget for satellite AI/ML payloads.
Key figurescomparable across the catalogue
- Mass
- 0.056 kg
- Power
- 30 W
- Largest dimension
- 110 mm
Normalised to common units so this product can be compared with others in its category. The supplier's own wording is below.
As published by the supplier
- MPSoC
- AMD Versal SoC
- Application CPU
- Dual-core ARM Cortex-A72 @ 1.7 GHz
- Real-Time CPU
- Dual-core ARM Cortex-R5F @ 750 MHz
- AI Engines
- 152 AI Engine-ML tiles
- System Logic Cells
- 820,000
- DSP Slices
- 984
- PS LPDDR4 RAM
- 8 GB
- PL LPDDR4 RAM
- 8 GB
- QSPI Flash (NOR)
- 1 GB
- NVMe SSD
- 2x 1,000 GB (independent buses)
- MRAM
- 1 MB
- GTY Transceivers
- 26.5625 Gbps (PCIe Gen 4, Ethernet, JESD204B)
- Control FPGA
- Microchip PolarFire
- Dimensions
- 90 x 110 x 20 mm
- Mass
- 56 g (without connectors)
- Power (Typical)
- 30 to 50 W
- Input Voltage
- 8 to 14 V DC
- Operating Temperature
- -40 to +60 deg C
- Operating System
- Yocto Linux BSP (LTS)
About
The Xiphos Q9 is a high-performance AI and edge-processing system-on-board built around the AMD Versal SoC, representing the first time a Versal-based platform has been available in such a small form factor. The Q9 extends the Xiphos Q-Card lineup to mission profiles demanding 30-50W of processing capability, particularly real-time sensor processing using AI/ML technologies.
The AMD Versal device integrates 152 AI Engine-ML tiles alongside dual-core ARM Cortex-A72 application processors at 1.7 GHz, dual-core Cortex-R5F real-time processors at 750 MHz, and 820,000 system logic cells with 984 DSP slices. This heterogeneous compute architecture allows simultaneous execution of control tasks, FPGA signal processing, and neural network inference workloads without the inter-chip latency of discrete AI accelerator solutions.
The Q9 includes 16 GB of LPDDR4 DRAM across two independent PS- and PL-accessible interfaces, 1 GB of QSPI NOR flash, and two independent 1 TB NVMe SSDs for high-speed mass memory. High-speed JESD204B, PCIe Gen 4, and 100G Ethernet connectivity via 26.5625 Gbps GTY transceivers enable direct connection to high-bandwidth sensors, RF front-ends, and downlink transmitters.
Documentation
Need the full ICD, test reports or a specific revision? Ask the supplier directly.