FPGA computing platform

M-KUBOS

To　achieve high-speed data processing,
such as edge computing FPGA computing platform

Features

Possibility of M-KUBOS

5G × MEC

Edge computing indispensable in the 5G era

By utilizing the large capacity, ultra-low latency, and multiple simultaneous connection communication infrastructure provided by 5G, some of the processing that was conventionally performed in the cloud can be performed in a small data center of the base station to control the factory. Timing-critical processing such as urban traffic control and power control can be efficiently realized. Such an edge computing method is called MEC (Multi-Access Edge Computing or Mobile Edge Computing), and by enhancing the functionality of the edge area, encapsulation of personal information, local management of IoT, and information congestion to the cloud are performed. It enables seamless cooperation with the cloud and information management through pre-processing to prevent it. For example, high-precision lesion detection in diagnostic imaging in hospitals, sensor data analysis to improve production efficiency in factory facilities, virtual reality (VR) in stadiums, and so on.

FiC × MEC

Multi-FPGA system FiC that enables high-speed processing of MEC

Currently, FPGAs are attracting attention as accelerators for high-speed calculations that execute algorithms in hardware. The multi-FPGA system FiC (Flow-in Cloud), which is led by Professor Hideharu Amano of the Faculty of Science and Technology of Keio University, is a large-scale system in which multiple medium-sized FPGAs with excellent price / performance ratios are connected by both high-speed links. The computing system can form a huge circuit as a whole. It also has the ability to partially reorganize while the system is running, allowing for flexible application changes and multi-user use. In addition, because it uses a circuit-switched method, it is also suitable for applications that require synchronization and predict data size, such as deep learning. By utilizing FiC for MEC, high-speed processing on the edge side can be realized, and quasi-real-time processing can be performed with low power consumption.

Equipment main specifications

item	specification
Onboard FPGA	XCZU19EG-2FFV1760
memory	PS：4GB DDR4-2400 PL: 1x DDR4-2400 SODIMM socket
storage	SATA (not implemented)
I/O	QSFP28 (not implemented) 4x GTY 4TX/4RX (max 28.125Gbps) 4x GTH 8TX (max 16.3Gbps) 4x GTH 8RX (max 16.3Gbps) USB3.0×1 USB-UART×1 1Gb Ether(RJ45) DP1.2
Expansion connector	FMC connector x 2 (LPC, GT are not assigned) PMOD×2
switch/LED	Mounted on the board
Power supply	ATX
Operating temperature	0~50℃
Humidity during operation	20-80% (no condensation)
External dimensions	MicroATX board size 　(243.8mm×243.8mm)

Related product

M-KUBOS can be connected to PALTEK's own developed product, PCIe platform "DATA BRICK" with FireFly ™, and can process 8K video data and 8K video.
It can be used as the AI platform used (selecting disease fragments from endoscopic images, etc).