Sintrones ABOX-5221G-M12X Intel Raptor Lake Refresh Industrial AI Computer with RTX Ada/Arc MXM GPU and M12 10GbE

Autonomous and driver-assist trials put GPU inference on moving vehicles, where the connectors have to hold as firmly as the model has to run. The Sintrones ABOX-5221G-M12X is the GPU industrial AI computer with M12 networking in the ABOX-5221 series, combining an MXM GPU with six Ethernet ports on screw-locking X-coded connectors.

Compute is an Intel Core i9-14900T, i7-14700T or i5-14500T with up to 24 cores and 32 threads, on the R680E chipset with up to 96 GB of ECC-capable DDR5-5600 and TPM 2.0. The MXM GPU sets the ABOX-5221G-M12X apart, taking an NVIDIA RTX Embedded 5000 Ada (16 GB, 682 AI TOPS at FP8), 3500 Ada (369 TOPS) or 2000 Ada (232 TOPS), or an Intel Arc A370E (56 TOPS at INT8) and adding two DP++ outputs beside the two HDMI 2.0b ports. Windows 11 IoT Enterprise LTSC and Ubuntu 22.04 LTS are supported.

Networking is two Marvell AQC113 10GbE ports and four Intel I226 2.5GbE ports on M12 X-coded connectors, whose screw-lock retention holds the cables under vibration, all with PXE boot and Wake-on-LAN, so a 10GbE uplink can carry aggregated camera or sensor traffic back to a server. Alongside sit eight USB 3.2 Gen 2 ports, two RS-232/422/485 ports, eight digital inputs and eight outputs. Expansion covers 4G or 5G with dual SIM, Wi-Fi and a CAN bus or GNSS module, with three M.2 SSD slots and two removable 2.5-inch SATA trays offering RAID 0, 1 and 5.

Power is 9 to 60 V DC on a five-pin terminal block with ignition detection and over-current, over-voltage, surge and reverse-polarity protection. GPU and PoE builds need a nominal 24 V supply or higher. With the GPU fitted, the ABOX-5221G-M12X operates from 0 °C to 60 °C or 70 °C depending on the module, is tested to MIL-STD-810H and EN 61373 and carries EN 50155, EN 45545-2 (R25) and E-Mark compliance. Elsewhere in the Sintrones series, the ABOX-5221PG-M12X adds PoE+ to this build and the ABOX-5221G keeps the LAN on RJ45.