Luca Marchese | Aerospace and Defence Machine Learning Company


	Who_we_are	What_we_do	Technology	Customers	Partners

ATM CENTURION® is the computer mounted on the German IFV PUMA

Why not use M2 or mini-PCI TPU on embedded computers?

TPUs and GPUs on M2 or mini-PCI cards inserted inside an embedded computer seem to be an easier solution than an edge-server connected with USB3. The main problem with this solution is related to the difficulty of obtaining effective thermal dissipation with passive cooling systems in IP68 and MIL-STD-810H computers. This issue is really very important when the TPU has to run at the maximum clock frequency for long periods of time.

For this reason, rugged edge-servers are the most flexible solution to accelerate AI in the aerospace, avionics and military sectors.

Testing setup for MIND™ as AI acceleration server of the ATM CENTURION®