Deep learning inference at the edge
(PresseBox) - The realization of deep learning inference (DL) at the edge requires a flexibly scalable solution that is power efficient and has low latency. At the edge mainly compact and passive cooled systems are used that make quick decisions without uploading data to the cloud.
The new Mustang-V100 AI accelerator card from ICP Deutschland supports developers by integrating AI training models successfully at the edge. Eight Intel® Movidius? Myriad? X MA2485 Vision Processing Units (VPUs) are integrated on a PCIe based expansion card. Due to its low power consumption of 2.5W per VPU, it is suitable for particularly demanding low-power KI applications at the edge. Each individual VPU can be assigned a different DL topology. The reason for this is the multi-channel execution capability of the VPUs, which enables the simultaneous execution of calculations. This allows different applications such as object recognition or image and video classification to be executed simultaneously.
In addition, the compatibility of the OpenVINO? toolkit from Intel® optimizes the performance of the training model and scales it to the target system at the edge. Software developers benefit in two ways - through fast and optimized integration without tedious trial and error. The Mustang-V100 is compatible with a variety of popular operating systems such as Ubuntu 16.04, CentosOS 7.4 and Windows 10 IoT and supports numerous architectures and topologies of artificial neural networks.
Specifications
? AI-Accelerator card with Intel® Movidius? Myriad? X MA2485 VPU
? Single Slot PCIe x4 interface
? Operating temperature: 5°C~55°C
? Low power consumption: <30W TDP
? Actively cooled
? Support of different ANN topologies
Applications
? Multi-channel excecution
? Acceleration of Deep Learning Inference
? Low power applications
Product Link https://www.icp-deutschland.de/industrie-pc/cpu-boards-cpu-karten/computing-accelerator/mustang-ca-cards/mustang-v100-mx8-r10.html
Unternehmensinformation / Kurzprofil:
Datum: 12.03.2019 - 14:32 Uhr
Sprache: Deutsch
News-ID 589477
Anzahl Zeichen: 2302
contact information:
Town:
Reutlingen
Kategorie:
Hazadous Materials Management
Diese Pressemitteilung wurde bisher 216 mal aufgerufen.
Die Pressemitteilung mit dem Titel:
"Deep learning inference at the edge"
steht unter der journalistisch-redaktionellen Verantwortung von
ICP Deutschland GmbH (Nachricht senden)
Beachten Sie bitte die weiteren Informationen zum Haftungsauschluß (gemäß TMG - TeleMedianGesetz) und dem Datenschutz (gemäß der DSGVO).
