Intelligent video and image analysis server system Sophon SS1 is powered by our customized tensor computing processor BM1680 and deep learning acceleration card SC1+. SS1 is designed for deep learning inference in AI applications such as video surveillance, internet videos and images, etc.
CPU | Intel E3 1275V6, 4 Cores, 3.8GHz (Max Turbo 4.2GHz) |
DL accelerating card | 2x SC1+ |
System power | 800W |
System memory | 2x 8G DDR4/2133-ECC (Max 64GB) |
Storage | 250G 2.5 inch SATA3 solid state disk (Max 6 SATA disks) |
Network interface | 2x GLAN (1000M/100M self-adaptation) |
Peripheral | 4x USB3.0 |
Display Interface | 1x HDMI |
Dimension | 380(L) x 425(W) x 177(H) mm |
Weight | 10Kg |
OS | Ubuntu 16.04 |
Toolkit | Preloaded software environment, including firmware, drivers, BMDNN libs, runtime libs...etc. |
Test sample codes | Preloaded models and testcase codes for Object Detection & Recognition |
Turn key solution for system integrators and end users
Server preload OS and deep learning environment
Holistic integration, deep optimization, co-design of hardware and software
Proven success with delivery of over 10 unique silicon chips; accumulated shipment of billions of chips
Parallel computing optimizations on multiple accelerator cards
Optimizations on data transportation and storage I/Os
Accelerate deep learning inference for specific applications
Strong improvement of inference throughput, while ensuring low latency.
Proven delivery success for over 10 silicon chips, accumulated shipment billions of chips
Cooperation with TSMC, accessing the leading semiconductor processes such as 16nm / 12nm / 7nm
Quick response and strong support for customers
Enhanced system performance
High performance, high reliability, thoughtful heat dissipation design, rigorous system testing
Adhering to the concept of open and collaborative ecosystems, all algorithms will be open-source
We seek to promote the application of artificial intelligence, deep learning algorithms and accelerated hardware integration solutions in more industries