Artificial Intelligence
System on Chip
Patents
17 US Patents
Filing Date
2009~2015
Field
Semiconductor Design and Production
Introduction
Background
•Hybrid-systems, Cloud (fixed resource) , Grid (dynamic resource)
•Data: text, media (graphics, sound, video, sensors, etc), big data, AI
Components
•Processors: CPU, GPU, DSP, FPGA, NPU, etc
•Memories: HBM, memory virtualization (L2 & L3)
•Able to use Vendor technologies
Hybrid-functions
•Application dependency: Consumer electronics (smart devices) vs data processing (systems: supercomputer, large servers, streaming servers. gaming server, autonomous machines, intelligent robots)
Hybrid-System & Data
•Hybrid-systems, Cloud (fixed resource) , Grid (dynamic resource), Streaming servers
•Unified Date Bus, Unified Data, Unified Memory
Core Technologies
For instance, when a CPU completes a data processing task, the data may still require graphical processing. This requires the CPU to pass the data from its memory space to the GPU memory, after which the GPU then processes the data and returns it to the CPU. This complex process adds latency and incurs a performance penalty, but shared memory allows the GPU to access the same memory the CPU was utilizing, thus reducing and simplifying the software stack.
Details
Hybrid Multicore Technology
Hybrid multifunction component system with component interface encapsulation including OS packet translator
for communication over unified data bus architecture:
A multiple generic microprocessor architectures with a set of two or more cores and groups of sub-processing components
and controlling components. Under this arrangement, different technology cores and functional components are organized
in a way that different technologies can collaborate as a single processor.
Unified Data Bus
The technology relates to multi-core integration, specifically for a multi-processing domain architecture.
Under the architecture, the first processing domain and the second processing domain can each comprise elements selected
from a group consisting of memory, I/O, cache, heterogeneous data buses, and processors.
3D Packaging
Self-organizing network with chip package having multiple interconnection configurations.
The technology relates to networks and chip packages specifically a self-organizing network with a 3D chip package
having back-end side interconnects and through silicon vias (TSVs). TSV technology is crucial for creating 3D packages
and 3D integrated circuits.
Memory and Process Sharing via Multiple Chipsets
A multiple chipset circuit design providing memory and process sharing via input/output (I/O) with virtualization.
The configuration reduces memory leakage and enhances overall performance and efficiency of the system.
Active Cache & Memory Virtualization on CPU and Memory Devices
This technology portfolio enables flexible and maximum utilization of cache memory and main memory system
that allows reconfigurable, dynamic cache, and works varies operation based on application demand from external/hybridized cores.
Low-Power Consumption for Mobile Devices
This technology portfolio improves the power consumption of AI chips on the mobile devices.
Reliability & Monitoring IN-Circuit Technology
This technology portfolio enables semiconductor degradation monitoring and modeling using a sensor within functional circuit blocks
and at critical locations.
Patents List
Will be provided upon request (17 patents)




