XuanTie's New Multi-Cluster Interconnect IP XL100 Supports Cache Coherence
TapTechNews July 3rd, today, XuanTie officially announced the launch of its first multi-Cluster interconnect IP - XL100 that supports cache coherence.
Multi-Cluster interconnection is a common technology in server distributed systems, which allows multiple Clusters to communicate via the network.
According to the introduction, XL100 provides 8 to 32-core processor expansion solutions and supports big.LITTLE core configuration. XL100 can reduce the bandwidth required for data transmission in multi-Cluster system interconnection and the difficulty of software programming, as well as reducing additional Cache access and optimizing system power consumption.
At the same time, XL100 provides interfaces for multiple IO devices to access the Cache to simplify the data exchange between CPUs and IO and improve performance.
Multi-core interconnection involves the efficient exchange and management of data, among which cache coherence is particularly important. This technology can ensure the consistency when multiple processor cores access shared memory, thereby reducing the complexity of software control and laying the foundation for the realization of high-performance multi-core systems.
XL100 is equipped with XT-Link - an on-chip network (NoC) interconnection technology that supports multi-processor cache coherence, which naturally supports XuanTie's multi-Cluster and has the following advantages:
Supports L2 & L3 cooperative fine-grained allocation
Supports hardware-selectable L3Cache
Supports half-power-down and full-power-down of Cache
Supports up to 7 Device Coherence Ports (DCP)
XuanTie XT-Link can meet scenarios with more than 4 cores, and can also meet scenarios with big.LITTLE core matching, and can work with XuanTie C920 and C908 processors in coordination, and in the future can support other Masters that adopt the standard ACE protocol to work together.
Currently, XuanTie XL100 can be mainly applied in scenarios with high energy efficiency requirements such as intelligent interactive terminals, AR/VR, etc., as well as in fields with high concurrent computing power requirements such as artificial intelligence, autonomous driving, network communication, etc. TapTechNews attaches the specification sheet:
