Taipei, March 13, 2003 --- Silicon Integrated Systems (SiS), a leading supplier of core logic and graphics chipsets, announced today that it has launched the HyperStreaming architecture, which is more powerful and suitable for the present demands of high-speed computing and for a wider range of applications of multimedia and Internet networking, following the technology of MuTOL®. SiS will demonstrate the new chipset, SiS748, which supports this technology at CeBit2003.

In order to upgrade the overall performance of PC system, SiS launched MuTOL® at the end of 2000, providing a throughput of 1.2GB between the internal North and South Bridges and the external PCI interface, promoting smoother operation to all external peripherals. At CeBit2003, SiS will launch the powerful HyperStreaming as an expansion of its mature MuTOL® technology. With HyperStreaming, the streams of a system are connected from a device control terminal to the connecting terminal of North and South Bridges, then from the connecting terminal of North and South Bridges to the FSB, and from the FSB to the memory or graphics interface, offering efficient control. It is a comprehensive and integrated solution for improving performance.

The HyperStreaming responds well to the demands of streaming technologies in movies and music playing. When the applications of Internet networking, online broadcasting become popular, the throughput demand increase dramatically as well. That is the motivity of HyperStreaming accelerate to come forth.

The HyperStreaming integrates four technologies:

First, Single Stream with Low Latency Technology enables a single stream to respond in prompt action and reduce latency in PC system. For a normal user’s access to the hard drive, through the DMA can lower the latency of a system by at least 5% compared to other platforms. When execute simple access, moreover, it can decrease the latency of a system as much as 43%. These advantages are significant and direct to users and application program that value response time.

Second, Multiple Stream with Pipelining and Concurrent Execution Technology with its integrated concurrent parallel structure effectively widens the bandwidth of data transmission and simultaneously processes non-sequential data to improve the performance of a system, bringing about more flexible operations of all peripherals. For example, when users copy and paste 1GB files between two hard drives, HyperStreaming can improve their performance by up to 15% compared to other platforms. However, with the file size increasing up to 4GB, HyperStreaming can reduce as much as 47 seconds, which represents a performance increase of 35% .

Third, Specific Stream with Prioritized Channel Technology distinguishes music and images downloaded from the Internet and allocates bandwidth and piping for listening or viewing to ensure smooth playing. Therefore, at present, Internet users can avoid from discontinuousness when using popular applications such as IP Phone, online digital broadcasting, and videoconferences. For instance, for FTP, currently a commonly used Internet data transmission protocol, throughput can be increased by at least 4% compared to other platforms and response time can also be reduced by at least 4.4%. For HTTP, another commonly used Internet protocol, throughput can be increased by at least 3% compared to other platforms, and response time can also be reduced by at least 2.6%.

Finally, Smart Stream Flow Control Technology with the smart arbiter can bring about more efficient streaming and even expand the space of processor access through effective throughput control, based on different protocols and characteristics, when dealing with the external chipset interface. This technology in turn increases the efficiency of throughput processing and execution of a system. For example, when dealing with memory access interface, this system can optimize read-write alignment based on the characteristics of the memory interface to reduce turnaround time, row conflicts, at the same time raise hit rate by access locality increase and relocation. The advantages are too comprehensive to be covered here. Please refer to all the test results of benchmark processes provided by SiS to amaze its performance.

The advantages of the HyperStreaming emphasize not only on the connecting bandwidth of the North and South Bridges, but also in the effective extension of its concurrency concept to the device terminals. For example, in the HyperStreaming, all the device control terminals have read-read-pipeline and read-write-concurrent. On the other hand, because a single device itself provides substantial concurrency, with additional HyperStreaming, it can distinguish access to every terminal and provide specific piping, enabling concurrent operations. This results in an adequate and better concurrency from device terminals to the connecting part of the North and South Bridges and the connection between the North and South Bridges.

Think about it. Obviously, the realization of such a concept in piping of the North Bridge and between the North Bridge and other interfaces constitutes an outstanding characteristic of the HyperStreaming. It is the aforementioned Smart Stream technology that contributes to such a characteristic. Smart Stream includes three functions: Smart Arbitration, Smart Streaming Flow Control and Fluent Data Transmission, which categorizes data, identifies its characteristics, increases the flow, and knows how to reduce waiting and read-write turnaround time. As a result, with commands and data scheduling, it cannot only bring about higher efficiency of the CPU and communication with its surrounding devices, but also increase the data transmission speed between the CPU and chipsets. It goes without saying that it further improves the performance of the PC system.

Are you interested in the HyperStreaming after this introduction? If so, please keep an eye on SiS748, the new chipset supporting the AMD platform at CeBit, and future series products. To learn more about HyperStreaming, please visit http://www.sis.com.