Montage Technology, a Chinese semiconductor IP company, has announced its 2nd Generation Jintide processors based on Intel’s Xeon CPU with some security-related customizations.
The new processor features more cores than the previous-generation model and runs at a substantially higher frequency, boasting 26 cores with simultaneous multithreading technology running at up to 3.20GHz as well as integrate 35.75MB of last level cache (1.375MB per core).
The 2nd Generation Jintide CPU is based on Intel’s 2nd Generation Xeon Scalable processor featuring the Cascade Lake microarchitecture and therefore supports Intel’s deep learning (DL) boost technology (the AVX512VNNI instruction).
- We've put together a list of the the best dedicated server hosting providers
- These are the best small business servers available
- We've also highlighted the best cloud storage services
Semi-custom Xeons for China
The key difference between Montage’s Jintide (opens in new tab) processor platform (opens in new tab) and Intel’s Xeon Scalable platform is support for hardware-based security enhancements on the platform level, including PrC (Pre-check) and DSC (Dynamic Security Check) technologies. To realize advantages of Montage’s Jintide, server makers have to use Montage’s HSDIMM (Hybrid Security Dual Inline Memory Module) available exclusively from the company.
DSC technology was originally developed by Tsinghua University. The method relies on a special security check engine that tracks CPU behavior, does memory tracing, and performs I/O tracing. In addition, the engine can offload certain cryptography workloads (using Chinese algorithms) using built-in or external hardware.
Montage’s Jintide processors are exclusively available in China. Montage has reportedly demonstrated several server models running Montage’s 2nd Generation Jintide CPUs. It is not clear when the new machines are set to be available, but it is logical to expect adopters of 1st Generation Jintide platforms to adopt their successors because of considerably higher performance and additional deep learning capabilities.
Source: cnTechPost (opens in new tab)