https://en.wikichip.org/wiki/hisilicon/microarchitectures/taishan_v110
Edit Values
TaiShan v110 µarch
General Info
Arch Type CPU
Designer HiSilicon
Manufacturer TSMC
Introduction 2019
Process 7 nm
Core Configs 324864
Pipeline
Type Superscalar, Superpipeline
OoOE Yes
Speculative Yes
Reg Renaming Yes
Decode 4-way
Instructions
ISA ARMv8.2-A
Extensions NEON
Cache
L1I Cache 64 KiB/core
L1D Cache 64 KiB/core
L2 Cache 512 KiB/core
L3 Cache 1 MiB/core
Succession
 

TaiShan v110 is the successor to the TaiShan v100, a high-performance ARM server microarchitecture designed by HiSilicon for Huawei's own TaiShan servers.

Brands[edit]

 

TaiShan-based CPUs are branded as the Kunpeng 920 series.

Release Dates[edit]

Kunpeng 920 CPUs were officially launched in early 2019.

Architecture[edit]

 

Overview

Key changes from TaiShan v100[edit]

  • TSMC 7 nm HPC process (from 16 nm)
  • 2x core count (64, up from 32)
    • Custom cores (from Cortex-A72)

      • ASIMD

        • double SP Vector throughput (2 inst/cycle, up from 1)
  • Memory
    • 2x memory channels (8, up from 4)
  • I/O
    • PCIe Gen 4 (from Gen 3)

This list is incomplete; you can help by expanding it.

Block Diagram[edit]

Entire Chip[edit]

Memory Hierarchy[edit]

  • Cache

    • L1I Cache

      • 64 KiB/core, private
      • 64-byte cache lines
    • L1D Cache
      • 64 KiB/core, private
      • 64-byte cache lines
    • L2 Cache
      • 512 KiB/core, private
    • L3 Cache
      • 1 MiB/core
      • Shared by all cores
    • System DRAM
      • 1 TiB Max Memory / socket
      • 8 Channels
      • DDR4, up to 2933 MT/s
        • 1 DPC and 2 DPC support
      • 8 B/cycle/channel (@ memory clock)
      • ECC, SDDC, DDDC

Overview[edit]

 

Overview

Though HiSilicon has a history of designing Arm processors. The TaiShan v110 core is HiSilicons' first custom homegrown high-performance ARM core and SoC design. The chip, which incorporates multiple compute dies and an I/O is a multi-chip package, is fabricated on TSMC's 7-nanometers HPC process and integrates up to 64 cores and up to 64 MiB of last level cache.

The SoC also incorporates a number of hardware accelerators. There is a crypto engine that supports AES, DES/3DES, MD5, SHA1, SHA2, HMAC, CMAC with throughputs of up to 100 Gbit/s. Additionally, there is also a compression engine supporting GZIP, LZS, LZ4 with compression throughputs of up to 40 Gbit/s and decompression of up to 100 Gbit/s.

Marketed as the Kunpeng 920, this SoC supports up to 4-way multiprocessing support through HiSilicon's Hydra interface. In order to keep the cores fed, eight DDR4 memory channels are incorporated per socket. Additionally, designed to facilitate an easy accelerator platform, there are 40 PCIe Gen 4 lanes provided per socket with CCIX support, enabling cache coherency.

Core[edit]

Each core is a 4-way out-of-order superscalar that implements the ARMv8.2-A ISA. Huawei stated that the core supports almost all the ARMv8.4 features with a few exceptions, including dot product and the FP16 FML extension. It features private 64 KiB L1 instruction and data caches as well as 512 KiB of private L2. Though light on details, Huawei says that compared to Arm's Cortex cores, their core features an improved memory subsystem, a larger number of execution units, and a better branch predictor.

ASIMD[edit]

Each core features a single 128-bit NEON unit. It is capable of executing single double-precision FMA vector instruction per cycle or two single-precision vector instructions per cycle. Operating at 2 GHz, a 64-core chip will have a peak compute of 512 GigaFLOPS of double-precision floating point. It's worth noting that compared to the TaiShan v100, the throughput for single-precision vector has been doubled from 1 to 2 instructions per cycle.

MCP physical design[edit]

The SoC itself comprises 3 dies - two Super CPU Cluster (SCCL) compute dies and a Super IO Cluster (SICL). The SCCL compute dies contains 8 CPU Clusters (CCLs), memory controllers, and the L3 cache block. There are eight CCLs on each of the SICL dies for a total of 64 cores. The CCLs are TaiShan V110 quadplex along with the L3 cache tags partition. The Super IO Clusters include the various I/O peripherals including PCIe Gen 4, SAS, the network interface controllers, and the Hydra links.

Scalability[edit]

See also: Hydra Interface

Each chip incorporates three Hydra interface ports. The Hydra interface facilitates the cache coherency between the dies on the chip. Every link supports 240 Gb/s (30 GB/s) of peak bandwidth for a total aggregated bandwidth of 720 Gb/s (90 GB/s) in a 2-way symmetric multiprocessing configuration.

With all three links, there is also support for 4-way SMP. In this configuration, one link from each socket is connected to another socket for an all-for-all connection.

Chipset[edit]

Along with the Hi1620 SoC, HiSilicon developed a number of integrated circuits as part of the chipset platform.

Chip Description
Hi1620 CPU, Kunpeng 920 series Chip
Hi1503 CPU interconnect chip, supports scaling-up to 32 sockets
Hi1812 SSD storage controller, for read/write I/O acceleration
Hi1822 Network controller chip, DC high-speed flexible interconnect
Hi1710 BMC management chip + enhanced RAS features chip

Die[edit]

  • TSMC 7 nm HPC
  • 20,000,000,000 transistors
    • 3-4 dies

All TaiShan v110 Chips[edit]

  List of TaiShan v110-based Processors
Model Launched Cores Arch Frequency L3 TDP
920-3226 26 April 2019 32 TaiShan v110 2.6 GHz 32 MiB 120 W
920-4826 26 April 2019 48 TaiShan v110 2.6 GHz 48 MiB 158 W
920-6426 7 January 2019 64 TaiShan v110 2.6 GHz 64 MiB 195 W
Count: 3

Bibliography[edit]

  • Huawei. Personal Communication. 2019
  • Huawei Connect 2018. October 2018
  • HiSilicon Event. January 7, 2019
  • Huawei, Supercomputing 2018
 
codename TaiShan v110 +
core count 32 +, 48 + and 64 +
designer HiSilicon +
first launched 2019 +
full page name hisilicon/microarchitectures/taishan v110 +
instance of microarchitecture +
instruction set architecture ARMv8.2-A +
manufacturer TSMC +
microarchitecture type CPU +
name TaiShan v110 +
process 7 nm (0.007 μm, 7.0e-6 mm) +

[转帖]TaiShan v110 - Microarchitectures - HiSilicon的更多相关文章

  1. 华为TaiShan 2280 ARM 服务器

    华为TaiShan 2280 ARM 服务器 华为TaiShan 2280 ARM 服务器  https://e.huawei.com/cn/products/cloud-computing-dc/s ...

  2. nginx负载均衡基于ip_hash的session粘帖

    nginx负载均衡基于ip_hash的session粘帖 nginx可以根据客户端IP进行负载均衡,在upstream里设置ip_hash,就可以针对同一个C类地址段中的客户端选择同一个后端服务器,除 ...

  3. [转帖]网络协议封封封之Panabit配置文档

    原帖地址:http://myhat.blog.51cto.com/391263/322378

  4. [转帖]零投入用panabit享受万元流控设备——搭建篇

    原帖地址:http://net.it168.com/a2009/0505/274/000000274918.shtml 你想合理高效的管理内网流量吗?你想针对各个非法网络应用与服务进行合理限制吗?你是 ...

  5. 3d数学总结帖

    3d数学总结帖,以下是对3d学习过程中数学知识的简单总结 角度值和弧度制的互转 Deg2Rad 角度A1转弧度A2 => A2=A1*PI/180 Rad2Deg 弧度A2转换角度A1 => ...

  6. [转帖]The Lambda Calculus for Absolute Dummies (like myself)

    Monday, May 7, 2012 The Lambda Calculus for Absolute Dummies (like myself)   If there is one highly ...

  7. [转帖]FPGA开发工具汇总

    原帖:http://blog.chinaaet.com/yocan/p/5100017074 ----------------------------------------------------- ...

  8. [Android分享] 【转帖】Android ListView的A-Z字母排序和过滤搜索功能

      感谢eoe社区的分享   最近看关于Android实现ListView的功能问题,一直都是小伙伴们关心探讨的Android开发问题之一,今天看到有关ListView实现A-Z字母排序和过滤搜索功能 ...

  9. AxureRP7.0各类交互效果汇总帖(转)

    了便于大家参考,我把这段时间发布分享的所有关于AxureRP7.0的原型做了整理. 以下资源均有对应的RP源文件可以下载. 当然 ,其中有部分是需要通过完成解密游戏[攻略]才能得到下载地址或者下载密码 ...

  10. 未能加载文件或程序集“Newtonsoft.Json, Version=4.0.0.0, Culture=neutral, PublicKeyToken=30a [问题点数:40分,结帖人u010259408]

    未能加载文件或程序集“Newtonsoft.Json, Version=4.0.0.0, Culture=neutral, PublicKeyToken=30a [问题点数:40分,结帖人u01025 ...

随机推荐

  1. Karmada:让跨集群弹性伸缩FederatedHPA突破新边界

    本文分享自华为云社区<Kubernetes多集群管理Karmada,跨集群弹性伸缩FederatedHPA突破新边界!>,作者:华为云云原生团队. 根据 Flexera 最新发布的< ...

  2. 教你如何用Keras搭建分类神经网络

    摘要:本文主要通过Keras实现了一个分类学习的案例,并详细介绍了MNIST手写体识别数据集. 本文分享自华为云社区<[Python人工智能] 十七.Keras搭建分类神经网络及MNIST数字图 ...

  3. Weex原理及架构剖析

    早期H5和Hybrid方案的本质是,利用客户端App的内置浏览器(也就是webview)功能,通过开发前端的H5页面满足跨平台需求.比如PhoneGap cordova ionic -- 该方案提升开 ...

  4. Spark 覆盖写Hive分区表,只覆盖部分对应分区

    要求Spark版本2.3以上,亲测2.2无效 配置 config("spark.sql.sources.partitionOverwriteMode","dynamic& ...

  5. Unable to open debugger port (127.0.0.1:53471): java.net.SocketException "Socket closed"

    21:59 Error running 'Vipsoft': Cannot run program "/Users/jimmy/Java/apache-tomcat-9.0.14/bin/c ...

  6. 如何安装和使用 Hugging Face Unity API

    Hugging Face Unity API 提供了一个简单易用的接口,允许开发者在自己的 Unity 项目中方便地访问和使用 Hugging Face AI 模型,已集成到 Hugging Face ...

  7. Hugging News #0203: 3.3 MB 的文生图模型、RHLF 训练框架、手机上能跑的 Transformer

    每一周,我们的同事都会向社区的成员们发布一些关于 Hugging Face 相关的更新,包括我们的产品和平台更新.社区活动.学习资源和内容更新.开源库和模型更新等,我们将其称之为「Hugging Ne ...

  8. JSP 学习笔记 | 五、MVC模式和三层架构 & JSP 案例实战

    前文:JSP 学习笔记 | 四.JSP标准标签库(JSTL)个人使用指南 前文:JSP 学习笔记 | 三.EL 表达式简述 前文:JSP 学习笔记 | 二.JSP 脚本 & 案例实现 & ...

  9. 基于 eBPF 的 Serverless 多语言应用监控能力建设

    作者:竞霄 监控能力作为基础运维能力和核心稳定性措施,开发运维人员可以通过监控系统有效进行故障定位,预防潜在风险,分析长期趋势进行容量规划和性能调优,是软件开发生命周期中必不可少的一环.与此同时,Se ...

  10. vue学习笔记 十二、通过计算属性获取定义的状态数据

    系列导航 vue学习笔记 一.环境搭建 vue学习笔记 二.环境搭建+项目创建 vue学习笔记 三.文件和目录结构 vue学习笔记 四.定义组件(组件基本结构) vue学习笔记 五.创建子组件实例 v ...