元宇宙挑战超高算力、超低延迟！Intel：我研究五年了

要说近期最火爆的概念，“元宇宙”(Metaverse)绝对当仁不让。

当然，当作家尼尔·斯蒂芬森 (Neal Stephenson)在1992年的科幻小说《雪崩》(Snow Crash)中提出这一概念的时候，肯定也想不到会有今天的局面。Speaking of the most popular concept in recent times, "Metaverse" definitely does my part.
Of course, when writer Neal Stephenson put forward this concept in the 1992 science fiction novel "Snow Crash," he certainly did not expect it to be where it is today.
With the passage of time, the concept of meta-universe is also different from the past. It is now described as the ultimate dream of the digital world. Ways to allow billions of users around the world to work, entertain, collaborate, and socialize.
After the World Wide Web and the mobile Internet, Meta Universe is likely to become the next major computing platform.
Since it is a computing platform, it needs strong software, hardware and computing power support, especially the demand for high computing power and low latency, which is exponentially increasing. At the same time, innovations must be made at the storage and network levels, and even the entire The basic technologies of the connected experience all need major upgrades.
At the moment, we are at the forefront of a new round of computing revolution. Movie screens, game animations, VR/AR equipment, decentralized digital finance, etc., are all changing people's lifestyles.
Metacosm challenges ultra-high computing power and ultra-low latency! Intel: I have been researching for five years. As the world's leading chip manufacturer, Intel is also actively participating in the world of Metaverse with its advanced technology and products, and has put forward its own unique concept.
Intel summarizes the technical cornerstones of Meta Universe into three layers, the highest of which is the meta intelligence layer, which can provide a unified programming model, open software development tools, and software development libraries, making it easier for developers to deploy complex Applications.
In the middle is the meta ops layer, as the computing infrastructure, focusing on providing users with available computing power beyond local computing.
At the bottom is the meta compute layer, which provides the primitive computing resources needed to realize the meta-universe experience.
Metacosm challenges ultra-high computing power and ultra-low latency! Intel: I have been studying for five years, Raja Koduri, Intel’s senior vice president and general manager of the accelerated computing system and graphics business unit, proposed in an interview to realize the virtual experience of "Avalanche" and "Player One" , It needs 1000 times the current level of computing power.
In the past five years, Intel has been working silently to prepare the computing architecture, which is also necessary to build the meta-universe, especially to provide Petaflop level (petaflops of floating point operations per second) computing power, and in real-time application scenarios The delay in is less than 10 milliseconds.
Why are you talking about meta universe at this time? Of course, Intel is not only following the industry concept, but has its own thinking.
Metacosm challenges ultra-high computing power and ultra-low latency! Intel: I have been researching for five years, Raja said, mainly considering that Intel's first high-performance graphics card used to build the cornerstone of the meta-universe technology will be released in the next few months.
If you talk about it prematurely, such as one year before the product launch, it feels a bit on paper. Intel will launch the first batch of high-performance graphics cards early in 2022, and is committed to deepening its popularity in the next 4-5 years.
The Xe GPU architecture built by Intel is gradually becoming a reality. Among them, the low-power Xe LP has been implemented, including the core display for the Core processor and the exclusive display for the entry-level consumer; the high-performance Xe HPG is coming soon, brand new Alchemist, the first product of the Ruixuan series, will arrive in the first quarter of next year, covering all desktops, notebooks, and workstations; the Xe HPC for high-performance computing is also fully advancing, and the first product Ponte Vecchio has been displayed many times. It will also be born in 2022 and will be used to realize exascale supercomputers.

时过境迁，元宇宙的概念也和以往有所不同，如今被描述成为数字世界的终极梦想，在信息丰富、实时连接、全球互联的虚拟现实(VR)和增强现实的(AR)世界中，以全新方式，让全球数十亿用户工作、娱乐、协作、社交。

万维网、移动互联网之后，元宇宙很可能会成为下一个主要计算平台。

既然是计算平台，就需要强大的软硬件和算力支撑，尤其是对高算力、低时延的需求，正在呈指数级暴增，同时在存储、网络等层面也必须进行革新，甚至整个互联体验的基础技术都需要进行重大升级。

而眼下，我们正处于新一轮计算革命的风口，电影画面、游戏动画、VR/AR设备、去中心化数字金融等等，都在改变人们的生活方式。

元宇宙挑战超高算力、超低延迟！Intel：我研究五年了

作为全球领先的芯片厂商，Intel也正凭借先进的技术、产品，积极投身元宇宙的世界，并提出了自己的独特理念。

Intel将元宇宙的技术基石概括为了三层，其中最高的是元智能层(meta intelligence layer)，可以提供统一的编程模型、开放的软件开发工具和软件开发库，方便开发者更轻松地部署复杂的应用。

中间的是元操作层(meta ops layer)，作为计算基础架构，重点向用户提供超越本地计算的可用算力。

底部的则是元计算层(meta compute layer)，为实现元宇宙体验提供所需的原始就算资源。

元宇宙挑战超高算力、超低延迟！Intel：我研究五年了

Intel高级副总裁兼加速计算系统和图形事业部总经理Raja Koduri在接受采访时就提出，要想实现《雪崩》和《头号玩家》(Player One)中天马行空的虚拟体验，就需要1000倍于现在的算力水平。

在过去五年，Intel一直在默默努力地准备计算架构，而这也是构建元宇宙所必需的，尤其是提供Petaflop级别(每秒千万亿次浮点运算)的算力，并且在实时应用场景中的延迟要低于10毫秒。

为什么在这个时候谈论元宇宙？Intel当然不仅仅是跟随业界概念，而是有着自己的思考。

元宇宙挑战超高算力、超低延迟！Intel：我研究五年了

Raja表示，主要是考虑到Intel用于构建元宇宙技术基石的首款高性能显卡，即将在未来几个月内发布。

如果过早地谈论，比如距离产品发布还有一年，多少有点纸上谈兵的感觉。Intel将在2022年早些时候推出第一批高性能显卡，并致力于在未来4-5年深入普及。

Intel全力打造的Xe GPU架构正在逐步化为现实，其中低功耗的Xe LP已经落地，包括用于酷睿处理器的核显、针对入门消费级的独显；

高性能的Xe HPG即将到来，全新的锐炫系列首款产品Alchemist(炼金术师)会在明年第一季度到来，桌面、笔记本、工作站全都覆盖；

高性能计算的Xe HPC也在全力推进，首款产品Ponte Vecchio已经多次展示，也会在2022年诞生，并将用于实现百亿亿次超级计算机。

元宇宙挑战超高算力、超低延迟！Intel：我研究五年了

Raja提出，假如出现元宇宙文明，Intel一直在幕后研究的便是通往这种文明的“道路、高速公路、火车线路”。这条道路的建设过程令人兴奋，但建成后可能鲜有人想起它的建设者。Intel的目标是提供构建元宇宙的道路建设——也就是技术基石。

在Raja看来，我们仍然处于元宇宙旅程的最初期阶段，不过一旦开启，它就会呈现指数级增长，期待在接下来的5-6年内，看到从1%到90%的发展。

Intel认为，让每个人在1毫秒的时间里获得1Petaflop算力、1PB数据，这一梦想终将成为现实。