更新对Intel 企业与高性能计算集团总经理Barry Davis的采访，其中有两个问题比较值得看一下
NH: What architectural directions seem appropriate for exascale and how did the existing Knights roadmap not fit with that vision?
Davis: We can talk about the right architecture for exascale in general but of course, I can’t be specific [about the Aurora deal] and I’m not going to say “Knights had X and the new Knights has Y”. The needs of the next generation architecture for exascale are fairly well-documented. There’s power, cooling, space, interconnect?and that is a tremendous issue, creating a low latency technology that communicate at that scale. There also compute and storage. There’s a lot to address here.
【可以谈谈正确的大致方向，不能明说 – “原本的Knights系列有X，而新的Knights有Y”。Exascale所需的下代架构已经有详细文献记载。功耗、散热，和互联 – 如何开发能在那么大的规模内通信的低延迟技术是个大难题。还有计算和存储。有许多问题亟待解决。】
NH: There are a lot of architectural contenders right now. How does what you’re working on for the bumped-up exascale chip fit into those?
Davis: You’re talking about hardware architectures but I like to think of things in terms of workloads. So there are, like you say, FPGAs, GPUs, vector machines and things like the PEZY chips in Japan (as one example). Many of those are offload for particular codes that have to be sent over a bus for execution. That’s fine but we are a CPU company. We like to think about codes executing on our platform without the latency and offload. There are areas where that works well, but it doesn’t work for everything and CPUs do.
【你说的是硬件架构，但我更倾向于从负载类型上考虑。的确像你说的一样，有FPGA, GPU, 向量机和日本的PEZY芯片（举个例子）这样的。其中许多都是针对特定类型代码进行负载分流，代码需要通过总线传到芯片后执行。这没什么，但我们是做CPU的。我们想让代码在我们的平台上无延迟无分流就执行。上面的手段在某些领域很有效，但不像CPU那样万能。】
作为路线图调整的一部分，Intel取消了之前宣布的代号为Knights Hill的Xeon Phi加速处理器。
Intel将专注于Exascale百亿亿次计算 – 下一代超大规模计算架构。计划之一便是用专为Exascale计算设计的新架构和新平台，来替换原计划的下一代Xeon Phi（代号Knights Hill）。
Knights Hill首次宣布是在2014年，该产品线已经从最新的路线图上消失。Knights Hill原本要使用Intel 10nm制程，并支持新一代Omni-Path互联。
最后出场的只有Knights Landing，而此前曾宣布过的Knights Mill不见踪影。
Unleashing High-Performance Computing Today and Tomorrow
Written by Trish Damkroger | November 13, 2017
Exascale at SC17
Around the world, excitement is growing for exascale computing and the new era of discovery and innovation it will enable. A consortium of European nations plans to deliver two exascale systems by 20221. The US Department of Energy recently announced intentions to by 2021 deploy the first US exascale system based on Intel® architecture, demonstrating their confidence in Intel’s roadmap and exascale technology portfolio2. China hopes to field an exascale system by 20203.
Exascale supercomputers ? systems capable of a quintillion calculations per second ? are now on the horizon. So, too, are unprecedented opportunities to transform industries, solve society’s toughest challenges, and drive extraordinary scientific discoveries via the convergence of artificial intelligence (AI), big data analytics, and traditional high-performance computing (HPC) workloads. It’s a thrilling time to work alongside my colleagues at Intel to deliver this new era in computing capabilities.
To meet our customers’ requirements and new timelines for global deployments of exascale, we are modifying our portfolio investment approach and accelerating technologies on the Intel high-performance computing roadmap. Intel has demonstrated the ability to enable efficient, multi-workload high-performance computing at scale across many different applications ? and we’re excited about the technologies we’ll bring to bear to further democratize massive scale computing. One step we’re taking is to replace one of the future Intel® Xeon Phi? processors (code name Knights Hill) with a new platform and new microarchitecture specifically designed for exascale. Combined with our comprehensive HPC solutions portfolio spanning compute, storage, I/O, and software, the updated roadmap is well poised to energize the exascale revolution.
Though exascale presents a compelling future, the first exascale systems are still three to four years away. Customers seeking nearer-term HPC solutions can turn to Intel® Scalable System Framework (Intel® SSF), our complete, balanced platform architecture for overcoming common system limitations like I/O and memory and running complex workloads like deep learning, modeling, and simulation. Intel SSF can be extended over time by adding current and future Intel® processors and accelerators, so it’s a great choice for deployments of any size — from small workgroup clusters to the largest supercomputers.
Introducing Intel Select Solutions for HPC
Intel SSF has been adopted throughout the industry as the foundation for HPC solutions from many leading system vendors. I’m pleased to announce the first Intel® Select Solutions for HPC, which build on and extend the value of Intel SSF.
Intel Select Solutions, developed and delivered by Intel and our partners, help customers more easily deploy the optimized configurations needed for today’s broad array of workloads and applications. These solutions help customers adopt the right system for their needs, with verified hardware and software configurations for targeted HPC workloads that reduce evaluation time and speed deployment, helping more users realize the benefits of HPC.
The initial Intel Select Solutions for HPC will focus on simulation and modeling clusters for manufacturing enterprises and an update to the genomics analytics solution in collaboration with the world-renowned Broad Institute of MIT and Harvard.
Intel Select Solutions for Simulation and Modeling will help more companies accelerate product design with greater levels of customization and optimization.
The Intel Select Solution for Genomics Analytics delivers performance, scale, and ease of deployment, providing scripts to deploy Broad Institute Genome Analysis Toolkit (GATK) Best Practices pipelines on optimized hardware to fuel further life science research to improve medical treatments.
Intel has partnered with E4 Computer Engineering, Atipa, Boston Ltd, Colfax, Exxact, Inspur, Lenovo, Megware, Nortech, RSC, and Supermicro to deliver the first Intel Select Solutions for HPC. For more information on Intel Select Solutions for HPC, please visit Intel Select Solutions.
HPC Leadership at the TOP
Supercomputers are a key tool for discoveries and innovations that help improve society. Intel is honored to once again be the processor supplier of choice for the vast majority of the systems in the November ‘17 TOP500 list, which was announced earlier today. Intel® Xeon® and Xeon Phi processors are the foundation for a record-breaking 94.2% of TOP500 systems, while our Intel® Omni-Path Architecture is the leading 100gbps fabric.
The number of Intel® Xeon® Scalable processor-based systems in the November ’17 TOP500 list is especially exciting. Released in July, the latest Intel Xeon processor represents a tremendous improvement for HPC with new features and performance enhancements. The November ’17 TOP500 shows 17 Intel Xeon Scalable processor based systems4. These systems deliver greater than 25 PetaFLOPS of performance; the most of any new Xeon processor on the 1st Top 500 list after launch.
Come See Us at SC17
These are just a few of the topics we can’t wait to discuss with members of the HPC community at SC17 this week in Denver, Colorado. Stop by Intel Booth #1301 for demos highlighting Intel HPC technology solutions, speakerships and hands-on interactions with industry thought leaders and networking with peers from the HPC community.
If you’re not able to connect with us at SC17, please visit http://intel.com/hpc for more information on Intel’s hardware and software solutions for high-performance computing.