Selected Clear All
Research Type - Insights Research by Market - Mobile & Telecom Category - Cloud & Data Center Service - Data Center Compute Intelligence ServiceBy Attribute
- Research Type
- Research by Market
- Category
-
- Analyst
- Geography
- Date Published
- Services
-
-
Cavium (now Marvell) is the next hope for ARM-based CPUs in the data centerJuly 17, 2018In 2017, Cavium, Qualcomm, and Applied Micro were all sampling ARM-based SoCs targeted at data center servers. With Applied Micro spinning off its server CPU business to newly founded Ampere Computing and Qualcomm demonstrating no tangible design wins after announcing commercial availability for its Centriq SoC in November 2017, hopes for success of an ARM-based CPU ecosystem moved to Cavium, and in 2017 Cavium kept the design wins and partnership announcements coming. In January, Atos announced Cavium’s ThunderX2 will power its HPC for the EU-funded Mont Blanc project. In March, Cavium announced a partnership with Microsoft and in November revealed designs for a Cavium-powered Project Olympus OCP server. In May, Gigabyte, Ingrasys, and Inventec announced and subsequently launched new servers based on ThunderX2. In June, Penguin announced its Open Compute server lineup will feature a ThunderX2 sku. Clients, please log in to view the full content.Subscribers Only
-
AMD partners with HPE and Dell EMC, turns up the heat on IntelFebruary 13, 2018In 2017, AMD changed the market dynamic for entry-level enterprise servers with its EPYC SoC, providing not just healthy competition but also opportunities for new designs utilizing the large number of memory channels and high I/O bandwidth available with EPYC. OEMs like Supermicro, Sugon, and Asus and white box server vendors like Wiwynn, Inventec, Gigabyte, and Tyan quickly backed EPYC, introducing over a dozen server models based on AMD’s SoC in 2H17. CSPs (cloud service providers) also embraced EPYC with Microsoft, Baidu, Tencent and JD.com all adopting EPYC-based servers. Baidu in particular boosted EPYC’s credibility in the market as it implemented a new 1-socket server design that utilizes EPYC’s memory channels and I/O bandwidth. Clients, please log in to view the full content.Subscribers Only
-
Moore’s law still alive in the data center, but is it enough?January 09, 2018When Gordon Moore made the observation that the number of transistors in an integrated circuit doubles every two years, known as Moore’s law, he also predicted technological limitations would eventually lead to the end of this era, with successive generations either seeing less than 2x growth in transistor count or taking longer than two years. An argument that we have already reached these limitations is currently supported by the lag between Intel’s 14 nm and 10 nm PC CPU introduction. Additionally, silicon manufacturers have questioned the investment required to maintain Moore’s law, with Samsung reporting Moore’s law affordability challenges in 2016 and TSMC’s quarterly earnings showing wafer revenue from advanced process technology dropping between 2008 and 2016. Clients, please log in to view the full content.Subscribers Only
-
Server architectures for a data-driven worldSeptember 20, 2017The nature of many applications is changing to become more data intensive as the number of data points that must be processed multiplies. In addition, many algorithms need to perform the same calculations on each data point in large data sets, introducing the opportunity for performing these calculations in parallel. The need for parallel computing became obvious with the work of the AlexNet team winning the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) and Google’s Brain Team advancing the science of artificial intelligence (AI) and machine learning (ML) and providing an open source machine learning library for neural network-based ML called TensorFlow. Other examples of applications needing parallel computation include advanced driver assistance systems (ADAS) used in self-driving cars and real-time rendering for virtual reality (VR), augmented reality (AR), climate analysis, and financial trend analysis. Clients, please log in to read the full insight.Subscribers Only
-
Co-Processors as a Service: GPUs, TPUs, and FPGAsJune 27, 2017Cloud service providers (CSPs) have begun to offer a variety of compute services to meet customers’ needs. Those needs can vary from general purpose computing, data mining, and website development to more complex requirements like artificial intelligence/machine learning (AI/ML), big data analysis, and real-time video processing. This has resulted in CSPs implementing different compute types, or instances, within their cloud infrastructures based on varying combinations of processors and co-processors, memory, storage, and networking equipment, providing customers with more options to choose the appropriate mix of resources they need for their applications. Clients, please log in to view the full content.Subscribers Only
-
On the Extreme Edge, a Double Triple: AT&T Domain 3.0 and Google Cloud 3.0June 22, 2017The “software-defined” focus of telecom means not only separation of the control plane and the data plane with a centralized view of the whole network or its major parts but also a distribution of intelligence—read compute—in stepping stones toward the extreme edge of the network. We hear “edge compute” embodied in such efforts as MEC, at first in “Mobile Edge Compute,” morphing into “Multi-access Edge Compute”.Subscribers Only
-
Outlook 2017: Artificial Intelligence and Machine Learning Transform Cloud Services and Data CentersJanuary 30, 2017Machine learning (or ML)—where typically the computer is presented with example inputs and their desired outputs in order to learn a general rule that maps inputs to outputs—and analytics—the discovery and interpretation of meaningful patterns in data—capabilities integrated with enterprise and mobile applications are set to bring more innovation, changing how enterprises and service providers will serve their customers. 3 trends have come together to make it possible for enterprises of all sizes to apply analytic techniques to business processes. Please log in to read the complete insight.Subscribers Only
-
NFV gains traction among mobile carriersMarch 23, 2015Software Defined Networking (SDN) and Network Function Virtualization (NFV) have been receiving an increasing amount of attention as innovative networking architecture solutions since the second half of 2013. These virtualization technologies achieved “buzz” status with several significant announcements at the Mobile World Congress (MWC) 2015 tradeshow in Barcelona in March of this year.Subscribers Only
