Intel Xeon Scalable

2024: Downfall vulnerability that allows attackers to steal user personal data

processors In families Intel released over the past eight years, vulnerability Downfall was found, allowing attackers to steal personal data users, including bank keys enciphering and violate privacy. Among the victims were individuals, companies and. data centers TAdviser Representatives told about this MTUSI on January 16, 2024.

MTUSI information security specialists tested the Downfall vulnerability, which was assigned the CVE-2022-40982 classification number, and presented a report on the consequences it leads to. As Vladimir Nikolaev, head of the information and software support sector of intellectual competitions in the field of information security MTUSI, explained, the Downfall vulnerability is a side-channel bug such as transient execution and, if properly exploited, opens up access to encryption keys, personal correspondence, user passwords and other confidential information for attackers. It is noteworthy that this data is protected using Software Guard eXtensions (SGX), Intel hardware encryption, which separates code and data in memory from software in the system.

Testing was carried out on 2 local servers equipped with Intel Skylake and Intel Ice Lake processors (Intel Xeon Scalable 3rd generation). The study confirmed the presence of a Downfall vulnerability in each case.

The Downfall vulnerability became known back in August 2023. The flaw was discovered by Google specialist Daniel Moghimi, who after studying it found out that the vulnerability could be exploited through AVX2 and AVX-512 instructions. He called the Downfall attack Gather Data Sampling (GDS), and a little later came up with a modified version of Gather Value Injection (GVI), which is a GDS technique combined with Load Value Injection (LVI). As evidence of the work of these exploitation methods, Downfall Mogimi published prototypes of ready-made exploits on his GitHub profile.

Downfall comes bundled with Intel's Skylake line processors (Skylake, Cascade Lake, Cooper Lake, Amber Lake, Kaby Lake, Coffee Lake, Whiskey Lake and Comet Lake subseries) and Tiger Lake. Downfall is also present in the Ice Lake and Rocket Lake server chips. Many of the processors mentioned above are discontinued. It follows that Downfall has existed for at least eight years, since Skylake processors were released in 2015.^[1]

To solve the problem that has arisen, there are certain software alternatives, but each of them has its own disadvantages: there is a 30% loss of performance, disruption of some applications up to their complete shutdown, - said Vladimir Nikolaev.

The corresponding patch, which Downfall can "patch," has long been ready, but, according to MTUSI representatives, negatively affects the performance of the device. At the same time, there is no vulnerability in the new Alder Lake and Raptor Lake processors for unknown reasons, although Intel has not announced its elimination, the researchers noted.

2022: Error in processor architecture that results in disclosure of sensitive user data

On August 10, 2022, it became known that a group of researchers had disclosed a method attacks for, CPU Intel which could allow to the malefactor obtaining confidential. information The "AEPIC Leak (ÆPIC Leak)" attack method is associated with an APIC controller that helps the CPU process interrupt requests from various sources to facilitate. When multiprocessing in xAPIC mode, APIC registers are accessed through the Memory Mapping I/O (MMIO) page.

Illustration: securitylab.ru

To carry out the ÆPIC Leak attack, an attacker requires privileged access (administrator or root access) to APIC MMIO. According to the researchers, ÆPIC Leak affects applications using SGX technology designed to protect data from privileged attackers.

AEPIC Leak exists due to an error in the processor architecture, which leads to the disclosure of confidential data without using any side channel. They described it as "the first CPU error capable of disclosing sensitive data."

According to scientists, it is enough to download the enclave's memory application in order to be able to leak its contents. AEPIC Leak can completely unload application memory in less than 1 second.

ÆPIC Leak (CVE-2022-21233) is described as an uninitialized memory read vulnerability, which Intel has characterized as a medium-level hazard problem associated with incorrect isolation of shared resources. The company published a recommendation and provided a list of affected products.

The following processors are affected:

• 10th Generation Intel Ice Lake;
• 3rd Generation Intel Xeon (Ice Lake SP);
• Ice Lake Xeon-SP;
• Ice Lake D;
• Gemini Lake;
• Ice Lake U, Y;
• Rocket Lake.

In addition, processors without SGX support are safe. According to experts, ÆPIC Leak applies only to Intel SGX enclaves. ÆPIC Leak requires access to the physical APIC MMIO page, which can only be achieved with high privileges. Traditional applications are secure. The researchers published the exploit code on GitHub. According to the report, the vulnerability was most likely not used in real conditions, but its exploitation may leave no traces in the log files.

Intel recommends enabling x2APIC mode in operating systems and virtual machine monitors, which disables the xAPIC MMIO page and instead provides APIC registers through model-specific registers, which fixes this issue in affected products. These actions apply only to access to Intel's xAPIC MMIO physical page^[2].

2021

Appearance in Yandex.Cloud

On August 26, 2021, Yandex.Cloud became a public cloud platform that provided users with access to solutions based on 3rd generation Intel Xeon Scalable processors (codenamed Ice Lake). This became known on August 26, 2021. Read more here.

3rd Generation Intel Xeon Scalable

On April 6, 2021, the company Intel introduced a platform for, data centers optimized for a wide range of workloads - from clouds to networks and intelligent computing peripherals. The scalable 3rd generation Intel Xeon processor family, codenamed Ice Lake, is the foundation of Intel's data center platform () using () DPCs technologies. artificial intelligence AI

3rd Generation Intel Xeon Scalable

According to the company, 3rd generation Intel Xeon Scalable processors provide performance gains - an average of 46% in common data center workloads compared to the previous generation. The processors also received a number of platform capabilities, such as a built-in security system with support for Intel SGX technology, acceleration of cryptographic computing with Intel Crypto Acceleration Technology and AI computing with Intel DL Boost Technology. These capabilities, combined with a wide range of specialized infrastructure solutions from Intel Select Solutions and Intel Market Ready Solutions, will enable users to optimize deployment in cloud, enterprise, high performance computing (HPC), and artificial intelligence, security, network, and edge computing applications .

3rd Generation Intel Xeon Scalable

With Intel's 10 nanometer process, 3rd generation Intel Xeon Scalable processors with up to 40 cores per socket deliver 2.65 times the average performance gain of a five-year-old system. The platform supports up to 6 terabytes of system memory, up to 8 DDR4-3200 channels and up to 64 PCIe Gen4 lines per processor socket.

3rd generation Intel Xeon Scalable delivers:

• Built-in AI acceleration: 3rd generation Intel Xeon Scalable processors deliver AI performance, performance, and availability, allowing users to extract more valuable information from their data. These data center processors, powered by built-in AI acceleration, extensive software optimization, and a wide range of off-the-shelf solutions, enable AI to be deployed across applications from the peripherals to the network to the cloud. Hardware and software optimizations optimize AI performance up to 74% over the previous generation.
• Built-in security: The technology Intel SGX protects sensitive code and data with the smallest surface attacks inside the system. It is now available on dual-socket platforms based on Xeon Scalable processors that support secure enclaves that can isolate and process up to 1 terabyte of code and data for common workload types.
• Built-in cryptographic acceleration: Intel Crypto Acceleration Technology provides the proper performance of a variety of cryptographic algorithms. Organizations with heavy workloads enciphering - such as online -retailers process millions of client transactions daily and can use this feature for protecting user data no significant reduction in response time or overall system performance.

In addition, near the end of 2021, developers ON will be able to optimize their applications to accelerate workloads on Xeon Scalable platforms with open cross-architectural software toolkit one, API offering freedom from the technical and economic constraints of proprietary products. The Intel oneAPI Toolkit helps you discover processor performance, AI capabilities, and encryption with advanced compilers, libraries, analysis, and debugging tools.

In addition, the Intel IoT Market Ready and Intel Select Solutions catalogs already contain more than 500 solutions that support Intel Xeon Scalable processors. Solution catalogs help accelerate the implementation of updated platforms; 80% of Intel Select Solutions will be updated by the end of 2021.

Intel Data Center Platforms are common on the market with storages data transfer and processing capabilities. The 3rd generation Intel Xeon Scalable platform technology portfolio includes Intel Optane PMEM 200 Series, SSDs Intel Optane SSD P5800X, and Intel SSD D5-P5316 NAND, as well as Intel 800 Series Ethernet Network Adapters and Intel Agilex Programmable Logic Arrays.

The 3rd generation Xeon Scalable platform is optimized for a wide range of tasks and market segments - from cloud to intelligent peripherals:

• For the cloud: 3rd generation Intel Xeon Scalable processors are designed and optimized for tough cloud workloads and support a wide range of service environments. More than 800 global cloud service providers use Intel Xeon Scalable processors, and cloud providers plan to deliver cloud services based on 3rd generation Intel Xeon Scalable processors in 2021.
• For the network: As part of the platform, Intel offers optimized N-series processors designed for a variety of network environments. 3rd generation Intel Xeon Scalable processors provide an average of 62% more performance across a wide range of deployed networks and workloads 5G than the previous generation. Working with an extensive ecosystem of more than 400 Intel Network Builders members, Intel introduces ready-to-use concepts for solutions based on the Intel Xeon Scalable N Series processor to accelerate qualifying tests and reduce deployment time for vRAN, NFVI, virtual CDN platform, and other networking solutions.
• For Intelligent Peripherals: 3rd generation Intel Xeon Scalable processors offer performance, security, and online manageability to build AI solutions, comprehensive video analytics, and consolidated workloads for intelligent edge computing. The platform provides up to 1.56 times the AI inference performance for image classification compared to the previous generation.

2020: Intel Xeon Processor Shortage

On January 20, 2020, Hewlett Packard Enterprise (HPE) sent letters to its customers warning of a shortage of Intel Xeon server processors. The company said that the shortage could continue throughout the year, so customers should take a closer look at alternative solutions.

Given the demand, we expect supplies to remain limited throughout 2020. This affects server platforms that use these processors. To minimize the impact of such limited supplies processors Cascade Lake, HPE encourages customers to consider alternative processors that are still available for sale, HPE said in a statement.

Hewlett Packard Enterprise (HPE) sent letters to its customers warning of a shortage of Intel Xeon server processors

According to The Register, the shortage problem especially affected the Xeon Silver 42xx series processors, as a result of which customers began to lack server platforms ProLiant DL360, ML350 and BL460; Apollo 4200 Synergy and 480.

Although Cascade Lake-SP remains the most powerful architecture in the Intel Xeon family by January 2020, the only alternative solution for customers who require new systems is the use of older Skylake-SP Xeon solutions, which were introduced back in 2017.

Intel still has difficulties supplying chips based on the 14nm process technology, as the company is in the process of converting its lines to 10nm technology. To cope with the shortage, the chipmaker increased capital expenditures, increased production of 14nm products, and also reduced the production of less profitable chips to order.

In addition, in 2019, Intel increased the number of crystals on a semiconductor wafer by 25%, and in 2020 another 25% growth is expected.

There has been a shortage of Intel processors since 2018. The company hoped to solve the problem by the third quarter of 2019, but the increased demand for PCs and servers only intensified it.^[3]

2019

Discontinuing Xeon Scalable Processors with Omni-Path Integration

On October 9, 2019, it became known that Intel is discontinuing the release of the first generation Xeon Scalable processors with the integration of the proprietary Omni-Path architecture - another sign that the company is moving away from developing high-speed interconnections that are designed for high-performance computing (HPC) workloads.

Intel announced its planned discontinuation in the Product Change Notice sent to customers and partners, listing eight processors from Xeon Gold 5117F to Xeon Platinum 8176F.

In its notice, the company indicates that processors with Omni-Path integration, which include the suffix "F," are discontinued due to a shift in market demand for "other Intel products." Customers and partners have the opportunity to order processors by April 24, 2020, and the latest deliveries will be made on October 9, 2020.

Other processors also discontinued are Xeon Gold 6126F, Xeon Gold 6130F, Xeon Gold 6138F, Xeon Gold 6142F, Xeon Gold 6148F, and Xeon Platinum 8160F.

Earlier, at the end of July, Intel confirmed that it had stopped development for the second generation of its Omni-Path Architecture platform, also known as the OPA 200 series, which would have a bandwidth of 200 Gbps instead of 100 Gbps for the OPA 100.

While Intel says it will continue to sell, support and maintain first-generation Omni-Path, channel partners say that in the absence of further development plans, it is hardly worth continuing to invest in a product line with Omni-Path.

Шаблон:Quote "If there is no future development plan, then there is no point in investing in Omni-Path," said one partner in July. "It's a one-trip ticket, so to speak.

At the same time, Intel said it will continue to invest in communication technologies such as Ethernet and semiconductor photonics and is evaluating high-performance Ethernet switches as a potential alternative to Omni-Path.

The Omni-Path architecture has faced intense competition from Mellanox, which has been engaged in interconnect technology for 20 years and is being bought by Nvidia for October 2019. Mellanox is already delivering a 200 Gb/s InfiniBand HDR connection solution and is preparing to release a 400 Gb/s version of^[4].

Intel Xeon Scalable announcement codenamed Cooper Lake

On August 7, 2019, it became known that the corporation Intel announced processors the Intel Xeon Scalable family (codenamed Cooper Lake), which will contain up to 56 processor cores per socket, as well as a built-in accelerator for learning tasks artificial intelligence in a standard socket processor. The processors are presented as part of the basic Intel Xeon Scalable platforms and will go on sale in the first half of 2020. The platform performance provided by Cooper Lake processors with a large number of cores will allow you to fully realize the capabilities inherent in the Intel Xeon Platinum 9200 series, which is gaining popularity among Intel's most demanding global HPC customers, including HLRN, Advania, 4Paradigm and others.

The Intel Xeon Platinum 9200 processor family, which we introduced as part of our 2nd generation Intel Xeon Scalable processor family, has been recognized among our customers who are deploying state-of-the-art technologies for high-performance computing, advanced analytics, artificial intelligence, and high-density infrastructures. Replenishing our core Intel Xeon Scalable platforms with processors with up to 56 cores will help us meet the needs of even more customers who need high processor performance and high memory bandwidth.

Intel Xeon Scalable processors (codenamed Cooper Lake) will include twice as many cores (up to 56), higher memory bandwidth, higher logic performance and AI learning tasks compared to standard Intel Xeon Platinum 8200 processors. The 56-core Cooper Lake processor will have lower power consumption than the current Intel Xeon Platinum 9200 processors. Cooper Lake will become an x86 processor that will be equipped with a built-in accelerator for AI learning tasks - including supporting the bfloat16 type, which is added to Intel Deep Learning Boost (Intel DL Boost) technology. In addition, Cooper Lake will be platform compatible with future 10nm Ice Lake processors.

2nd Generation Intel Xeon Scalable

On April 2, 2019, Intel unveiled the second generation of Xeon Scalable server processors. The flagship Xeon Platinum 9200 has 56 processing cores, a 12-channel memory controller and the highest RAM bandwidth of any Xeon CPU. The average performance of this chip is claimed to be 1.33 times that of the previous generation.

In early April 2019, Intel unveiled the flagship server processor. It received 56 cores and 1.33 times the speed increase compared to the first generation Xeon Scalable

The processors are made in the BGA5903 form factor, which implies soldering the CPU directly to the motherboard, and allow dual-processor configurations. Intel highlights several key features of the new products:

• Integration of Intel Deep Learning Boost (Intel DL Boost) technology, which is optimized to accelerate logical inference workloads in AI algorithms such as image recognition, object detection, or photo segmentation tasks in data centers, enterprise environments, or intelligent devices on the periphery. 

• Support for Intel Optane DC non-volatile memory technology (maximum capacity - 36 TB), which, according to the developers, helps to realize a threefold increase in addressable memory compared to the previous generation Intel Xeon Scalable.

• Support for Deep Learning Boost (provides 14X more bandwidth), Turbo Boost 2.0, and Speed Select

• Add a range of hardware security features to help create a more robust foundation for computing, including  third-party channel attack protection , implemented directly at the hardware level.

New products at Intel are called "three processors in one." The fact is that they can be configured in such a way that the system takes into account primarily the number of cores or clock speed.^[5]

2018: Cascade Lake AP

On November 6, 2018, Intel announced the expansion of its processor portfolio with the Cascade Lake AP (Advanced Performance) family of chips in configuration to 48 computing cores.

Cascade Lake AP is a class of Scalable series multi-core high-performance processors Intel Xeon focused on high-performance computing (high-performance computing), HPC c, applications artificial intelligence as well as infrastructure-as-a-service solutions, IaaS noted in Intel.

Developer representatives announced that Cascade Lake processors are expected in a Multi-Chip Package (MCP) that combines up to 48 processing cores per chip using the UPI (Ultra Path Interconnect) internal bus, with support for up to 12 DDR4 memory channels per processor socket. This is significantly more than the 28-core Xeon Platinum 8180 processor (-SP generation) with Skylake 6-channel DDR4 memory. servers In a dual processor socket (2P) configuration, the system will support up to 96 processing cores and up to 24 DDR4 memory channels.

Intel has not announced whether Cascade Lake processors will support Hyper-Threading technology, but if this option is implemented, each processor will process up to 96 threads, noted in. CNews

According to tests conducted at Intel, Cascade Lake processors in the Linpack test are 1.21 times faster than Intel Xeon Scalable 8180 chips and 3.4 times faster than AMD EPYC 7601 chips. In the Stream Triad test, Cascade Lake chips are 1.83 times faster than the Intel Scalable 8180 and 1.3 times faster than the AMD EPYC 7601. In the AI/Deep Learning Inference test with the Cascade Lake chip, image processing is 17 times faster than with the Intel Xeon Platinum chip.

In addition to the presented Cascade Lake processor architecture, Intel also offers a storage hierarchy stores for heavy computing tasks. It is assumed that Intel Optane DC "permanent" memory is located in the hierarchy below, DRAM followed by Intel Optane, SSD then, and 3D NAND SSD only after them winchesters , tape drives. With this architecture, it is assumed that the Intel Optane DC DIMM "permanent" memory works together with Xeon processors in two special modes. One is App Direct, where applications optimized for this type of memory can take full advantage of "persistent" memory combined with its large capacity. The second mode is Memory Mode, which allows you to use the Intel Optane DC DIMM as a regular DRAM. At the same time, the application can store the most critical data in the Intel Optane DC DIMM, and they will remain even after the system is turned off. Thus, the cost of reboot time can be databases reduced from several hours to several seconds, according to Intel.

Cascade Lake AP family processors will be available for order in the first half of 2019, according to Intel officials.^[6]

2017

Xeon Scalable Platform Overview

Around the world there is talk of a fourth industrial revolution that will change the way leading digs use ever-growing amounts of data. Humanity continuously accumulates huge amounts of information, and old technologies can no longer cope with storage or processing. Therefore, the new platform and the Intel Xeon Scalable processor family are of particular business interest.

According to IDC, the amount of information accumulated worldwide in 2017 exceeded 20 zettabytes and continues to grow. In connection with the digitalization of all spheres of the economy, analysts expect that by 2020 there will be more than 100 zettabytes of data, and by 2025 - over 160 zettabytes. In order to cope with such an influx of data, fundamentally new technologies are needed, and one of these solutions was offered by Intel to the market in mid-2017. The new Intel Xeon Scalable platform includes not only processors and a new chipset, but also network interfaces and innovative storage solutions - NVMe, VROC and VMD.

The platform maximizes the synergy of all elements of the computing system and maximizes performance to support heavy workloads such as complex analytics, machine learning, artificial intelligence, and processing huge amounts of data. According to tests conducted by leading companies, Intel Cherry showed an increase in the performance of Huawei enterprise cloud applications by 1.62 times, SAS business analytics by 2 times, IBM analytical systems by IBM times, video encoding in Ericsson by 1.5 times. Every day the number of records of the new platform continues to grow. Intel Xeon Scalable is already rated by the likes of Amazon, AT&T, Google, Microsoft, Montefiore, Technicolor and Telefonica. The products of the new series are already used in equipping data centers, for example,Selectel and RSK.

Current applications

The need for new technologies for data centers is experiencing all industries without exception. Companies need new, more powerful, faster, flexible, and secure solutions. Each task is characterized by its own needs, but today growing markets show great demand for new, high-performance technologies.

According to IDC Cloud Infrastructure Tracker, up to 2020, the cloud infrastructure market will grow by an average of 18% per year. Cloud providers are interested in minimizing server footprint and linearly increasing performance by simply adding new computing nodes. Large dynamics is also observed in the big data processing segment. According to the IDC forecast, by 2020 companies will already spend $210 billion to cope with the influx of information.

The new processors and Intel platform are already demonstrating the benefits of fifth-generation networking. And the demand for powerful platforms will only increase, as SDNNFV the market/continues to grow and, according to analysts, Technology Business Research will exceed $150 billion by 2020. The increase also occurs in the segment of distributed data warehouses. According to Gartner, hyperconverged infrastructures will occupy 24% of the entire integrated solutions market by 2019. But in order for such systems to work efficiently, it is necessary to use high-performance network interfaces and minimize delays in data transmission.

New technologies

The Intel Xeon Scalable processor family is a new generation of chips based on the Intel SkyLake microarchitecture. Processor innovations such as Intel AVX-512, Intel Omni-Path, and a fundamentally new crystal topology have achieved 1.6 times the performance of the previous ^generation.1 The line is divided into different levels - from platinum, which includes chips with 28 computing cores to bronze, designed for the mass market. The special arrangement of components on a chip allows you to quickly receive data from RAM and cache belonging to other cores.

Intel Xeon Scalable Processor Structure

Given the need to create powerful multiprocessor systems, the developers have invested in the new platform a wide range of possibilities for interaction between several chips. The presence of a special Intel Omni-Path interconnect creates a level playing field for all computing cores when accessing any memory location up to a DDR4-2666 that can operate in 6-channel mode. The Mesh architecture opens up new possibilities for data transfer between processors in one system. Thus, the new Intel platform improves the efficiency of working with large and heterogeneous data sets for a variety of tasks.

Intel Mesh Architecture Diagram - Right

One of the most important add-ons is the Intel Optane Solid-State Drive. They create a new level of storage, intermediate between DRAM and high-capacity drives. Low latency and fast access speeds allow you to achieve significantly better server system performance. And when you optimize your workloads with special Intel software, you'll see more than one increase in critical storage options. So, installing 6 Intel Optane Solid-State Drives while optimizing Intel SPDK along with the Intel Xeon Platinum processor allows you to achieve IOPS growth up to 5.2 times, reduce latency by up to 3.3 times, and reduce the average storage response time by up to 40 times.

Intel Xeon Scalable Memory Subsystem

When building cloud systems, the integrated Intel Ethernet controller with high speed and built-in support for traffic protection and data encryption deserves special attention. Thanks to its presence, computing modules gain access to network resources with minimal delays and with a guarantee of data privacy. Meanwhile, Intel Ethernet Controller's 4-channel 10 Gb/s bandwidth provides benefits for cloud and distributed storage.

Prospects

By introducing its most powerful and versatile data center platform, Intel has demonstrated a system approach and a comprehensive vision of today's data center infrastructure that is relevant to different market segments. As the number of virtualized platforms grows, there is an increasing demand for powerful and flexible systems that can support a large number of processes simultaneously and redistribute the load in real time. With up to 28 high-performance cores on board, the new Intel Xeon Scalable platform is capable of supporting 4.2 times more virtual ^machines2 and 5 times more transactions per ^second3 than four-year-old systems. In a storage configuration, it can perform up to 5 times more I/O operations with a 70% reduction in ^signal latence4, enabling faster data storage and retrieval for analysis. In addition, the Intel Xeon Scalable platform can serve as the basis for the development of 5G networks, demonstrating an increase in the performance of key network applications up to 2.5 ^{times. 5}

The Intel Xeon Scalable platform is characterized by a significant increase in performance over the previous generation of processors. At the same time, all processors in the line are socket-level compatible, which simplifies the process of equipping, upgrading and maintaining racks. In addition, the platform offers new opportunities for organization storage systems through the use of Intel Optane drives SSD and specialized software. These features, as well as an advanced partner ecosystem, make the Intel Xeon Scalable processor platform and family a good choice for creating a demanding data centers future.

^{1The result} is achieved 1.65 times when comparing the normalized performance of Intel® Xeon® E5-26xx v4 compared to Intel® Xeon® Scalable (using OLTP Brokerage, SAP SD tests 2-Tier, HammerDB, Server-side Java, SPEC * int_rate_base2006, SPEC * fp_rate_base2006, Server Virtualization, STREAM_PAD, Black-Blade, PDLD

^{2,3,4,5C corresponds} to 1,2,3,4 here: https://newsroom.intel.ru/editorials/intel-xeon-scalable-processor-family-data-center/

Xeon Scalable Platform Announcement

On July 11, 2017, the company Intel introduced what it claims to be the most powerful server processors on the market. Xeon Scalable solutions strengthen Intel's position in the market of chips for cloud data centers services, notes. Reuters^[7]

Naveen Rao, vice president of Intel Artificial Intelligence Products Group, told the agency that the presented processors received closer support for next-generation computing applications such as artificial intelligence and self-driving cars.

Intel Xeon Scalable Processor

Compared to the previous generation, the Intel Xeon Scalable platform provides a 1.65-fold increase in system performance, and 2.2-fold increase in artificial intelligence applications. Significant benefits include Intel Advanced Vector Extensions 512 (Intel AVX-512), which accelerates resource-intensive tasks, new Intel Mesh architecture that reduces system latency, Intel QuickAssist hardware-based cryptography acceleration and data compression, and integrated high-speed Intel Omni Path Architecture for cost-effective cluster deployment.

Xeon Scalable processors are divided into five lines depending on performance: Xeon Platinum 8100, Xeon Gold 6100, Xeon Gold 5100, Xeon Silver 4100 and Xeon Bronze 3100. The most powerful chips can have up to 28 processing cores with support for multiprocessor systems with more than eight sockets. All products except the Intel Xeon Bronze 3100 support Intel Turbo Boost Technology 2.0 and Intel Hyper-Threading Technology.

Intel says the Intel Xeon Scalable platform is designed to help companies accelerate the deployment of cloud and transformation of communication infrastructure and create the conditions for using artificial intelligence technologies.

At the same time, Intel introduced a series of optimized Intel Select Solutions configurations to accelerate the assessment and deployment of data center and communications network infrastructure. The solution will provide the market with a selection of proven Intel configurations to enable customers to accelerate their return on investment in the Intel Xeon Scalable-based infrastructure for priority user workloads.^[8]

212

Notes