Supermicro, Inc., a Total IT Solution Provider for Cloud, AI/ML, Storage, and 5G/Edge, is announcing the first in a line of powerful yet quiet and power-efficient NVIDIA-accelerated AI Development platforms which gives information professionals and developers the most powerful technology available today at their deskside. The new AI development platform, the SYS-751GE-TNRT-NV1, is an application-optimized system that excels when developing and running AI-based software. This innovative system gives developers and users a complete HPC and AI resource for department workloads. In addition, this powerful system can support a small team of users running training, inference, and analytics workloads simultaneously.

The self-contained liquid-cooling feature addresses the thermal design power needs of the four NVIDIA A100 Tensor Core GPUs and the two 4th Gen Intel Xeon Scalable CPUs to enable full performance while improving the overall system's efficiency and enabling quiet (approximately 30dB) operation in an office environment. In addition, this system is designed to accommodate high-performing CPUs and GPUs, making it ideal for AI/DL/ML and HPC applications. The system can reside in an office environment or be rack-mounted when installed in a data center environment, simplifying IT management.

In recent years, government institutions have been funding the development of home-grown hardware that will power the government infrastructure. This trend was born out of a desire to design chips with no back doors implemented so that no foreign body could monitor the government's processes. Today, Russian company Baikal Electronics managed to boot up the Baikal-S processor with 48 cores based on Arm Instruction Set Architecture (ISA). The processor codenamed BE-S1000 manages to operate 48 cores at a 2.0 GHz base frequency, with a maximum boost of 2.5 GHz clock speed. All of that is achieved at the TDP of 120 Watts, making this design very efficient.

When it comes to some server configurations, the Baikal-S processor run in up to four sockets in a server board. It offers a home-grown RISC-V processor for safe boot and management, so the entire SoC is controlled by a custom design. Baikal Electronics provided some benchmark numbers, which you can see in the slides below. They cover SPEC2006 CPU Integer, Coremark, Whetstone, 7Zip, and HPLinkpack performance. Additionally, the company claims that Baikal-S is in line with Intel Xeon Gold 6148 Skylake design and AMD EPYC 7351 CPU based on Zen1 core. Compared to Huawei's Kunpeng 920, the Baikal-S design provides 0.86x performance.

The 56th edition of the TOP500 saw the Japanese Fugaku supercomputer solidify its number one status in a list that reflects a flattening performance growth curve. Although two new systems managed to make it into the top 10, the full list recorded the smallest number of new entries since the project began in 1993.

The entry level to the list moved up to 1.32 petaflops on the High Performance Linpack (HPL) benchmark, a small increase from 1.23 petaflops recorded in the June 2020 rankings. In a similar vein, the aggregate performance of all 500 systems grew from 2.22 exaflops in June to just 2.43 exaflops on the latest list. Likewise, average concurrency per system barely increased at all, growing from 145,363 cores six months ago to 145,465 cores in the current list.

Unlocking the full potential of 5G requires transforming network infrastructure from core to edge. As the world's leading network silicon provider, Intel is at the forefront of driving this transformation. Today, the company made a sweeping set of hardware and software announcements, including the launch of the new Intel Atom P5900, a 10 nm system-on-chip (SoC) for wireless base stations, which is a critical early deployment target for 5G networks.

"As the industry makes the transition to 5G, we continue to see network infrastructure as the most significant opportunity, representing a $25 billion silicon opportunity by 2023," said Navin Shenoy, executive vice president and general manager of the Data Platforms Group at Intel. "By offering customers the fastest and most effective path to design, deliver and deploy 5G solutions across core, edge and access, we are poised to expand our leading silicon position in this growing market."

All Intel CPU microarchitectures since 2013 are vulnerable to a new class of "Zombieload," attacks, chronicled under "Zombieload v2" (CVE-2019-11135). This is the fifth kind of microarchitectural data sampling (MDS) vulnerability, besides the four already disclosed and patched against in Q2-2019. The vulnerability was kept secret by the people who discovered it, as Intel was yet to develop a mitigation against it. There is no silicon-level hardening against it, and Intel has released a firmware-level mitigation that will be distributed by motherboard manufacturers as BIOS updates, or perhaps even OS vendors. While Intel's latest enterprise and HEDT microarchitecture, "Cascade Lake" was thought to be immune to "Zombieload," it's being reported that "Zombieload v2" attacks can still compromise a "Cascade Lake" based server or HEDT that isn't patched.

"Zombieload v2" is an exploitation of the Asynchronous Abort operation of Transactional Synchronization Extensions (TSX), which occurs when malware creates read operation conflicts within the CPU. This reportedly leaks data about what else is being processed. "The main advantage of this approach is that it also works on machines with hardware fixes for Meltdown, which we verified on an i9-9900K and Xeon Gold 5218," reads the latest version of the Zombieload whitepaper that's been updated with "Zombieload v2" information. TSX is a requisite for "Zombieload v2," and all Intel microarchitectures since "Haswell" feature it. AMD processors are inherently immune to "Zombieload v2" as they lack TSX. Intel downplayed the severity or prevalence of "Zombieload v2," but dispatched microcode updates flagged "critical" nevertheless.

GIGABYTE showed off the C422-WS, a socket LGA2066 motherboard which supports both client Core i5/i7/i9 X-series and enterprise Xeon Gold and Xeon Platinum processors compatible with socket R4 (a variation of LGA2066). The board supports up to 128 GB of quad-channel memory on Core X processors, and up to 512 GB of DDR4 LRDIMM memory on Xeon chips. It is based on the Intel C422 chipset. The board draws power from 24-pin ATX, two 8-pin EPS, and 6-pin PCIe power connectors. Expansion slots include seven PCI-Express 3.0 x16, of which four are wired to the socket. Storage includes a mix of M.2, U.2, SATA-Express. Networking includes a 10 GbE interface, and a second 1 GbE, both driven by Intel-made controllers. 8-channel HD audio makes for the rest of it.

Intel is preparing a new nomenclature for its Xeon line of enterprise processors based on core counts. The Xeon Bronze-3000 series will consist of chips with less than 10 cores, Xeon Silver-4000 series with 10 to 12 cores, Xeon Gold-6000 series with 14 to 22 cores, and the top-dog Xeon Platinum-8000 series with 24 to 28 cores. The company will address a variety of enterprise workstation and server markets with these chips. All chips in this lineup will be built in the new LGA-2066 package, and the Xeon Bronze and Xeon Silver parts are confirmed to be based on the "Skylake" microarchitecture.

What is also characteristic of these chips is their vast memory controller with six memory channels, and support for three DIMMs per channel. The top-tier Xeon Platinum-8000 series 28-core chip features 1 MB of dedicated L2 cache per core, and 38.5 MB of shared L3 cache. Its TDP is rated at 208W, and it will be built on Intel's 14 nm process. The 28-core parts come with clock speeds under 2.50 GHz. The 34 Xeon Gold and Xeon Platinum series processors Intel plans to launch over the year are tabled below.

Hot on the heels of AMD Ryzen, Intel is planning to launch the pro-consumer targeted Xeon Gold line of processors. The company is reportedly "freaked out" at the cost/performance of AMD Ryzen R7-1800X in creative productivity applications, and is preparing a new line of processors targeting that niche of the market, which uses MacPro desktops and media production workstations using HEDT processors, but needs a bit of "reliability." The Xeon Gold series will be based on the 14 nm "Skylake-EP" silicon, and will feature up to 18 CPU cores. The first model in the series is the Xeon Gold 6150.

The Xeon Gold 6150 features 18 CPU cores, with HyperThreading enabling 36 logical CPUs. Intel increased the L2 cache amount per core to 1 MB, from 256 KB found on other implementations of the "Skylake" architecture. The chip features 25,344 KB (24.75 MB) of shared L3 cache. The core clock speed is 2.70 GHz, with a maximum Turbo Boost frequency of 3.70 GHz. The chip features a quad-channel DDR4 memory interface. Going by its specs, the Xeon Gold 6150 will clearly occupy a higher market segment than the Ryzen R7-1800X, but could lure orders from the likes of Apple, for its next-generation MacPro workstations.