It’s only two years late, but the fourth generation of Xeon Scalable processors, aka Sapphire Rapids, is hitting the ground running, with every major OEM offering new servers featuring the chips.
The 4th Gen Xeon Scalable is notable because it contains a number of specialty computing engines in addition to its x86 cores, and it also has lots of cores as well; up to 60. One of the special engines is for AI acceleration, as Intel is determined to make the CPU viable as an AI processor instead of GPUs. So not surprisingly, many of the new servers are built with AI processing in mind.
Here’s the list.
Dell Technologies introduced a total of 13 PowerEdge servers, covering enterprise data centers, large-scale public clouds, and edge locations in rack, tower, and multi-node designs. They come with Dell Smart Flow, a new feature in the Dell Smart Cooling suite to increase airflow and reduce fan power by up to 52% compared to previous generation servers.
Dell claims its PowerEdge R760 delivers up to 2.9x greater AI inferencing with the new Xeon Scalable processors, while the PowerEdge R760 also offers up to a 20% increase in VDI users and over 50% more SAP Sales & Distribution users on one server, compared to the previous generation.
In addition to updating existing lines, Dell introduction the PowerEdge HS5610 and HS5620 servers delivers optimized solutions tailored for cloud service providers managing large-scale, multi-vendor data centers. They are two-socket servers that come in both 1U and 2U form factors.
All of the servers come with updated Dell CloudIQ, which combines proactive monitoring, machine learning and predictive analytics while offering a comprehensive view of servers wherever they reside. They also come with Dell ProDeploy services for rapid deployment of the hardware and Dell iDRAC9 for simplified deployment and diagnostics.
HPE announced a major update to its Alletra line of enterprise storage products. Alletra is a rebranding of its Primera and Nimble storage arrays with a storage-as-a-service model for its 9000 and 6000 series, respectively. The Alletra 4000 is a rebranding of its Apollo 4000 data storage server line, with two new products: the Alletra 4110 and 4120.
The Alletra 4110 is a 1U, dual CPU server with up to 3TB of DDR5 memory and up to 307.2TB capacity with 20x 15.36TB SSDs. The Alletra 4120 is a hybrid NVMe/HDD higher capacity 2U chassis with up to 6TB of DDR5 memory. It supports a mix of NVMe, SAS, or SATA drives.
Both systems are aimed at large enterprises with a range of data-centric workload use cases from data lakes and archives to high-throughput and in-place analytics and AI/ML, hyperconverged infrastructure (HCI), and cache-intensive workloads like databases.
One thing about Lenovo, they do things in a big way. The company introduced 25 new ThinkSystem, ThinkAgile servers and hyperconverged solutions running the new Xeon Scalable processors. They are sold under the banner of the Lenovo Infrastructure Solutions V3 portfolio and span uses like in-memory and large transactional databases, real-time analytics, ERP, CRM, and virtualized and containerized workloads.
Some systems are updates to existing lines, while others are new, such as the ThinkSystem SR850 V3, a four-socket 2U design for high-intensity workloads. The new Lenovo ThinkAgile V3 HX, MX and VX hyperconverged infrastructure products come with Microsoft, Nutanix, and VMware software pre-installed for a HCI environment.
Some systems are cooled by the fifth generation of Lenovo’s Neptune direct water-cooling technology, which it claims can reduce power consumption by up to 40% vs. using fans. Lenovo also said select models support liquid cooling for High Bandwidth Memory (HBM), a first in the industry. HBM sits next to the CPU or GPU, and air cooling them is a challenge. But Lenovo says it will support Intel’s Xeon Max Series CPUs and Intel’s GPU Max, which are not yet shipping.
Supermicro launched more than 50 servers and storage systems across 15 families collectively known as X13. They cover the gamut of AI, HPC, cloud computing, media, enterprise, and 5G/telco/edge workloads. They range from single-socket to eight socket designs and support GPU cards along with the CPU.
Among the families:
- SuperBlade (high-performance, density-optimized blades),
- Hyper (high-performance rackmount servers),
- BigTwin (for cloud, storage, and media workloads),
- GrandTwin (an all-new architecture for single-processor performance)
- SuperEdge (single-processor nodes in a 2U, short-depth form factor for the edge)
- CloudDC (meant for cloud environments)
- Petascale Storage (all-flash NVMe storage systems)
The Chinese ODM Inspur Information, popular with hyperscale cloud providers, announced that its G7 server platform would support Xeon Scalable processors with a total of 16 different types of servers. Inspur said performance is up to 61% faster, and power consumption is up to 30% less compared with the previous generation of Intel-based products.
The servers are designed around a variety of use cases, such as general-purpose computing, critical computing, and AI. The G7 design is focused on green energy, with both cold plate and immersion cooling schemes. Inspur also claims its intelligent fault diagnosis has an accuracy rate up to 95%.
The G7 server family ranges from the high-end TS860G7 with eight sockets in a 6U space and 128 DIMM slots of DDR5 memory, to the NF5180G7 general-purpose 1U rack server, which supports immersion. Other servers come in two- and four-socket designs.
There is also the NF5266G7 high-density storage server with a three-tier storage architecture in a 2U space. Inspur says its computing performance is up to 60% faster than the previous generation.
Giga Computing, the enterprise spin-off of consumer component maker Gigabyte Computing, announced servers and server motherboards based around the 4th Gen Intel Xeon Scalable processor as well as the Intel Xeon Max Series.
Giga is targeting AI, cloud computing, advanced analytics, HPC, and networking, as well as storage applications. Giga Computing has 14 new series of servers with a total of 78 configurations for customers to choose from.
The new lines include:
- Single and dual socket M-Series motherboards
- R-series 1U and 2U rackmount servers
- G-series GPU servers with thermal and mechanical design for the hotter GPU and built for HPC, AI, and high-performance parallel processing.
- H-series high-density 2U 4-node servers
- E-series edge servers with short form factors for small settings
Copyright © 2023 IDG Communications, Inc.