On-Prem

HPC

HPE goes Cray for Intel's Sapphire Rapids Xeons in latest supers

Of course the IT giant's Epyc 4 systems will still ship first


For the first time in years, Intel's CPUs will be at the heart of Hewlett Packard Enterprise's Cray supercomputing platforms, following the launch of new systems based on the chipmaker's Sapphire Rapids Xeon Scalable processors in early 2023.

Going back to HPE's acquisition of Cray in 2019, the company has favored AMD's Epyc CPUs in its EX series supercomputers. However, with the launch of the company's EX2500 platform, HPE will offer compute blades based on both Intel's 4th-Gen Xeon Scalable and AMD's Epyc 4 Genoa CPUs.

According to HPE, the EX2500 has similar architecture to the EX4000 used in Oak Ridge National Laboratory's 1.1 Exaflop Frontier supercomputer, but occupies 24 percent less space than its predecessor. The system consists of three chassis, each with 24 blades for a total of 140kW of power density, HPE's VP and GM of HPC Gerald Kleyn told The Register.

The EX2500 chassis equipped with Intel 4th-gen Xeon Scalable processors are slated to power the Crossroads supercomputer at the Department of Energy's Los Alamos National Laboratory. The system will eventually be used by the National Nuclear Security Administration to ensure the reliability and safe storage of the US strategic stockpile using high-resolution simulations of nuclear warheads.

Intel's Sapphire Rapids CPUs are attractive to the high-performance compute (HPC) community for a number of reasons including the availability of on-chip high-bandwidth memory (HBM) on certain SKUs, said Kleyn, who confirmed that the chips will be the first Intel CPUs deployed in a Cray system since HPE bought the company.

"We also have some customers that are optimized for Intel libraries."

But Cray isn't the only vendor with its eye on Intel's Sapphire Rapids CPUs. For its Hopper GPU generation of DGX servers, Nvidia ditched AMD's Epyc for Intel's 4th-Gen Xeons.

Cray absorbs HPE Apollo

The announcement sees Cray pull HPE's Intel-based Apollo HPC platform under its umbrella. The Apollo 2000 and 6500 Gen 10 systems now become the XD2000 and XD6500. Both get an upgrade to Sapphire Rapids, but the 2U XD2000 will also be available with AMD's Epyc 4, and the larger XD6500 will get Nvidia's H100 GPUs.

While the EX2500 relies on direct liquid cooling to keep each node at a a safe operating temperature, HPE's XD series can be ordered in both liquid and air-cooled configurations. HPE says this allows the systems to be deployed more easily in traditional enterprise datacenter environments that may not be equipped to support liquid-cooled infrastructure.

All of the systems announced today can make use of HPE's HPC-focused hardware and software suite, including its Slingshot interconnects, Cray Clusterstor E1000 storage arrays, Cray programming environment, and liquid-cooling infrastructure, it said.

While Intel may be back on the menu, the first new systems from HPE's Cray division will be running on AMD's Epyc 4, starting with the XD2000 later this month. Customers itching for a crack at Intel's 4th-Gen Xeon Scalables will have to wait until Intel gets around to actually launching them in Q1 2023 — more than a year and a half after they were supposed to hit the market.

And it's a similar story for HPE's enterprise lineup. Earlier this month, HPE admitted that its first x86-based ProLiant Gen11 systems would ship with AMD CPUs starting in December, months before its Intel-based systems would be ready. ®

Send us news
Post a comment

AMD slaps together a silicon sandwich with MI300-series APUs, GPUs to challenge Nvidia’s AI empire

Chips boast 1.3x lead in AI, 1.8x in HPC over Nv's H100

HPE says impact of AI on enterprise not 'overstated.' It must be hoping so

Company counting on widespread business adoption to counter server declines

Intel shows off backside power and stacked transistors at IEDM

Chip giant claims demo tech could 'significantly' improve device density

HPE targets enterprises with Nvidia-powered platform for tuning AI

'We feel like enterprises are either going to become AI powered, or they're going to become obsolete'

Intel scores a reprieve in $2.18B VLSI patent case after court orders retrial

The never-ending IP story goes on

AMD thinks it can solve the power/heat problem with chiplets and code

CTO Mark Papermaster lays out the plan for the next two years

HPE to start pumping AI capabilities into Greenlake under Project Ethan

OpsRamp now natively available through IT-as-a-service platform

Server sales down 31% at HPE as enterprises hack spending

Customers still 'digesting' shipments bought in 2022, says exec in mixed fiscal year for hardware biz

DoE watchdog warns of poor maintenance at home of Frontier exascale system

Report says new QA plan currently being worked up

China's Loongson debuts processor that 'matches Intel silicon circa 2020'

Best not to dismiss it, as Asus looks to be onboard and advances are promised

Intel drops the deets on UK's Dawn AI supercomputer

Phase one packs 512 Xeons, 1,024 Ponte Vecchio GPUs. Phase two: 10x that

German budget woes threaten chip fab funding for Intel and TSMC

Constitutional court tells govt: Er, about that €60B you handed out... it's not legal