Dell’s enterprise computing playbook has diversified in the last year, with new additions like quantum computing and high-performance computing-as-a-service to its portfolio.

The company on Tuesday announced what it characterized as a major expansion of its server portfolio. The updates include multi-socket servers that fit into the deep learning and datacenter infrastructures, and also systems targeted at cloud providers.

“We’re refreshing the core portfolio with the latest and greatest technology,” said Rajesh Pohani, vice president of PowerEdge, high-performance computing, and core compute technologies at Dell.

The new lineup of PowerEdge servers is a step up from the previous generation, and includes Intel’s new fourth-gen Xeon chips codenamed Sapphire Rapids, and AMD’s Epyc chips called Genoa, both of which were launched in recent months. The new servers also support Nvidia’s new GPUs codenamed Hopper.

The overall upgrades provide a major performance boost with storage and memory technologies that include DDR5 memory and PCI Express 5.0, which are supported by Sapphire Rapids and Genoa.

Customer needs are getting more specific with applications such as AI, and high-performance systems in datacenters are trying to make sense of telemetry and other data coming in from servers and devices on the edge, Pohani said.

The AI-focused high-performance systems include the PowerEdge XE9680, which can pair eight Nvidia H100 GPUs with support for the SXM5 connector and the NVLink interconnect. The 6U server has Sapphire Rapids CPUs and 10 PCIe Gen5 slots. The 4U PowerEdge XE8640 can support up to four H100 GPUs with SXM5 and NvLink support, and is paired along with Sapphire Rapids CPUs. Both are air-cooled systems.

Another new server, the 2U PowerEdge XE9640 server supports up to four of Intel’s Max series GPUs, codenamed Ponte Vecchio. The four-way GPU system supports Sapphire Rapids chips, and can also be configured with Nvidia’s Hopper GPUs, which will come with an AI Enterprise license so customers can write AI applications based on the CUDA parallel programming framework.

Dell also upgraded the mainstream server offerings, which include the R960, R860, R760, the R660 and the R760xa. The R660 offers up to 104% faster compute performance than the previous generation R650, while the R860 delivers 184% better performance in the past generation, Pohani said.

The new chips and support for faster interconnects also played a role in boosting server performance.

“The new RAID controller that we’re introducing gets you 2x the performance of the previous generation,” Pohani said.

Pohani declined to say whether the new servers with Sapphire Rapids will support Intel’s on-demand feature, which allows turning on chip capabilities for a fee.

Intel most likely has not sorted out how it will implement on-demand through traditional server makers like Dell or HPE, said Jim McGregor, principal analyst at Tirias Research.

“For the most part, Intel is probably just going to be selling [Sapphire Rapids] with everything enabled, because on-demand would be a whole new business model for them,” McGregor said.

Dell servers will be offered through the company’s APEX (computing-as-a) service, or as bare-metal deployments either on-premises or via colocation providers.

Dell is getting more active in commercial HPC with business workloads tilting to AI and ML training, and the new systems fit into Dell’s move up the computing ladder into HPC, said Charles King, principal analyst at research firm Pund-IT.

“An interesting thing about Dell APEX in regards to HPC is how a ‘pay as you go’ model could benefit businesses considering AI and similar solutions,” King said.

The other servers introduced by Dell include the PowerEdge HS5620 and HS5610 servers, which are targeted at hyperscalars.

“The HS line is a new part of our portfolio expansion to meet the requirements of cloud service providers,” Pohani said.

The Dell PowerEdge R760 servers will be globally available next month and the PowerEdge HS5620, HS5610 are slated to be widely available in April. Other  Availability for other servers and APEX services will come later in the year.