Trending Articles
Trending Topics
Trending
IBM LinuxONE is the company’s range of pre-configured servers, and they have just been updated to the latest version: IBM LinuxONE Emperor 5. They have been designed for our AI world, with built-in AI accelerators to handle these popular new workloads.
These are rack-mounted machines with a modular design. If you want the balance to towards compute power, that’s fine; likewise if you need extra AI acceleration, more memory or heaps of storage.
IBM LinuxONE servers are based around IBM’s proprietary Telum II processor and Spyre AI accelerator, and come with a choice of Linux distributions: Red Hat Enterprise Linux (RHEL), SUSE Linux Enterprise Server (SLES) and Canonical Ubuntu Server.
In effect, the LinuxONE 5 is a pre-packaged version of its z17 mainframe, which launched in June.
The main reason IBM created LinuxONE in the first place was to help businesses take advantage of open-source technology. It’s essentially everything you need in one box.
With the LinuxONE 5 in particular, IBM claims it had three major goals in mind.
“The first is the continued drive for security,” said Marcel Metron, IBM Fellow and CTO Cloud Platform IBM LinuxONE during a webinar introducing the technology.
“The second is the increased infrastructure efficiency, and the third is enhanced AI acceleration, and that’s all on top of the amazing reliability and availability built into the LinuxONE 5.”
Keeping data safe has always been important, but now it’s more key than ever. You need to train your AI models on confidential data, so how do you stop that leaking out through AI queries from employees or attackers?
One answer is taking a Confidential Computing approach. “LinuxONE 5 can run confidential containers at scale on its highly secured enclave technology,” said Metron.
“When we pair that confidential container technology with LinuxONE 5’s significant crypto bandwidth, its cutting-edge hardware security module, and its standardized quantum safe algorithms, LinuxONE 5 is designed to offer end to end pervasive encryption of workloads and data protection that is unmatched and ready for the post quantum age.”
When IBM talks about efficiency on the IBM LinuxONE 5 Emperor, it’s really talking about consolidation. Moving your existing workloads, whether they sit on-premise or in the cloud, to a single piece of hardware.
“In a recent IT economics study, we demonstrated that consolidating X86 workloads to LinuxONE 5 resulted in an 83% reduction in physical footprint, and 85% reduction in power to run that workload,” said Metron.
IBM provides a TCO and CO2e calculator if you want to see how your workloads may be affected.
It’s AI acceleration where things get interesting. Two pieces of silicon matter here: the first is IBM’s Telum II processor with its integrated AI accelerator, the second its dedicated Spyre AI accelerators. The latter are coming in Q4 2025.
The Telum II offers “up to seven and a half times of the AI throughput that we had on z16”, according to Tina Tarquinio, Chief Product Officer for IBM Z and LinuxONE.
But if AI workloads are key then you may want to wait for the Spyre’s release.
“This is [an] IBM Research purpose-built engine with up to 32 cores on the Spyre accelerator, and that will be attached via PCIe card,” said Tarquinio. “And in the initial release we can have up to 48 of these Spyre accelerator cards on the system.”
| Maximum number of engines | 208 |
| Maximum number of drawers | 4 |
| Maximum number of I/O drawers | 12 |
| Number of frames | 4 |
| Sizes | 43, 90, 136, 183 and 208 |
| MAXIMUM MEMORY PER SIZE | |
| Max43 | 16TB |
| Max90 | 32TB |
| Max136 | 48TB |
| Max183 | 64TB |
| Max208 | 64TB |
You can find more detailed configuration details via this document.
The IBM LinuxONE Emperor 5 is available to order now, but as mentioned above you will have to wait until the fourth quarter of 2025 if you want to take advantage of the Spyre accelerators.
