For more than two decades, we have maintained a goal to improve the computing power delivered for each kilowatt-hour of electricity consumed for new server products as compared to equivalent, previous-generation products with a valid upgrade path.
As an important part of our comprehensive portfolio of sustainability solutions, IBM z16™ and IBM LinuxONE 4 systems continue to be developed with consideration of their environmental impacts. In 2023, we introduced the IBM z16 single frame and IBM LinuxONE Rockhopper 4, both with new rack mount options, designed to deliver strong sustainability benefits when moving workloads from x86 servers.
Like their multi-frame predecessors, the IBM z16 single frame and IBM LinuxONE Rockhopper 4 utilize the dual-chip IBM Telum® processor. IBM Telum contains two separate processor chips acting as one through a high-speed communications bus, for a performance increase. The largest IBM z16 single frame provides approximately 14% more IBM z/OS® capacity than the largest IBM z15® T02 (both with 6 configurable processors).1 However, when compared to the immediate previous generation systems, while the re-architecture of the chip and the feature size reduction increased performance, they also contributed to a decrease in compute power per kilowatt delivered of 11% in an IBM z16 single frame with 6 configurable processors and 5% in an IBM LinuxONE Rockhopper 4 with 68 configurable processors.2
Consolidating Linux workloads on an IBM z16 single frame or IBM LinuxONE Rockhopper 4 instead of running them on compared x86 servers with similar conditions and location can reduce energy consumption by 75% and space by 67%.3 This can have the associated benefit of reducing the amount of cooling required in data centers and potentially alleviating physical IT growth so that clients that are constrained for space can defer or avoid expanding or building new data centers.
IBM continues to utilize innovations in semiconductor, hard drive/storage system and networking technologies to improve server and storage system performance for each unit of power consumed by the equipment.
The company also continues to certify products to the U.S. Environmental Protection Agency's ENERGY STAR® program.
1Based on internal measurements. Results may vary by customer based on individual workload, configuration and software levels. For capacity sizing for your systems, use the IBM zPCR Capacity Planning tool and refer to the IBM Z® Large Systems Performance Reference (LSPR) website.
2System capacity based on data available at the IBM Z LSPR website. Power consumption published in the IBM 8562 Installation Manual for Physical Planning and the IBM 3932 Installation Manual for Physical Planning. Single-thread-based MIPS are used. For IBM Z, the performance is LSPR Data (Average RNI Based) (GCP-IBM Z) for the maximum number of customer general purpose cores. For IBM LinuxONE, the performance is LSPR Data (Low RNI Based) (IFL-LinuxONE) for the maximum number of customer IFL cores. All the systems are externally air cooled. Calculations use worst-case power conditions with the maximum system power configuration at the maximum utilization and for the system environment driven maximum power condition. Results may vary.
3Compared IBM Machine Type 3932 Max 68 model consisting of 2CPC drawers and an IO drawer to support network and external storage with 68 IFLs and 7 TB of memory in 1 frame versus compared 36 x86 servers (2 Skylake Xeon Gold Chips, 40 Cores) with a total of 1440 cores. IBM Machine Type 3932 Max 68 model power consumption was measured on systems and confirmed using the IBM Power estimator for the IBM Machine Type 3932 Max 68 model configuration. x86 power values were based on Feb. 2023 IDC QPI power values and reduced to 55% based on measurements of x86 servers by IBM and observed values in the field. The x86 server compared to uses approximately .6083 kWh, 55% of IDC QPI system watts value. Savings assumes the Worldwide Data Center Power Utilization Effectiveness (PUE) factor of 1.55 to calculate the additional power needed for cooling. PUE is based on Uptime Institute 2022 Global Data Center Survey. x86 system space calculations require 3 racks. Results may vary based on client-specific usage and location.