IBM has launched its new IBM z16 mainframe today. Based on its IBM Telum Processor, announced in August last year, the new system has been based on over 1,100 hours of co-creation with more than 70 of its mainframe customers and partners. The new server extends the technical leadership IBM already has for providing ‘systems of record’ to the largest enterprise organizations in the world and will be available from the very end of May.
The Figure above is ITCandor’s assessment of IBM zSystems hardware revenues and shipped BIPS (Billions of Instructions per Second) for the generations stretching back to 2003. It could be extended almost endlessly to the left, as IBM has been making mainframes for 58 years! In addition to the server hardware shown it has made significantly more revenue from linked operating systems, applications and storage systems.
IBM zSystems – used by two-thirds of the Fortune 100
IBM zSystems is already the most important server used for transaction processing worldwide by:
- 45 of the 50 top banks,
- 8 of the 10 top insurance companies,
- 7 of the top retailers,
- 8 of the to telecommunication companies and
- many central governments around the world.
IBM has continuously modernized IBM zSystems to play a pivotal role in a hybrid multi cloud infrastructure by embracing open source software (especially on the IBM LinuxONE version) and has significant advantages over large x86-based servers for running compute-intensive applications at scale.
Integrated AI processing
The IBM z16 is the first server to include on-chip AI inferencing in its main CPU, which gives it headline advantages for realtime fraud detection – a long term aim for the payment platforms of banking and insurance companies. It can process 300 billion AI inference requests per day at 1ms latency. This function will help financial companies and retailers speed up credit card purchases, loan requests and payment clearance times. It will also help medical companies process and analyze stored x-ray and MRI scanned images and help governments to identify fraudulent claims and analyse land surveys for tax purposes.
The first quantum-safe server
One day almost all current encryption will be hackable by criminals using large quantum computing… and we can’t ignore the threat today, as some criminals are already harvesting data for future exploitation. In response IBM has been developing quantum-safe cryptography for some years and has already presented two candidates (SIKE and Dilithium) to the US National Institute of Standards and Technology (NIST) third round PQR standardization process. IBM z16 servers with Crypto Express 8S cards installed provide APIs for accessing SIKE and Dilithium and Kyber quantum-safe algorithms. IBM z16 is protected by quantum-safe technology across multiple layers of firmware, making it the first server to be protected against quantum-attacks as a standard feature.
Telum’s technical leadership
Designed by IBM the Telum processor for the new mainframe has a number of important features. In particular:
- It is a 7nm processor manufactured by Samsung
- Has on-chip acceleration for realtime AI inferencing
- Has 8-processor cores with a ‘deep super-scalar out-of-order instruction pipeline’ running at 5.2GHz
- A cache and chip-interconnection infrastructure with 32MB cache per core, scalable up to a maximum of 32 Telum chips
- A dual-chip module contains 22 billion transistors and 19 miles of wire on 17 metal layers
It maintains IBM’s proprietary approach to chip design in a world in which almost all other server vendors build systems using x86 processors from Intel or AMD.
How successful will the IBM z16 be?
IBM’s server strategy has been very different from its competitors. It offloaded its x86 servers to Lenovo in 2013 in order to allow it to focus on designing its proprietary processors – Power and zSystems. It is the only manufacturer of mainframe systems, while Oracle’s Sparc chip is its only rival in the RISC market.
My Figure above shows my annual assessment of its server market share from 2003 to 2021. While this hardware line fell with its x86 offload and the popularity of RISC servers has continued to decline throughout this period, it has maintained its mainframe revenues well.
I expect it to post big increases in revenues with this mainframe refresh in 2022, as has always happened in the past. The medium term prospects are also good as it focuses on bolstering the key ‘systems of record’ applications its customers run, demonstrating its growing advantages for large scale workloads against x86 servers, helps its customers use open source and container-based applications through Red Hat OpenShift management and fully integrating the mainframe into its advanced vision of the ‘hybrid multi-cloud’ infrastructure. It’s long term prospects will be affected by how its competitors react and how successful it is in selling future Telum-based IBM LinuxONE servers to new customers.