In the realm of ARM64-based server CPUs, the abundance of cores can present unforeseen challenges for Linux operating systems. Ampere, a prominent player in this space, has recently launched its AmpereOne data center CPUs, boasting an impressive 192 cores. However, this surplus of computing power has led to complications in Linux support, especially in systems employing two of Ampere’s 192-core chips (totaling a whopping 384 cores) within a single server.
The Core Conundrum
According to reports from Phoronix, the ARM64 Linux kernel currently struggles to support configurations exceeding 256 cores. In response, Ampere has taken the initiative by proposing a patch aimed at elevating the Linux kernel’s core limit to 512. The proposed solution involves implementing the “CPUMASK_OFFSTACK” method, a mechanism allowing Linux to override the default 256-core limit. This approach strategically allocates free bitmaps for CPU masks from memory, enabling an expansion of the core limit without inflating the kernel image’s memory footprint.
Tackling Technicalities
Implementing the CPUMASK_OFFSTACK method is crucial, given that each core introduces an additional 8KB to the kernel image size. Ampere’s cutting-edge CPUs stand out with the highest core count in the industry, surpassing even AMD’s latest Zen 4c EPYC CPUs, which cap at 128 cores. This unprecedented core count places Ampere in uncharted territory, making it the first CPU manufacturer to grapple with the limitations of ARM64 Linux Kernel 256-core threshold.
The Impact on Data Centers
While the core limit predicament does not affect systems equipped with a single 192-core AmpereOne chip, it poses a significant challenge for data center servers housing two of these powerhouse chips in a dual-socket configuration. Notably, SMT logical cores, or threads, also exceed the 256 figure on various systems, further compounding the complexity of the issue.
AmpereOne: A Revolutionary CPU Lineup
AmpereOne represents a paradigm shift in CPU design, featuring models with core counts ranging from 136 to an astounding 192 cores. Built on the ARMv8.6+ instruction set and leveraging TSMC’s cutting-edge 5nm node, these CPUs boast dual 128b Vector Units, 2MB of L2 cache per core, a 3 GHz clock speed, an eight-channel DDR5 memory controller, 128 PCIe Gen 5 lanes, and a TDP ranging from 200 to 350W. Tailored for high-performance data center workloads that can leverage substantial core counts, AmpereOne is at the forefront of innovation in the CPU landscape.
The Road Ahead
Despite Ampere’s proactive approach in submitting the patch to address the core limit challenge, achieving 512-core support might take some time. In 2021, a similar proposal was put forth, seeking to increase the ARM64 Linux CPU core limit to 512. However, Linux maintainers rejected it due to the absence of available CPU hardware with more than 256 cores at that time. Optimistically, 512-core support may not become a reality until the release of Linux kernel 6.8 in 2024.
A Glimmer of Hope
It’s important to note that the outgoing Linux kernel already supports the CPUMASK_OFFSTACK method for augmenting CPU core count limits. The ball is now in the court of Linux maintainers to decide whether to enable this feature by default, potentially expediting the timeline for achieving the much-needed 512-core support.
In conclusion, Ampere’s 192-core CPUs have thrust the industry into uncharted territory, necessitating innovative solutions to overcome the limitations of current ARM64 Linux kernel support. As technology continues to advance, collaborations between hardware manufacturers and software developers become increasingly pivotal in ensuring seamless compatibility and optimal performance for the next generation of data center systems.
[…] list of the suspected Emerald Rapids CPUs that Intel plans to release on the server market has been revealed by hardware detective momo us. […]
[…] platform. Buildroot is used to build the standard software stack for Linux. Creating a bootloader, Linux kernel, and bootable file-system image for the RISC-V virtual platform is a pretty simple task when you […]
[…] with a myriad of compelling features. Let’s delve into the core attributes that make this AMD CPU an exceptional […]
[…] cutting-edge strategy for improving single-threaded CPU speed is Single Instruction Multiple Data […]
[…] powers the powerful performance of the Galaxy Book4 series, maximizing productivity. A speedier central processing unit (CPU), a more powerful graphics processing unit (GPU), and a recently added neural processing unit (NPU) […]
[…] end-to-end message authentication to legal form digital signatures to cryptocurrency mining’s CPU-draining […]