Introducing CLK: the enterprise kernel built for fast-moving hardware

New hardware doesn't wait for traditional enterprise kernel update schedules. The new server that you bought to get the best performance for your applications doesn’t always have support in traditional kernels. The newest security and performance features are developed in the upstream kernel long before being backported to older kernels. CIQ launched CLK, the CIQ Linux Kernel, to close those gaps. CLK tracks the upstream kernel.org long-term releases, delivering current hardware support and the performance improvements that come with it.

For AI inference clusters, HPC environments, and GPU-accelerated pipelines, the time between "new hardware ships" and "your distribution fully supports it" means lost performance, manual driver work, or unsupported configurations.

When hardware outpaces your kernel, production pays for it

When a new piece of hardware (network cards, GPUs, storage adapters) ships, upstream kernel.org gets driver support first. Memory allocation, connection protocols, the kernel interfaces that determine how efficiently accelerators run. All of it lands upstream before it reaches an enterprise distribution.

Traditional enterprise kernels usually get new hardware support eventually. The question is when, and whether "when" falls inside your supported lifecycle or if it supports the feature you want. For many hardware generations, the answer is: it doesn't.

For teams running AI, HPC, or any workload where hardware recency matters, that lag means one of three things: wait for vendor backports, rebuild driver stacks manually, or run in an unsupported configuration.

CLK tracks the latest kernel.org longterm releases directly, which means hardware support arrives with the hardware, not months after it. The same applies to scheduler improvements like EEVDF (Earliest Eligible Virtual Deadline First) that improve throughput and reduce latency, networking stack advances like TCP BBR congestion control, and emerging compute capabilities like io_uring, eBPF, CXL memory tiering, and confidential computing extensions (Intel TDX, AMD SEV-SNP).

Running RLC Pro or RLC Pro Hardened? CIQ closes the exposure window there too. RLC Pro Hardened ships accelerated CVE remediation and proactive kernel-level threat detection for organizations that need enterprise Linux security. RLC Pro Hardened →

Upstream LT coverage

Upstream kernel.org longterm kernels and CIQ's CLK are not the same thing. CIQ applies kernel configuration optimizations for throughput, enables secure boot for dynamic lockdown, and carries the ability to create targeted variants for specific workload profiles. Because CLK tracks a single upstream LT branch, CIQ engineers can validate each update against the Enterprise Linux userspace with a cleaner dependency chain than multi-year backport trees typically allow.

Because CLK is paired with the Rocky Linux userspace, most organizations can adopt it without requalifying their application stack. Only the kernel is different. Applications certified for the Rocky Linux userspace run unmodified — the consistent userspace ABI is preserved across kernel versions.

Newer LT kernels also carry efficiency improvements that older enterprise kernels miss. Adaptive network polling, introduced in kernel 6.13, reduces CPU energy consumption by up to 30% in high-throughput network environments. For organizations operating GPU clusters at scale, that reduction translates directly to lower power bills and cooler racks.

CLK is designed for environments where hardware recency is the primary requirement. It is available today in RLC Pro AI, with broader availability planned across the RLC Pro family.

Running RLC Pro AI today? You're already running CLK 6.12. See what's in the RLC Pro AI stack →

CLK is built for workloads where the hardware support and performance are the top concern — GPU-accelerated AI and HPC environments in particular, where teams deploying new accelerator generations need driver support the day hardware ships, and not months later.

Today, CLK ships as the standard kernel for RLC Pro AI. Teams running both compliance-driven and AI workloads can run both products in parallel on separate deployments.

Designed for	Modern hardware support, AI/HPC workloads
LTS pinning	kernel.org LT branch pinning points
Hardware support cadence	Upstream longterm (current)
Available in	RLC Pro AI (broader RLC Pro availability planned)

A current kernel paired with a stack that's been validated to work together

CLK ships as the foundation of RLC Pro AI, which validates the complete AI software stack on top of it.

RLC Pro AI ships with and validates popular ML/AI framework combinations before release and as a turnkey solution. This means, for example, that CUDA toolkit versions are matched to driver requirements. Updates are tested against each other before they ship.

The kernel keeps pace with the hardware. The validated userspace keeps the frameworks from breaking each other.

Get started with RLC Pro AI

RLC Pro AI launches in the coming weeks through CIQ, with availability on AWS, Microsoft Azure, and Google Cloud. It is part of the Rocky Linux from CIQ Pro product family alongside RLC+, RLC Pro, and RLC Pro Hardened.

For deeper background on kernel security and the limitations of the frozen vendor kernel model, CIQ has published a white paper: Vendor kernels, bugs, and stability and a technical analysis of why frozen kernels aren't the safest choice for security.

Request a demo →