Join us for a webinar, September 17 11am PT | 2pm ET: What's New in Fuzzball 4.2
ProductsSolutionsWhy CIQResourcesPricing
Contact

RLC Pro AI

Your GPUs run only as fast as the operating system under them

RLC Pro AI tunes the layer between your accelerators and your models. The CIQ Linux Kernel tracks the upstream long-term kernel, so new accelerators run the week you rack them. CUDA, DOCA-OFED, and PyTorch ship validated together, and the same stack runs on bare metal, in the cloud, and as an OCI container. Move a workload to wherever it runs cheapest or fastest. Enterprise Linux binary compatibility.

CIQ trusted by:

RLC Pro AI delivers more from every GPU by optimizing the layer that drives it all: the OS

Start with a validated stack

CUDA Toolkit, DOCA-OFED, and PyTorch ship pre-integrated and compatibility-tested as a unit. Drivers, kernel modules, and runtime dependencies are baked in before release. No manual assembly. No version conflicts at deployment. From install to first inference in under four minutes.

Day-one hardware support

Current NVIDIA GPU hardware is supported from day one. No backport cycles. No waiting for the OS to catch up to the infrastructure you are already running.

More output from every GPU

PyTorch flags and CUDA configurations ship pre-optimized for AI/ML workloads. More throughput from the infrastructure you already own, with no manual tuning required.

What's in RLC Pro AI

Enterprise Linux, enhanced with AI/ML workload optimizations, that deliver leading performance for your AI workloads with a golden, validated image that makes scaling fast and repeatable.

CIQ Linux Kernel (CLK)

The CIQ Linux Kernel tracks the upstream long-term kernel instead of a frozen snapshot kept alive by backports. Hardware support arrives as upstream delivers it, so new accelerators run the week you rack them, not a release cycle later.

Pre-validated stack

CUDA Toolkit and DOCA-OFED ship commercially authorized and pre-integrated. PyTorch and vLLM framework combinations are validated before release.

Deployment flexibility

One validated stack, one performance profile, across bare metal, virtual machines, and AWS, Azure, and GCP — including sovereign, on-premises infrastructure.

Ships as a container

The validated Pro AI stack ships as OCI container images. A container by itself doesn't make a GPU fast — run it on RLC Pro AI and the containerized stack gets the same tuned, validated performance as the bare-metal and VM builds. bootc brings that same stack to immutable, image-based deployment. Runs on x86_64 and aarch64.

Faster model loading, workload-tuned scheduling

CLK 6.18 tunes the memory and storage path for the large sequential reads that dominate model loading, and lets a workload-specific CPU scheduler load at runtime — no patch, no reboot — for inference engines that general-purpose scheduling starves.

Ready-to-run workloads

Ollama ships in container form on the same validated stack. Pull the image, serve a model, no CUDA environment to assemble.

Enterprise security and support

Every package is CIQ-built and cryptographically signed with an SBOM alongside each image. Secure Boot and Confidential Computing included. Backed by Standard and Premium commercial SLAs.

Enterprise Linux binary compatibility

Built on rock-solid Rocky Linux. Get the reliability of Enterprise Linux with cutting-edge performance.
Meet CIQ Enterprise Linux Manager (ELM)

Pin a known-good stack across your GPU fleet

CIQ Enterprise Linux Manager (ELM) comes with all RLC Pro subscriptions. Mirror and curate content for your GPU and CUDA nodes, pin the exact versions that work, and stage the next before it touches production. A bad driver bump rolls back in one move.

See what ELM does

CIQ trusted by:

Built for what comes next

RLC Pro AI ships the pre-built, pre-tuned stacks your AI/ML workloads require today. Deploy on bare metal, virtual machines, OCI containers, or immutable images through bootc. What's next: the Ollama container post and immutable substrates for Warewulf and Fuzzball.

Find the right edition

Rocky Linux

RLC+ NVIDIA

RLC Pro

RLC Pro Hardened

RLC Pro AI

Enterprise Linux binary compatibility
Bundled GPU drivers (NVIDIA, AMD)
CUDA Toolkit / DOCA-OFED
Long-Term Support (LTS)
FIPS 140-3 validated packages
Bug fixes
Indemnification
Enterprise support SLAsRequires purchase of Standard or Premium support
Kernel runtime protection (LKRG)
Automated STIG / CIS compliance
AI / HPC kernel optimization
Capabilities at a glance
CompatibilityBinary compatible with Enterprise Linux
Architecturesx86_64, aarch64
GPU accelerationNVIDIA CUDA Toolkit, DOCA-OFED, GPU drivers
NetworkingRDMA and InfiniBand via DOCA-OFED
KernelCIQ Linux Kernel (CLK), built from upstream kernel.org longterm
CloudAWS, Azure, GCP
FormatsISO, KVM and cloud images, OCI container images, and bootc images
Supply chainCIQ-built, cryptographically signed, SBOM shipped with each image
LicensingAnnual subscription, per-node pricing (not per-GPU)

Deployed with confidence. Backed by CIQ.

Secure supply chain

Every package is CIQ-built, cryptographically signed, and shipped with an SBOM. Packages are stored in US-based servers, transmitted via secure channels, and validated end-to-end. CVE remediations are delivered on time, to the right environments.

Support from experts

RLC Pro AI is backed by engineers with decades of experience optimizing Linux for the world's most demanding compute environments, from national laboratories to hyperscale AI infrastructure.

Indemnification

CIQ stands behind every Rocky Linux component in its repositories and cloud marketplace listings. Your legal and compliance teams get the coverage they require. CIQ takes accountability so you don't have to.
Download the RLC Pro AI Guide

Learn more in the RLC Pro AI guide

Questions

Frequently asked

The CIQ Linux Kernel (CLK), currently the upstream 6.18 long-term series. CLK tracks the upstream long-term kernel instead of a frozen snapshot kept alive by backports, so GPU hardware support arrives as upstream delivers it.

No. CLK pairs with the Rocky Linux userspace, so applications certified for Rocky Linux run unmodified. Only the kernel changes.

Yes. It ships as OCI container images on x86_64 and aarch64, with GPU access from the host through the NVIDIA Container Toolkit and CDI. A container by itself doesn't make a GPU fast — run it on RLC Pro AI and the containerized stack gets the same tuned, validated performance as the bare-metal and VM builds. bootc brings that same stack to immutable, image-based deployment.

Yes — it's one of the two situations teams arrive from. Moving GPU workloads back on-premises is a solid use case for RLC Pro AI. You get one validated stack across bare metal, virtual machines, cloud, and containers, so the workload moves without a re-platforming project. Licensing is per node, not per GPU, so what you own is predictable as the fleet grows.

Annual subscription, per node, not per GPU. Adding GPUs to a node doesn't raise the software bill.

No. The CLK kernel is not FIPS 140-3 validated. FIPS 140-3 validated cryptographic modules are available on the RLC Pro LTS line — route FIPS requirements there. Kernel runtime protection and automated STIG and CIS compliance belong to RLC Pro Hardened. Teams running mixed estates often combine editions under one commercial relationship.

It's built on the RLC Pro commercial foundation and sits alongside RLC+, RLC Pro, and RLC Pro Hardened. Long-Term Support and FIPS are RLC Pro capabilities; kernel runtime protection and automated STIG and CIS compliance belong to RLC Pro Hardened.

CUDA Toolkit, DOCA-OFED, and PyTorch are rebuilt and benchmarked against the shipped driver and kernel before release, then ship pre-integrated and cryptographically signed. The assembly work happens before the image reaches your cluster, not at deployment.

Still have questions?

More throughput. Production-ready today.

RLC Pro AI is designed for organizations where GPU utilization, deployment speed, and stack consistency are business-critical.

Available on