Built for How Business Buys.
Built for How Business Buys.
The Kalyron Ruby is the AI workstation you don't outgrow: a liquid-cooled 24-core Threadripper PRO on the AMD WRX90 platform, Blackwell-generation RTX 5090 graphics, PCIe 5.0 storage on both drives — and enough empty DIMM slots, PCIe x16 slots, and drive bays to absorb your next three upgrade cycles.
One look inside the ASUS ExpertCenter Pro ET900A X9 explains the "you don't outgrow it" claim — hover each marker (or read the legend) for what the expansion actually buys you.
1
sTR5 socket · AMD WRX90Home of the liquid-cooled 24-core Threadripper PRO 9965WX.
2
8 DIMM slots128 GB ECC Registered DDR5 installed — expandable all the way to 2 TB.
3
7× PCIe x16 slots (Gen5)Six x16 links plus one x8 link — room for future accelerators, capture cards, and NICs.
4
4× 3.5″ internal baysGrow bulk storage for datasets and project archives without touching the NVMe drives.
5
2000 W 80 PLUS Platinum PSUPremium-efficiency headroom for the RTX 5090 and everything you add later.
Workstation-class platform with the PCIe lanes to feed every slot — the liquid-cooled 24-core Threadripper PRO 9965WX lives here.
Ships with 128 GB of ECC Registered DDR5; grow into terabyte-class memory for giant datasets and in-memory analytics.
Six x16 links + one x8 link. Add accelerators, capture cards, or faster NICs years from now.
Bulk SATA/SAS storage for corpora and archives, alongside the PCIe 5.0 NVMe pair.
Higher-efficiency Platinum rating — quiet, cool, and ready for full-load runs.
The exact workloads this configuration was built for — no hypotheticals.
Serve mid-size open-weight models privately in 32 GB of GDDR7 — order it with the Qwen 3.5 27B package and the OpenClaw agent framework pre-installed and skip setup week.
ETL, feature engineering, and CPU-bound model work spread across a liquid-cooled 24-core Threadripper PRO — with 128 GB of ECC DDR5 keeping big frames in memory, honestly.
CFD, FEA, and scientific computing with workstation-class accuracy — ECC memory end to end, and 6 TB of PCIe 5.0 NVMe as blistering scratch space.
Blackwell-generation ray tracing and DLSS 4 on the ASUS TUF RTX 5090 deliver photoreal design and simulation visuals with speed to spare.
9th-gen NVENC encoders accelerate live production and editing pipelines while the 24-core CPU chews through exports in the background.
Engine and shader builds love 24 cores; testing on Blackwell-generation graphics with DLSS 4 keeps the play-test rig and the build box in one tower.
Scene photos are AI-generated illustrations of typical deployments, not photographs of this exact product.
Where this box refuses to be ordinary is headroom — what's installed is only the starting point:
| Dimension | Installed Today | Room to Grow |
|---|---|---|
| System Memory | 128 GB DDR5 ECC Registered | 8 DIMM slots — up to 2 TB |
| Expansion Slots | ASUS TUF RTX 5090 installed | 7× PCIe x16 (Gen5; six x16 + one x8 link) — room for up to 4 GPUs total |
| NVMe Storage | 2 TB (OS) + 4 TB (data), both PCIe 5.0 | 4× PCIe 5.0 M.2 slots total, plus 2× SlimSAS ports for industrial NVMe |
| Bulk Storage | — | 4× SATA 6Gb/s ports and 4× 2.5″/3.5″ drive bays |
| Networking | Intel dual 10Gb Ethernet (2×10 GbE) | Faster NICs via free PCIe slots |
| Front & Rear I/O | USB Type-C, Type-A, and audio within easy reach | USB4® 40Gbps with DisplayPort™ Alt Mode for high-speed external expansion |
| Power Budget | 2000 W 80 PLUS Platinum | Platform supports PSU options up to 2000 W 80 PLUS Titanium |
Platform photography of the ASUS ExpertCenter Pro ET900A X9, provided by ASUS.
The ASUS ExpertCenter Pro ET900A X9 platform ships with fleet-management software and a verified software compatibility list — this is capable hardware that's also ready for a managed IT environment.
Fleet health, utilization, and system status at a glance across every workstation on the network.
Deploy updates and policies across every workstation from one console instead of walking to each desk.
Helps prevent unauthorized access and data leaks with centralized policy control.
ASUS Control Center Express ships with the ET900A X9 platform — no separate license to buy.
Adobe® Creative Cloud (Illustrator, Photoshop, Premiere Pro, After Effects), Autodesk® 3ds Max, Autodesk® Maya.
Autodesk® Revit, Autodesk® 3ds Max, Autodesk® AutoCAD, Chaos® V-Ray.
Autodesk® AutoCAD, Autodesk® Fusion 360, Autodesk® Inventor.
ASUS partners directly with these independent software vendors to test and verify compatibility on the ET900A X9 platform.
Illustrative deployment profiles — and the Build-to-Order options that fit each one:
Order it with Qwen 3.5 27B pre-configured and OpenClaw pre-installed: a private assistant and agentic workflows running in week one, no cloud bills.
24 ECC-protected cores and PCIe 5.0 scratch space handle solver runs; TAA compliance and USA assembly clear institutional purchasing.
RTX 5090 graphics for photoreal viewports and DLSS 4 testing, NVENC for delivery, and a CPU that treats an engine build like a warm-up.
Start at 128 GB; the 8 DIMM slots take you to 2 TB when your datasets demand it — no platform swap required.
7× PCIe x16 slots mean next year's accelerator, capture card, or 25/100 GbE NIC drops straight in.
Every unit is assembled and stress-tested in the USA to your configuration — request a quote or customize from this page.
Yes — the RTX 5090's 32 GB of GDDR7 comfortably serves mid-size open-weight models; the optional Qwen 3.5 27B (or higher) package ships pre-configured for private on-prem inference, and the optional OpenClaw framework adds agentic AI out of the box.
The liquid-cooled 24-core Threadripper PRO 9965WX on the AMD WRX90 platform brings workstation-class parallelism, ECC Registered memory support, and the PCIe lanes that feed 7× PCIe x16 slots and dual PCIe 5.0 NVMe drives — desktop platforms can't.
8 DIMM slots take system memory up to 2 TB; 7× PCIe x16 slots (Gen5) accept future accelerators and NICs; and 4× 3.5″ internal bays add bulk storage — all inside the ASUS ExpertCenter Pro ET900A X9 chassis.
Ubuntu comes pre-installed — the native home of CUDA, PyTorch, vLLM, and the open-source AI stack.
It is TAA compliant and assembled in the USA, with compliance documents available for purchase review.
1-year parts and labor limited warranty backed by ABS manufacturer support. As a Build to Order system, it is custom built and stress-tested after purchase.