- Lenovo ThinkCentre X Tower 1TB AI Fusion Card Enables Local Post-Training for Massive Models
- Dual RTX 5060 Ti cards provide 32GB of VRAM for extended context lengths
- Intel Core Ultra 9 CPU and 256GB memory avoid preprocessing bottlenecks
At CES 2026, Lenovo introduced the ThinkCentre X Tower, its newest workstation-style desktop computer designed for AI inference and data-intensive computing.
The device supports a single high-end Nvidia GeForce RTX 5090 with 32GB of memory or a dual RTX 5060 Ti setup.
It marks a return to an SLI-style multi-GPU approach, a setup that became viable around 2010 when controller support improved and dual-card scaling reached 60 to 80% in real applications.
Dual GPUs Return for Modern Inference Workloads
Instead of focusing on raw frame rates, the dual RTX 5060 Ti setup emphasizes memory capacity by combining 32GB of GDDR7 VRAM on both cards.
This capability allows large language models to run locally with extended context lengths that exceed what many single-card systems can support.
However, the dual RTX 5060 Ti card setup requires careful system planning, including enough PCIe slots, strong airflow, and a capable power supply.
The ThinkCentre X Tower uses an Intel Core Ultra 9 processor and supports up to 4 x 64GB DDR5 6400 UDIMM memories.
This design prevents CPU-side preprocessing and memory-intensive tasks from becoming immediate bottlenecks during inference operations.
Expansion options remain extensive, with multiple PCIe slots, up to 3 x 2TB PCIe 2280 M.2 SSD storage bays, and legacy connectivity options.
The system also includes 1x Thunderbolt 4, 1x USB-C, 6x USB-A, 2x Ethernet, 1x HDMI 2.1, 1x DisplayPort 1.4a, audio in and out, and optional PS/2 or COM ports.
In practical testing, the dual GPU setup pushes expert mixing models like Qwen3-MoE 30B to over 100,000 tokens.
It holds around 131,000 tokens reliably, while dense models like the Qwen3 32B reach up to 45,000 tokens.
FlashAttention improves responsiveness and reduces sharing during large-context inference tasks.
Optimized runtimes such as ExLlamaV3 and TabbyAPI help close performance gaps, making the system more responsive even at context lengths from 32K to 44K.
However, stability under sustained inference loads stands out as the main advantage rather than absolute leadership in performance.
Token generation performance also continues to be limited by bandwidth limitations, particularly as context sizes expand.
Beyond the GPU selection, Lenovo includes support for a 1TB AI Fusion card, which remains one of the least clearly defined elements of the system.
This component allows for local post-training and fine-tuning of models reaching up to 70 billion parameters.
Cooling for these workloads is based on a biomimetic fan design within a 34-liter chassis, supporting high airflow for dual-GPU configurations.
Lenovo also includes support for a Sensor Hub assistant that integrates cameras, microphones, radars, and environmental sensors.
The system processes data locally to adjust performance characteristics, privacy behavior, and energy efficiency in real time.
Although these adaptive systems promise efficiency gains, their real-world value will likely depend on the maturity of the software and the transparency of user control.
Security features include DTPM 2.0, ThinkShield, a Kensington security slot, an optional chassis intrusion switch, an electronic lock, and a smart cable lock.
At over 20kg, the ThinkCentre X Tower prioritizes modularity and airflow over physical convenience.
The dual RTX 5060 Ti configuration is more attractive to users who need large context windows and model flexibility.
The device starts at $1,500 and will be available in March 2026, while the Sensor Hub costs an additional $99 and will be available in June 2026.
TechRadar will cover this year’s edition extensively CESand will bring you all the important announcements as they happen. Go to our CES 2026 News page for the latest stories and our hands-on verdicts on everything from wireless TVs and foldable screens to new phones, laptops, smart home devices and the latest in artificial intelligence. You can also ask us a question about the show on our CES 2026 Live Q&A and we will do our best to answer it.
And don’t forget follow us on tiktok and WhatsApp For the latest from the CES fair!
