How-To Geek’s review of the Lenovo Yoga Pro 9n 15 Gen 11 reported strong game results from its 32GB configuration, alongside crashes, stutters and compatibility problems. Those findings belong to the tested 18-core CPU, 5,120-core GPU model; NVIDIA lists a separate 20-core tier with a much higher memory ceiling, and Windows-on-Arm software support depends on using the right versions and packages.
What the Yoga Pro 9n review measured
Zachary Boddy’s How-To Geek review tested a Yoga Pro 9n with NVIDIA’s 18-core RTX Spark N1X processor, a 5,120-core Blackwell GPU and 32GB of unified LPDDR5X memory. Unified memory is a shared pool for the CPU and GPU. The review reported Geekbench 7 scores of 2,599 single-core, 21,581 multi-core and 115,017 GPU, plus Cinebench 2026 scores of 541 single-core and 6,543 multi-core.
In games, the review reported an average of 68 FPS in Forza Horizon 6 at 1600p, High settings, with DLSS enabled. Cyberpunk 2077 averaged 75 FPS at High settings and 46 FPS with Ray Tracing set to Medium. How-To Geek also reported crashes, compatibility problems, stutters and dropped frames during its game testing. Those results describe this review unit and its stated settings, not every Yoga Pro 9n configuration or every Windows game.
NVIDIA’s two laptop RTX Spark tiers
NVIDIA lists two laptop N1X configurations. Its specifications give both a 45–80W TDP range, but differ in CPU and GPU cores, maximum unified memory and FP4 AI-compute rating.
| Laptop N1X tier | CPU cores | Blackwell GPU cores | Maximum unified memory | FP4 AI rating | TDP |
| 18-core | 18 | 5,120 | 32GB | 827 TFLOPS | 45–80W |
| 20-core | 20 | 6,144 | 128GB | 1 petaflop | 45–80W |
These are NVIDIA’s RTX Spark laptop specifications, not benchmark scores from the Yoga review. The 32GB review unit uses the lower-memory tier; How-To Geek’s results do not establish how the 64GB or 128GB configurations perform.
US prices and what each step changes
How-To Geek’s review, published November 2, 2026, listed three US Yoga Pro 9n configurations:
| Yoga Pro 9n configuration | Reported US price | Unified memory | Storage |
| 18-core CPU, 5,120-core GPU | $3,000 | 32GB | 512GB |
| 20-core CPU, 6,144-core GPU | $4,250 | 64GB | 1TB |
| 20-core CPU, 6,144-core GPU | $6,100 | 128GB | 2TB PCIe Gen5 |
The step from $3,000 to $4,250 costs $1,250, or about 41.7% more. It doubles memory, adds two CPU cores and 1,024 GPU cores, and increases storage by 512GB. The next step costs $1,850 more. It doubles memory and storage, moving from 1TB Gen4 to 2TB Gen5, while keeping the same 20-core CPU and 6,144-core GPU tier. Those are configuration changes, not measured performance gains.
For a separate RTX Spark workload, NeoTeo’s earlier Blender analysis covers a memory-heavy render on a different laptop. It offers context on another task, rather than a direct comparison with the Yoga’s game results.
Battery life and Windows-on-Arm software
How-To Geek reported no meaningful change in game performance between plugged-in and battery use on its sample. The same review estimated about 50 minutes of battery life at full GPU load and an average of 5–6 hours for ordinary, CPU-focused use. Its experience also included the game compatibility issues and interruptions noted above.
For developers, NVIDIA documents a CUDA Toolkit 13.4.2-or-newer requirement and a CUDA-enabled PyTorch installation path for native Windows Arm64. The operating system and processor architecture matter: NVIDIA says Windows x64, Windows Arm64 and Linux Arm64 packages are not interchangeable. Its RTX Spark developer FAQ and Windows-on-Arm guide describe those supported paths.
TensorRT’s Windows-on-Arm DLA support has tighter limits: NVIDIA’s support matrix lists it as beta in TensorRT 11.4.0 and 11.4.1, restricted to a limited layer set and batch size 1, with no GPU fallback for those DLA engines. The Yoga’s RTX Spark hardware and its reviewed game results therefore do not, by themselves, establish that every development or gaming workflow will run as expected.
