153 tok/s
207 tok/s
48 GB DDR5
👍 What we like
- ✓The RTX 4060 Ti offers outstanding performance in tokens/s, surpassing previous models.
- ✓The Core i9-13900K maintains a great balance between performance and energy consumption.
- ✓Kingston Fury Beast RGB combines the speed of DDR5 with low noise levels for home labs.
- ✓The complete configuration offers good value for money in 2026.
👎 What to watch
- ✕Individual component costs remain high, limiting access to some users.
- ✕Some configurations may overheat under heavy load, requiring adequate ventilation.
- ✕Technical implementation complexity can discourage beginners.
🏆 Our picks
Affiliate links · same price for youGeForce RTX 4060 Ti
Amazon · price and availability on the listing
Intel Core i9-13900K
Amazon · price and availability on the listing
Kingston Fury Beast RGB 32GB
Amazon · price and availability on the listing
See all our recommended gear in the shop.
📑 Contents ▾
In 2026, the self-hosting of large language models (LLM) has reached unprecedented levels of performance. This comparison on the RTX 4060 Ti gathers indicative orders of magnitude, without a cited source, on LLM execution speed. We also explore how the Intel Core i9-13900K processor and Kingston Fury Beast RGB memory contribute to this performance.
Why this comparison
This comparison aims to evaluate the ability of local systems to run large language models (LLM) efficiently in terms of tokens per second. Technical specifications and public tests alone do not provide a complete picture of real-world performance in home lab or self-hosted NAS scenarios. Metrics such as energy consumption, latency, and performance-to-price ratio are crucial for determining which hardware is best suited to run LLMs.
Typical configuration
Typical configuration, for illustration only: no test bench was set up on our side and the figures on this page do not come from this configuration.
- RTX 4060 Ti : 16 GB VRAM
- Intel Core i9-13900K : 24 cores, 32 threads, max speed up to 5.8 GHz
- Kingston Fury Beast RGB 32GB : DDR5-6000 MHz
- OS : Ubuntu Server LTS 2026
- Ollama version 1.9
- Price: see live price (RTX 4060 Ti, Core i9-13900K, Fury Beast RGB)
Method
No test bench was set up on our side, and the figures on this page are not tied to any cited source: they are indicative, unverified orders of magnitude that do not replace the manufacturers’ spec sheets. Check those spec sheets before buying.
Indicative results
| Product | RTX 4060 Ti | Intel Core i9-13900K | Kingston Fury Beast RGB |
|---|---|---|---|
| Tokens/s (ollama) | 153 tok/s | voir fiche | - |
| Full load consumption (W) | 165 W | 125 W (base) / up to 253 W (max turbo, see spec sheet) | - |
| Max operating temperature (°C) | see spec sheet | 100 °C (TJunction) | - |
| Memory latency (ns) | - | - | see spec sheet |
Note: indicative orders of magnitude, unverified and without a cited source.
Reading the results
The RTX 4060 Ti stands out with its performance in tokens/s, offering a remarkable efficiency. The Core i9-13900K maintains its crucial role in running LLMs, though energy consumption is higher than expected. The Kingston Fury Beast RGB memory (DDR5-6000) provides high bandwidth; its latency depends on the timings of the chosen kit (see spec sheet).
Limitations of the analysis
This comparison does not cover the impact of software optimizations, cooling quality or total cost of ownership over several years. Here again, no testing was done on our side: the values given here are indicative and unverified. These figures are not tied to any cited source.
Practical verdict
The RTX 4060 Ti remains the ideal choice for those seeking top performance in tokens/s. For a balance between performance and energy consumption, the Intel Core i9-13900K is recommended, while the Kingston Fury Beast RGB offers fast and quiet memory perfectly suited to home lab needs. Check our page /materiel-recommande/ for more personalized advice.