CPU Offloading
CPU Offloading is a clever safety net built into local AI engines like Ollama and LM Studio that lets you run a model even when it is slightly too big to fit inside your graphics card's VRAM.
Instead of crashing with an Out-Of-Memory error, the software loads as many layers of the model onto your fast GPU as possible and dumps the leftover layers into your motherboard's regular system RAM, where your CPU processes them.
The upside is flexibility—you can test out a massive 70B model on a modest 16GB or 24GB rig without buying new hardware. The downside is a steep drop in tokens per second. Standard DDR4 or DDR5 system RAM has a fraction of the memory bandwidth of dedicated VRAM, so offloading even 20% of a model can cut your generation speed in half or worse.
Looking for verified hardware rigs?
Check our real-time £/GB VRAM index to find systems optimized for local model inference.