What is RAM, and how is it different from storage?
RAM (random-access memory) is the fast, temporary memory a computer works out of. When you open a program or load a model, it goes into RAM so the processor can reach it quickly. The catch is that RAM is volatile: switch the power off and it is empty. That is the line between memory and storage. A disk keeps your files after a reboot but is far slower; RAM is fast but forgets everything. The machine keeps the lasting copy on disk and pulls the active copy into RAM to actually use it.
How does RAM relate to running models?
On a normal desktop with a discrete graphics card, the model lives in the GPU’s own memory and the operating system lives in system RAM, so the two rarely fight over space. A DGX Spark works differently: there is one pool of memory, and both the GPU and the operating system draw from it. That is how the box fits a large model in a small chassis, but it has a consequence worth remembering. The memory your browser and background services hold is memory your model cannot. Run past what is free and you OOM (hit an out-of-memory error). The fix is usually to free something else first, not to buy more.