MC-WIKI.net
Nvidium Wiki › Guides › Nvidium Memory Limits and Chunk Culling Explained

Nvidium Memory Limits and Chunk Culling Explained

guide

Nvidium manages GPU memory through a configurable ceiling and responds to that ceiling by dropping chunk geometry from view. Two related settings control this behaviour: Automatic memory limit and the maximum GPU memory allowance.

Automatic memory limit

Automatic memory limit inspects how much video memory your graphics hardware has available and picks a memory ceiling for you based on that figure. You do not type a number yourself when this is active — the mod derives one from your hardware's capacity.

One practical detail matters here: the maximum memory field cannot be adjusted while the automatic option is in charge. If you want to set your own ceiling, you must leave the settings screen and open it again before the maximum memory value becomes editable. Plan for that extra step rather than expecting the field to respond immediately.

Max GPU memory and culling

The maximum GPU memory value is the hard ceiling on how much video memory Nvidium is permitted to consume. When usage reaches that ceiling, the mod begins culling chunks — that is, it stops drawing some chunk geometry to bring consumption back under the limit. This is the mechanism that keeps the mod from overrunning your card's memory: rather than letting usage climb unchecked, it sacrifices visible chunks.

The trade-off is straightforward. A lower ceiling means culling starts sooner and you see less terrain at once. A higher ceiling delays culling but asks more of your hardware. If you notice chunks disappearing as you move or turn, the limit is being hit and culling is doing its job.

Several other options affect memory use and visual results, and they interact with the limits above.

  • Region Keep Distance sets how far loaded regions are retained. Its own tooltip warns that it can cause problems and suggests relying on a separate mod such as Bobby, leaving this behaviour to Vanilla instead. Treat it cautiously.
  • Enables temporal coherence removes visual artifacting that appears when you rotate the camera.
  • Translucency Sorting controls how transparent surfaces are ordered, with each level trading performance against visual accuracy. None skips sorting entirely — no performance cost, but results can look poor. Sections sorts at section level, matching Sodium's quality with minimal cost. Quads sorts incrementally across several frames, which is correct but can look odd while the sort completes. Sodium defers to Sodium's own translucency handling.
  • Use sodium vertex format switches to Sodium's vertex layout. It costs performance and consumes more VRAM, but can resolve incompatibilities with certain other mods.
  • Cull degenerate triangles discards zero-area triangles to improve performance.
  • Use NV fragment shader barycentric improves performance when switched on.
  • Statistics logging level controls how much detail is recorded about how many cull layers are visible.
  • Asynchronous BFS chunk section loading greatly reduces frame time while you move, and the benefit grows with higher render distances.
  • A separate toggle decides whether fog is drawn at all.

Operating these options

All of the above are changed through Nvidium's settings interface, reached via the corresponding hotkey — assign it in the controls menu if nothing is bound. Changes to the maximum memory value require closing and reopening the settings screen, as noted. If chunks vanish while you play, raise the maximum GPU memory or disable Automatic memory limit and set a figure manually, then reopen the screen to confirm the value took effect.