This setting is mostly related to how the software handles large datasets. By large datasets we mean datasets for which a single timepoint cannot easily fit in the video memory of the GPU.

Viewing data in 3D and rendering videos are both highly dependent on the GPU. We cover the topic of how arivis handles large data in the 4D viewer in this article, but in short, since most GPUs have a finite amount of video memory available and a finite ability to render 3D datasets within a given amount of time, arivis typically down-samples data to make possible to render quickly in 3D. This down-sampling can lead to a noticeable drop in in the level of detail we can render and can also lead to noticeable down-sampling artefacts. As we explained in this article about rendering HD screenshots, it is always possible to render both screenshots and videos with the highest level of details, but for large datasets this comes at the cost of high processing times.

The data resolution slider adapts both to the PC configuration and to the dataset.

The scale goes from 64MB of VRAM usage on the left to the full resolution of the dataset on the right, and each graduation mark typically represents a doubling of the memory usage.

Note that since the 3D viewer renders 8bit version of the images the VRAM requirement may be significantly smaller than the actual dataset. Also, since we can only render one timepoint in any given frame, the amount of memory required is limited by the amount of data in a single timepoint.

The colour coding reflects the hardware configuration. The green part of the scale represents the amount of VRAM available on the GPU. Sticking to the green part of the scale reduces the loading times and also leads to much faster renders.

Since most computers typically have more system memory that video memory, we can use this as temporary storage for the image data instead, up to the amount of system memory available. The loading time is usually comparatively faster because loading 1MB from the disk in the RAM is usually faster than loading that data in the VRAM, but the rendering time suffers because the GPU is now reading data from the RAM rather than the faster and closer VRAM.

But since some datasets can be larger than even the RAM available, arivis also allows creating high resolution videos even at native resolution by using a hybrid rendering approach, though this is done at the cost of rendering time. 

If we are rendering a very large dataset, it is recommended the minimum amount of data resolution required to see the detail that we need to see. Some experimentation using HD screenshots may be worthwhile to find out what those optimal settings might be. It may also be worth considering splitting the animation into several movies and using volume clipping to narrow down the rendering to specific regions of interest for the high level of detail required and then, if necessary, merging the movies into a single file using video editing software.