Credit to Argonaut McFancypants PhD on the official forums.
By using server side visuospatial and linguistics pre-processing, transhumanist designers can achieve up to 4x increase in brain computing efficiency by eliminating the costly unconscious object recognition processes. Furthermore, this completely frees up the occipital, temporal, and parietal lobes of the brain which, with use of frontal temporal lobe emulation, allows us to further multiply previous brain efficiency increases by 3×. This, multiplied with the natural x5 efficiency increase in removing physical latency, leads to up to 60x increase in brain efficiency without increasing brain processing power.
Complications arise when trying to interact with the outside world from a simulspace. Data external to the simulspace cannot be pre-processed by the server. This requires re-enabling the occipital, temporal, and parietal lobes, and requires a ego to preform object recognition itself. Additionally, rerouting of mesh data in a time dilated simulspace requires synchronization that greatly slows perceived mesh access. Research shows a significant productivity drop do to frustration normally negates any productivity gains from working in a time dilated simulspace.
Studies show that subconscious processes in creativity and imagination are less effective in time dilated simulspace as time dilated subjects were no more creative or productive as non-time dilated subjects over the same real-time period. With the increased stress and alienation an ego can experience in a time dilated simulspace, it is not normally recommended for most simulspaces. Despite that, many hypercorps insist on the cost saving benefits of time-dilated simulspace and continue to use it.