The cache structure for dynamic display is similar to the ArcGIS Server cache structure. In general, Cache information for non-dynamic layer is generated by a background thread. During navigation, if the cache is not available for that area, the background thread will try to create it, which leads to CPU usage. To improve the performance, there are two options to persist the cache information in dynamic display:

 

Make sure to enable the dynamic map, then save the layers as layer files or save the maps as mxds. In this way, the existing cache information will be reused during the navigation. This option applies to ArcGIS 9.2 and ArcGIS 9.3.

           

As a fact, Layer files and Mxd files will persist the cache information as long as they are saved while in dynamic mode. In other words, the way to persist the cache is to save a MXD or layers while dynamic display is enabled. The next time when dynamic display is enabled on the same session or on a different session, dynamic display will identify the existing cache and then make use of it.

 

Keep in mind, the cache information is not based on one map scale but a range of map scales. Once an area in the map is visited (zooming/roaming), the cache information will be generated for that area (on a background thread). When you revisit the same area next time, dynamic display would know where the cache is and reuse that cache for the non-dynamic layers.

 

Practically if you know the area and map scale you will visit ahead, at one session of your mapping application, you could pre-cook caches by zoom in/ roaming to the area and then save the mxd file, be sure under dynamic mode though. Those caches will be reused, without having to use any CPU, when you perform the navigation on the same map on the consequent sessions.