Any evidence that 1 gig is a good maximum cache size

Another study for complex environments, for which I did some subjective tests, seemed to suggest that the cut-off time (i.e. the completed picture) doesn't change much. Regardless of what is in the cache or what its size is, you still have to wait for the cut-off time to get a complete view.

HOWEVER, the larger the cache, the more of the scene appears to be rezzed quickly. There are always a few objects and textures that are missed due to dropped packets.

As I say, this becomes rather more subjective because I'm going on my personal impression of how much of the scene has been rezzed over 10-second periods.

Suffice to say, there appears to be an inverse relationship between the size of the cache and the time it takes to rezz half of the view. It would suggest that doubling the cache reduces half-life of the rezzing required and that 1GB is not the optimum setting. Now I'm stuck because there's no way I can test beyond the client's 1GB setting, but the improvement as the cache size is doubled becomes relatively neglible.

The optimum size of the cache is lower than 500MB in sandboxes, where many changes can occur between visits. It still takes around a minute to fetch the whole scene! I conclude that for residents who:

EITHER...

(a) mostly go to a few fairly static but complex environments frequently,
(b) have a consistently lower than 1Mbps link,

OR

(c) are charged for data throughput by their ISP;

the largest (1GB) cache is best. As for the argument for a larger cache, the cacheing algorithm and disk speed will significantly affect speed for a cache several times that size and as I've previously noted, the improvement becomes insignificant around this point.

For the most of us, I CANNOT envisage a larger cache making ANY difference to the speed that a scene is completely rendered. The time taken entirely depends on the scene being rezzed and a 500MB or 1GB cache provides adequate versitility.

Move out of the sticks!
jk.