08-13-2006, 10:52 AM
Quote:Yeah. I think a number of fights can easily be scaled down. I.e. only 6 adds for Domo, 2 fewer sons of Rag, 30 spawns for Razorgore, Vael's BA every 20 seconds, etc. The number 40 is nothing magic.
I can see these types of changes also benefitting the processing loads on the servers and the loads on most players clients since there would be less of the N! problem to deal with.
I know one thing that was killing a lot of players in BWL was that the first rooms were right under the suppression room (with a 'low' ceiling). Since the game engine (server and client versions) does not do much to differentiate the z-axis from the x- and y-axis when doing distance checks for things like the AoEs or graphical drawing, this would lead to some heavy loads on the server or client as they keep trying to do camparisons to all the whelps up there even if they will not actually be interacted with at the time of the first two encounters. The clients failing can be helped a lot by turning down the distance at which deals with objects at a distance so at least it can effectively ignore trying to track a lot of whelps for drawing purposes (before it gets to the check that there is a floor/ceiling blocking the view).
I had wondered how they were going to handle things in Medivh's Tower. I had been in it during the closed beta and had seen that it was many levels stacked together as close as or closer than the ceiling hight to the suppression room in BWL. Now I know that they have changed the tower since beta (its noticably smaller now) but that does not mean that there is overall less to the tower. Most likely there are as many levels as there had been before but now with less running from floor to floor. This most likely tighter packing of the levels would have potentionally lead to even more lag inducing effects in that tower if any of the rooms ended up being populated even similar to how the suppression room is populated in BWL.
Now consider for those not as math knowledgeable what the reduction like fractaled might end up impling. A reduction from 40! to 30! (this is in some ways less than what is really the load being looked at by the way) is not 3/4 of the load, it is 30!/40! which works out to be about 3.25x10^-16. Basically it will amount to a fairly small fraction of the load that is at times required for the way they set up some of the current 40 man content. Yes there will likely be more of these instances being run than there are right now, but it is not likely that there will be enough being run at one single time to be able to make up that difference. This is because to make up the difference between say 25! to 30! would require 26x27x28x29x30 additional concurent runs to match a similar load problems.