The 1/3 slower speed I'm seeing in RTW vs MTW isn't due to the graphics in my case. I pull the camera back so all the men are sprites. It's RTW sprites vs MTW sprites that I'm comparing, and RTW is much less efficient at it. There is no network or AI involved when watching a replay. I've put all graphics setting on minimum, and it barely makes any difference in fps from high graphic settings since I have a good video card. It suggests either the cpu or the i/o is the bottleneck. It's hard to think of a reason RTW should be less efficient than MTW under these conditions except inefficient programming. For example, sequencing through a multi-dimension data array in memory can be very efficient or very inefficient depending on the order of the array indecies used. It's very easy to make a mistake when programming and use an inefficient addressing method. Another example could be floating point calculations being done in double precision when you could get by with single precision or even using floating point variables when interger would be sufficent. I suppose there could be more parameters associated with each man in RTW that have to be updated, but I doubt is 3 times as many as in MTW.