Too long for a comment: I would say it depends on the PCR. I have personally used maximum volumes of 50ul, if I needed to prepare higher volumes, I made one mix which subsequently got distributed to more tubes with 50ul max.
The problem with big volumes is to get a fast and homogeneous heating and cooling. If the outside of the tube is faster, the reaction there starts already while this may not be the case in the middle. This can lead to uneven denaturation and amplification of your template. The later is especially problematic when you amplify short sequences. There the amplification time might already be over and the PCR program is moving on before the template is finished, which will result in less of your full-length product and the enrichment of shorter versions.
There is another problem that might occur in big reactions: Due to gradients in reactivity because of temperature gradients I think it is possible that reagents (dNTPs, Mg-ions) can form gradients which further inhibit a good yield.
Finally I tested one thing on one of our PCR machines: The PCR tube is set into the heating/cooling block. It is not completely set into the block but stick out partly, so if you fill it up to maximum capacity, the upper 75-100ul (approximately) would be outside the heating block and we have a thermal problem. The reaction mix there would only heat up due convection.