Am I understanding correctly? You say that you ran out of sufficient processor bandwidth to untilize one additional address bit?

What else is the uC having to manage? If not very much, you may need to look at streamlining your code.
An 8MHz '51 should have no trouble addressing any channel (or even groups of channels) within a full DMX universe give that the processing and output requirements are reasonable.

The reason that 8MHz and 16MHz clocks work is that UART is running in mode 2 for DMX reception. If you look at your datasheets, you'll see that the UART baud rate in that mode is fixed at either (osc/32) or (osc/64) depending on the state of SMOD. You CAN however, get a 250KHz baud rate from a 32MHz clock on some of the Hyper-51 variants because of the inclusion of a prescaler.

Joe.