If you run a micro at 24MHz you cannot simply see ground noise by the help of a scope. What you see is always a mix of voltage drop between probe contacts AND 'absorption' of electric fields which hit the probe wires AND effects of magnetic fields which intersect the loop formed by probe wires. So, what you observed tells 'everything and nothing'. Measuring pure ground noise is very difficult and demands sophisticated tools (differential probes, etc.) and experience.

If you have your micro only doing internal affairs, then I guess that ground noise will be negligible at all, if you use a solid ground plane instead of a narrow copper ground track. And the last portion of noise which might develop by the work of oscillator can very easily be overcome by just inserting a simple ferrite bead where the Vcc line hits the Vcc pin of micro. This ferrite bead in combination with decoupling capacitor, which must directly sit accros supply pins of micro, forms a very effective low pass filter which nearly totally eliminates any HF ground return current running from ground pin of micro back to voltage regulator.
In mixed applications I do always use LRC filters with each digital chip, or at least one LRC filter between the 5V regulator and digital circuitry. This tremendeously helps to minimze digital ground return currents, which otherwise would contaminate analog ground somehow.

Most noise in fast digital applications comes from magnetic fields formed by digital current loops, anyway. Then, only the use of a 4-layer board (or even multilayer) will provide adequate help.

If you need noise-free operation with your 24MHz micro, then first reduce clock frequency, then use solid ground plane (4-layer board), LRC filters in power supply and put the analog circuitry at some distance to micro. Of course, in super noise-free applications the use of splitted analog and digital ground planes is a must, at least from my experience...

Forget 40pin DIP packages, if you expect very low noise. These chips have several centimeters long internal wires, which radiate a lot. Keep in mind, that these wires can never profit of the benefits of a solid ground plane!! So, do only use as tiny packages as possible. By the way, this is very probably the reason, why you noticed different performance with your setup: The much smaller PIC package shows less noise than the bigger one of 89C51, of course, that's no surprise at all...

Kai