I would be highly surprised if it actually worked, even for a second. Motherboards have some pretty high frequencies, and the PCB routing is intricately designed to minimize capacitance so that they can actually carry these signals.
Changing the fluid that is around the board from air (dielectric constant = 1.00059) to water (80.4) is likely to introduce a lot of capacitances that weren't designed for and would be way out of tolerance, especially for channels like CPU to RAM. The additional capacitance just wouldn't allow the signal to switch fast enough to be able to reliably transmit the data. By the way, mineral oil has a dielectric constant of 2.1, so much less capacitancy than water, and some people have had success with submersion in that.
If you would be doing this so that you can overclock everything, then the higher dielectric constant works against that by reducing the maximum frequency that the board can operate at.
The Cray computers didn't have nearly the same challenges to being submerged, since the highest fundamental frequency signal on the board was 125MHz, and modern machines potentially have ~4000MHz signals, with common RAM being just below 2000MHz, with harmonics extending to >5x the fundamentals to form the waveform accurately.
I agree with the others here that have noted that metals are slightly soluble in water (especially copper), so the water would start to become conductive immediately. Voltage differences would also cause electrolysis through the water and H2 + O2 would be produced, as well as forcing ions into aqueous solution.