The voltage thing is easy: don't do it. Remember power = Voltage**2 / Resistance, so the difference won't be 1.9/1.65 or about 15% more power (heat), it will be 1.9**2 / 1.6**2 or about 33% more power (heat).
Intel has really good brief data sheets as far as the processors go. The chip set data sheets are almost as good.
The processor and chip set have to agree on Front Side Bus speed, but it is between the chip set and the memory how fast that bus will go. In either case, the negotiated speed is the lower of the two, so if your chip set will only do 1.333 GHz, that's how fast your 1.800 GHz memory will go.
Did a little research on the i5-750, which is a very slick little package. Forget about FSB, this thing seems to have the chip set embedded -- memory bus and PCIe included, making FSB is irrelevant. According to the data sheet, it supports 1.033GHz (1.066?) and 1.333GHz memory, so your memory will be negotiated down to 1333. The voltage caveat remains.
A little processor history. In the days of the big mainframes, the fastest logic was ECL -- a non-saturating logic. What that means is you could set the clock rate to 0 and it would still consume about the same power. CMOS is saturating logic, which means that the electrical state becomes static at the clock. The power is consumed going between states, so clock speed is critical: set the clock to 0 and it's power consumption will be pretty close to nil, depending on how the chip talks to the rest of the world. Clock it too fast and it's effective resistance goes down, probably in inverse proportion to the clock rate. P=E**2/R, so if the resistance goes down, the power goes up.