Overclocking RAM for simulation and compliation

[Altera_Forum]

A bit of a curve ball but I'm about to build a new computer and wondered if it is hugely beneficial to overclock the systems RAM (>1600MHz) when it comes to running compilation and logic simulations of systems?  

 As a quick spec the system will have 32GB RAM with core i7 (Haswell). If any body has recommendations regarding the best RAM to go for these will be very much appreciated.   Many thanks in advance

[Altera_Forum]

Main memory accesses are so much slower than cache accesses that increasing the cache size (or decreasing the size of the working set) is likely to have a greater effect than minor changes in memory speed. 

Memory with a higher clock speed often requires more clocks before the data is returned - so the speedup isn't as great as you might imagine. Clearly running the whole system in a bath of liquid nitrogen so that the clock speed can be increased by 50% will make things significantly faster! You might also find that compilation/synthesis is unable to keep all 8 (with hyperthreading) cpu cores busy. Since the busy thread keeps changing the OS may well schedule them of different cores - and then reduce the cpu clock frequency because the cores are idle! So look at the power (safe) settings - they may well have a larger effect that the DRAM clock speed.

[Altera_Forum]

Also, be aware that simulation (if you are referring to modelsim) is artificially slowed by the licence you buy. It also used to be the case (and still may be the case) that the 32 bit version of modelsim actually ran faster than the 64 bit version. And Modelsim generally doesnt use more than a single core (again, this used to be the case, Im not sure with the newest versions). 

 So theres too many variables that affect the speed, much more greatly than system clock speed.

[Altera_Forum]

Many thanks, I had a feeling that the *potential* gain with overclocked RAM would be far outweighed by other system factors