Dear all
I wonder whether overclocking the CPU or the memory can significantly reduce the processing time of optimisation with tools like Galapagos or Wallacei in complex multiple or single-objective optimisation simulations?
On a separate note, I appreciate it if anyone who has done this let me know whether it was any useful?
Best
I can only give you a partial answer - a few years ago I overclocked my system because I was going to edit a bunch of HD video clips, and I wanted to skip the “intermediate” conversion step that basically doubles the total length of processing time. So I did the overclock and tested it with all my usual applications.
However, I encountered repeated BSOD’s (blue screen of death) during my video editing runs. For reasons I was never able to identify there was something about the HD video compression./decompression algorithms that, when overclocked, caused a BSOD every time. So I went back to standard clock speeds.
My conclusion was that overclocking is tricky business - even though my system passed all the overclock tests it crashed with the video stuff. So my sense is you can try it, but you can’t be assured it will work for what you want. And it might not.
Single threaded processes will of course benefit from overclocking (OC) as long as the bottle neck is with the CPU processing (disk speed and memory speed etc may also be bottle necks).
But as with all overclocking, you’ll have to step-wise increase the OC and find out at which point the CPU starts to misbehave, and then step back one step. I’ve constantly run my ol’ i7 3930 with slight OC and have had no problems.
I had to stop at the first step of GigaByte’s OC schemes because at the second step I’d suffer from BSOD like @Birk_Binnard did. Just find a stable highest level, then off you go. I recommend using OC-software with predefined OC schemes, like I do in the screenshot below:
I’ve run my computer like this for nearly ten (10) years by now (from 2011) with no problem.
// Rolf
Wel…not really. Overclocking is fine, but you’re not going to gain more than…5% on irrelevant benchmarks without going completely berserk with cooling.
Modern CPUs already “overclock” single-threaded tasks, they’re constantly adjusting speeds, there isn’t a bunch of extra performance just sitting there waiting for you to turn on–like when I turned my…486 DX50 (NOT DX2!) into a crazy 66Mhz monster by overclocking every part of the system by 1/3rd it’s a wonder it worked. Progress in raw CPU speeds has drastically slowed down if you haven’t noticed, they are making incremental gains by trying to squeeze out every hertz they can already.
I attended a short Webinar with a Wallacei developer last year, they said they are thinking about cuda implementation, so that would speed things up.
Reducing calculation time per solution & well defined criterias are key factors in my opinion, but I think that topic could be discussed better on the wallacei group: https://www.grasshopper3d.com/group/wallacei
Thanks a lot for your reply Rolf. It was very helpful.
Just one question, are you using a liquid cooling system for your PC as you overclock it? Or you just use normal air-cooling system? And If you use Air cooling, does it get very hot or ever cause a problem due to overheating?
Many thanks
No liquid cooling. I actually used a liquid cooling system some three years ago, which started leaking and smoked my graphics card which was located below the leaking hose and… well, the good thing was that the store down the street had a sell out of some GTX1080Ti graphics card to a very low price that day… 
Today I’m using only a big fat CPU-fan (well some extra fans on the box as well) and no temperature problems whatsoever, not even with that GTX 1080Ti stove installed. As said, no cooling problems.
// Rolf
