Economy of Effort

Twitter LinkedIn GitHub Mail RSS

Overclocking: Getting The Low Hanging Fruit

I have started playing a game called Day Z (a mod for the game ArmA 2). I will write about this at a later time. One thing I will mention, though, is that ArmA 2 is a very CPU-demanding game. As a result, I found myself wanting to get some extra performance out of my system’s processor. I decided that I would overclock it, as I have done on many of my systems before.

One thing to understand about CPU production is that Intel and AMD don’t design completely separate cores for each CPU product they sell. Rather, they’ll make one CPU core and clock it at different speeds to make a range of products.¬†What this means to overclockers is that it’s often trivially easy to buy a CPU from the low-end of that range, and overclock it to run at the speed of the high-end.

To be honest, this barely qualifies for “overclocking”. It’s more like removing an artificial restriction that makes the CPU run slower than the higher speeds of which it is capable. “Real” overclockers like to take CPUs and see just how far they can push those cores, way beyond the normal range that Intel or AMD are willing to clock those chips for wide production. Overclockers utilize more aggressive methods of keeping the chip cool, increase voltage beyond the stock levels, and other such tweaks to push a core to its maximum stable (or, sometimes, not-so-stable) limits.

This is not the level of overclocking that I participate in. I am simply interested in getting the low-hanging fruit. Taking a chip clocked at the low-end of the range of CPUs using the same core, and turning it up to the high-end, is a very easy way to spend less on a CPU and end up with the same level of performance. Sometimes, depending on how much cushion is left, a chip can go even higher relatively painlessly. (Intel/AMD don’t necessarily always max out a core’s capable range. Sometimes they simply move on to the next core.) When buying CPUs, I tend to buy on the low-end of the core’s range, knowing that I will probably have an easy overclock ahead of me if I need more CPU power.

My current system runs an AMD Phenom II x6 hex-core CPU. It’s the 1055T, giving it a stock clock of 2.8 GHz. The other chips that use the “Thuban” core that this one does run up to 3.3 GHz. What’s more, overclockers found it trivial to clock them up much higher - AMD definitely did not release any CPUs that reack the peak range of this core. So, I bought a nice big heatsink/fan combo (Cooler Master Hyper 212 Plus), replaced the crappy stock AMD heatsink, and clocked the chip up to 3.5 GHz. It’s a nice 700 MHz boost, yet overall it is still a very conservative overclock. With the new heatsink, my CPU temperatures are chilly cool (a good 15 Celsius cooler than they were at 2.8 GHz with the stock AMD fan), so I know heat is no factor. I hear 3.8 GHz is reasonably achievable, though with some slight voltage bumps. Nothing I’d be afraid of, but at this point, I start to reach diminishing returns. I have successfully picked the low-hanging fruit.