Post History

Like always with overclocking, it is not guaranteed to work or be reliable. When one as a hardware manufacturer makes a claim about supported speeds, there's a lot of things to consider:

- Does it work within the whole specified temperature range? 

  Which is perhaps something like 0° to +85°C for a motherboard. They'll have to make the guarantee for what whole range, while in practice your motherboard might be some 30-50°C most of the time.
- Under which premises was it tested for electromagnetic compatibility (EMC)? 

  The speeds used in the test will have to be used in the real product or else the CE/FCC marking turns invalid and the product may turn illegal, as it might cause electromagnetic interference (EMI) with other devices or turn vulnerable against EMI. These are major concerns for high speed buses.

- Is the hardware clock provided accurate enough to support the speed?

  All data communication is based on clocks from oscillators, some manner of crystal oscillator in this case. An oscillator is picked based on which speeds you wish to support and what inaccuracy you may tolerate. Too bad clock margins mean that the data sent risk getting corrupted. Which in this case would lead to _way_ slower speed, rather than faster.

These are all hardware design decisions to be made in advance. Good engineers always design with margins, which is why overclocking is a thing - the part may work reliably outside the specified range up to a certain point where it doesn't.

The most probable error mode of RAM would be data corruption, in which the worst case is that the OS might decide that it should lay down and die (BSOD etc). You aren't going to cause any hardware damage by it. You may however provoke unexpected side effects like "coil whine" in some hardware, when you go outside the tested frequency ranges. PSUs are prone to this for example. Coil whine is harmless most of the time, but obviously very annoying.