What’s the science that stops from getting higher overclocks?

P

plutoniumman

New Member
I know all CPUs have their limit to how high they can be overclocked. Like temperature, quality of electricity... ...And is that it?

I could see how noise on the electrical input could mess-up the chips (thanks collage & Parallax!), but why is heat a limiting factor? Obviously one reason why is because it’ll literally melt/burn the chip or board or etc., if there isn’t sufficient cooling. But why does it cause more errors?

I noticed when I overclock the VRAM in my GPU, it starts producing artifacts/errors. Is the reason a CPU can’t overclock too high, is because the L1 & L2 cache start producing errors, and can’t ‘keep-up’? (L1 is generally the same frequency as the CPU’s main clockspeed; more overclock = faster cache)

I wish my collage course covered overclocking better... Er, maybe I should’ve paid better attention...
 
Is that what prevents from getting higher overclocks? It seems more like it’s what causes a CPU to cease functioning from over-voltage, or of old age.

Is electromigration similar to electrosis? ie to plate one metal with another (like gold plating). If it is, this makes sense why overclocking, or higher clock speeds reduce the life-span of the CPU. But still I don’t quite understand why someone can’t get, say, 5 ghz if they can keep the CPU cool enough.
 
It's just inefficiency of the CPU. If the power getting to the chip isn't being turned into calculation performance, then it's either going to be turned into heat or noise (in some older cases).
 
plutoniumman said:
Is that what prevents from getting higher overclocks? It seems more like it’s what causes a CPU to cease functioning from over-voltage, or of old age.

Is electromigration similar to electrosis? ie to plate one metal with another (like gold plating). If it is, this makes sense why overclocking, or higher clock speeds reduce the life-span of the CPU. But still I don’t quite understand why someone can’t get, say, 5 ghz if they can keep the CPU cool enough.
Click to expand...

Electrolysis is using electricity to migrate ions, basically running a reverse battery. Electromigration is electrons bumping into and displacing atoms, so it's not the same.

And electromigration still happens at low temperatures, just not as much. And also, in order to achieve high overclocks, a CPU needs to run at higher voltages. However, the circuits inside the CPU are not designed to withstand such high voltages, so they break down faster the more you overclock.
 
None of the above is correct in terms of the question. It has nothing to do with electromigration, and certainly nothing to do with electrolysis.

As heat rises, so does resistance. As each nano-meter transistor switches on and off it is resisted due to friction. The faster we do this, the more friction per time unit. The more friction the more heat. As heat rises, gate resistance across the transistors change as does the break over voltage - differently due to the minute differences in the architecture and the doping materials. This means you get significant and relative timing changes across the CPU die leading to errors. Heat also causes on die memory malfunctions. The point is it is beyond design specifications.
 
Last edited:
You must log in or register to reply here.
Back
Top