Question Explain modern Nvidia (Maxwell and newer, Pascal, Turing, etc.) overclocking to me.

[VirtualLarry]

I guess, I sort of don't get it.

I mean, AMD's Ryzen 3000-series / Zen2 CPUs, have an aggressive "Core Performance Boost" setting that is default to "Enabled" in most BIOSes. This basically causes the CPU to "clock up" when there is thermal and power headroom to do so.

My understanding, was that NVidia's modern GPU "Boost method" was largely similar in application.

So what I don't get, then, is why manual overclocking of Mem or Core in Afterburner, or different "Factory OC settings", have any affect at all.

If a card has a Core clock of 1500, and a Boost clock of 1700, but the chip can go to 1900, is that why people OC the Core +200Mhz, because then it will "Boost" to 1900, rather than 1700? Is that what I'm missing here?

I guess, I don't get why "manual" OC even has an effect (if "Boost Clock" was working to full effectiveness), unless NVidia sets some pre-existing limits to the temp/power OC ability of a card (aka. "Boost Limits"), that are less than the actual limits of the silicon.

 

[Bouowmx]

Base and boost frequency are just some numbers to put on a 'box'. The real stuff is in the voltage-frequency curve of GPU Boost 3 (Pascal and later). In MSI Afterburner, click bars icon near Core Clock, or press Ctrl + F. When running a load and without limits, GPU will run at the highest frequency (up) and the lowest voltage for that frequency (left), usually ~1800-1900 MHz and ~1050 mV. An offset overclock, such as +200 MHz, moves the curve up; at same voltage, higher frequency. Without good sub-ambient cooling, there isn't much more to gain with just offset overclocking; the limit is ~2000-2100 MHz.