This method shown in this post is actually leaving the stock curve entirely and just capping, like you said.
Another method seen on this subreddit instead raises the whole curve and caps, this is basically overclocking/undervolting the entire curve then capping. The most common method I've seen online actually underclocks/overvolts at low voltage regions, overclocks/undervolts near some peak threshold, then caps the curve so it never exceeds that peak. The reason this is true is because the curve is enforced to be monotonic up and to the right. With V as the x axis, shifting the curve up overclocks the freq for each voltage point but is identical to a left shift of the curve which is an undervolt, reducing voltage for same clock speed.
The difference is just whether you have V as the x axis or freq as the X axis. I'm 99.999% sure that undervolting and overclocking are actually the same thing in theoretically, but sort of refer to different things in practice. So I started undervolting this week and spent a lot of time thinking about this.
0 kommentar(er)
