kanecvr wrote on 2016-11-02, 15:50:
I'm probably in the minority here, but I don't even see the point of overclocking old PCs, unless you're into that sort of thing and that's your primary objective to begin with.
I mean, if a game runs too slow on a K6-2 450MHz it will probably not be that much faster on the same CPU at 600MHz (law of diminishing returns of SS7 and all that). If you need more performance, just build a faster Pentium III or Athlon machine.
Agreed. From 400Mhz to 500Mhz you will see a ~ 18-20% difference, but going higher, performance percentage will diminish greatly.
From 300 to 400, you'll see like 11% differance, but from 400 to 500 18-20%? From where you have these percentages, because they are wrong. Between 400 and 500, amd k6-2 rise very bad. Basicaly, it's optimal limit is around 300 mhz and 100 FSB speed, after that, it raise really slow. Because cache stays at 100 mhz, it is holding it back.
The differance between amd k6-2 300 and 400 is around 11%-13%, while you'll have 33% increase in clock speed.
Between 400 and 500, it will be even worse ratio (because cache staying at 100Mhz, holding it back even more). So with 25% increase clock speed, my guess would be 7-10% gain in fps in games, not 18-20% 🤣. It is worth to take k6-2 high, no doubt, if you can, but differances are really tiny. There will be increase, but not so big. I would rather undervolt it, and puting it low , around 350 Mhz, than taking it to 500 or 550 with increased voltage. Really, differance is not so big. Amd k6-2 limit is around 300, after that it is raising really uneffectively.
Here's another example:
Differance between K6-2 300 and 400 is tiny (around 11%), but with K6-3, it scales much better. (thanks to cache), particulary in more demanding benchmarks, as the last one. Conclusion - for me, I would not take K6-2 too high, rather, preserve it, by undervolting it, so it lasts longer.
K6-3 on other hand, is worth to take at high speeds, because it scales much better thanks to on-die cache. Same with K6-2+.