Min/Max CPU cache ratio helping stability

[kbalaz]

So I have been testing my system trying to maximize performance/power.
I can have it stable 43@1.18V, 44@1.22V and not at all at 45 any volt.
Watching another tutorial from JJ at Asus and explaining setting the Min and Max CPU cache ratio to help stability by setting them lower than the intended frequency ie; setting min/max to 43 and intended frequency at 44. by doing this I was able to lower voltage to 1.21V from 1.22V, not a big deal but I actually managed to set 45@1.275V with min/max cpu cache ratio set at 44 and maintain stability. Still find this a big jump to gain that extra 100mhz to go up .05V but it did it. Max temps at 1.275V were 71-72C using OCCT at the end of 30min.
In the end I backed it off to 44@1.21V min/max cpu cache at 44 and having max temps at 62-63 using OCCT.
thought I would share with the community, most probably already knew this, but it took me a bit to find the answer.

 

[metalgun2000]

I've been doing a fair amount of cpu performance testing with the cache / Uncore ratio at different settings to see if it had in if at all impact. What I've found is that what JJ and Asus have been saying about getting it to within 300 - 500 mhz below the cpu ratio is enough to not "bottleneck" the cores themselves. You don't have to get it to a 1:1 ratio like some people argue. My 4770k is currently at 4.2ghz and the cache is at 3.9ghz and taking that up any more isn't improving performance.

Intel has given us free reign over the cache ratio to help increase cpu core stability at higher multipliers. Haswell has twice the cache bandwidth compared to Ivy, this is the reason it doesn't need to be at 1:1 ratio. Actually my cpu falls off ever so slightly when it's at 1:1. But if you can get the same clocks with slightly lower voltage, that's awesome, that is the very reason Intel did what they did. I hope this helped.