Per-CCX OC results sanity check

Joined
Apr 5, 2016
Messages
2,249
Hey guys! Finally got the courage to attempt a proper overclock on my 3900x. I'm using an Asus Prime X570 Pro board.

I wound up with:

CCD0:
CCX0: 4375
CCX1: 4400
CCD1:
CCX0: 4375
CCX1: 4375

Vcore is at 1.3v, LLC level 4, temps well under control - 65c peak under Prime95 testing (large FFT with AVX2 enabled). Stability is 30 minutes of P95 with zero errors.

What do you guys think? Are the values good? Do I have headroom to keep going, or is this a good stopping point?
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
6,435
LargeFFT is really easy to run load wise. It's mainly a memory test vs a load test fyi. In other words it will be easier to get higher clocks with large vs small/smallest.
 
Joined
Apr 5, 2016
Messages
2,249
LargeFFT is really easy to run load wise. It's mainly a memory test vs a load test fyi. In other words it will be easier to get higher clocks with large vs small/smallest.
So my "stable" clocks may not actually be that stable if they don't validate with small/smallest?
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
6,435
So my "stable" clocks may not actually be that stable if they don't validate with small/smallest?

On the surface yea. Large is good for testing memory/imc stability. But it won't be a realistic load for everyday stability. Each setting small/smallest/large has a specific purpose. And blend combines all the load settings into one which is best for testing stability in all potential loads but it takes a while longer to cycle the differing loads.
 
Top