Originally Posted by cekim
1. I haven't used OBS, but it compresses and encodes, does it use AVX instructions to do so?
2. Windows is really terrible at resource management, it both under and over subscribes CPU resources such that the default behavior has high core count processors failing to reach full speed when load is uneven. I have no issue testing "by usage" setups in linux. Windows often loads up "unused cores" for reasons I can't explain or predict and drives things to the "all cores" clock rate.
3. Even when its working well, you are going to see fluctuation based on I/O bottlenecks (disk, network, filesystem, etc... ) So, some of this is normal.
OBS doesn't utilize AVX to my knowledge, I'll double check w/ the creator though gimme a few hes away atm.
Shouldn't be any bottlenecks as far as filesystem / disk go since 960 pro however I could try making it utilize the 850 pro instead but no drops in the write times and only testing for 1 -3 minutes at a time.
Originally Posted by Jpmboy
the adaptive per core settings are based upon the individual core's VID. Remember, adaptive cannot run below the VID for a given core and I was able t get the *cores to 4.6 with only a slight bump in the 4.5VID. Yeah - it's an odd setup, but had to try. Synching all cores is very straight forward. When pushing things a bit more than 24/7 work, I go with manual override running below the VIDs, c-states and the turbo residence settings in that screen pack.
RUns fine as a per core.. and free benefit is 2 cores at one notch higher with little more voltage. I wnat to try a per core manual override and drop the vcore for each as far as stability will allow.
And.. I do not use p95 on this processor. It's a left over from the Jurassic cpu epoc.
x264, x265, HCi memtest, GSAT (which uses AVX). For me, getting really solid ram stability is the most important. I can adjust core voltage on the fly if ever needed.
Edit: VCCIN... stay (well) below 2V. I have not needed more than 1.8 for normal use, 1.95 for some benches (like the final two modules in geekbench 3 (x64).
Ah ok thanks, noted for VCCIN as I've seen a few variance between 1.8 and 1.9 with some people. Also I noticed you're running at 200% vs 140% is that safe I thought 200+ was for LN2? ( based on the text file you uploaded )
As for OC wise, any settings I could try offhand just to test this thing, ie vcore, SA , VCCIO, and cache? and or whats safe to try?
I'm grabbing those apps you mentioned now, already grabbed cinebench, realbench, handbrake, and purchasing HCI. Also grabbed h264Edited by LunaP - 11/20/17 at 11:06am