Hi,
 
So I just got around to looking at GPU Core voltage and noticed that I can't get passed 1.075v. Checked software and with MM
 
Even with the classy tool settings NVVDD @1.1375 and VVDD Loadline @ level 5.
 
Yes I installed the latest version of PX1 and performed the firmware update.
 
I want to make sure I figure this issue out so that I can move on to some DICE and LN2 testing. Would suck to waist $$$ stuck at a silly stock voltage. Tested on all 3 bios.
 
Any idea's?

Ya the only thing I have not done is the V/F point in PX1 or are you setting V/F in afterburner?
 
thx for the info

V/F curve doesn't do anything if you override voltage directly.

Get a candy from candythread here.

So I spent a couple hours logging incremental adjustments with the classy tool, PX1, and "AB 4.60 Final" and got all kinds of inconstancy.  At one point my KPE would drop to 469mhz GPU core and mem idle and then boost to desired clocks under load.
 
I Did see 1.089v on gpu core for a minute but wouldn't hold and Frequency was hopping all around like an Easter bunny. logged everything via google docs spread sheet as is my methodology...
 
To be clear my process when attempting to get passed 1050/1060mv was as follow:
 
Environment =
ambient temp ~20c
Bench -open case all max fans - fans = Noctua IPPC 3000rpm x5 +evga KPE clc fans - Moving enough air to blow down the green screen... 
temp sensors at PCIE ~28c
MAX GPU temp was 41c
MAX VRAM temp was 50c.
Idle GPU temp was ~27c
Idle vram temp was ~39c
During the majority of this session GPU temp at load was ~36c and VRAM temp was >50c
 
Workflow =
1. max out pwr limit ect on PX1
2.Launch the "tool" and set the slider to any VDDC voltage from 1.07 to 1.15 -save
3.soft reboot.
4.max out pwr limit ect on PX1
5. Launch 3DMark and set process priority
6. Start Custom runs of FSX GT1 & GT2 only
 
Ran this workflow on all 3 bios's
 
Sometimes this process would even nurf voltage as I was originally hitting 1079mv without any tweaks and mostly only saw 1050/1060mv via 3rd party software and multi-meter but PX1 would show 1010mv to 1014mv
 
Was up pretty late and at the end I decided to DDU nvidia drivers and uninstall AB and PX1 rebooted a few times and things went mostly back to normal. I think ive decided to wait for the Z390 Dark to arrive on weds and then will give a Fresh bench OS a go. When the dark gets here I will go with NO CASE and rig everything up on a Streamcom BC1 powered by a 1300 G2.
 
I am probably just tooo focused on something and am missing something simple. I can be pretty dense sometimes or the bench OS I was using is a ruined.
 
 
 
 

Do we still change the Memory and GPU in PX1?
What about the NVVDD Loadline? Keep at 10?
 

TiN_EE
Before you go any deeper, lets do some sanity checks.
a. If you using candytools then "Voltage" slider in PX1 does not do anything. You can leave it alone and ignore it altogether.
b. When you have no 3D load card will drop down clocks/p-state to low power, (300MHz when not overclocked).
In your workflow, you don't need to set pwr target slider, NVDD and then reboot, you may need reboot only if you touch memory voltages.
To monitor actual voltage use PX1, select to show GPU NVVDD sensor on OLED, or use DMM attached to ProbeIt port. Don't forget to uncheck Auto on candytool
At this point, before you get things behaving I would recommend to avoid using any 3rd part software, as anything else that access GPU likely to cause interference/issues instead. Keep things simple and test one at the time.

@TiN_EE
 
Thank you for the sanity check. Not that it needs to be said but of course I will memorialize the info you provided, follow these guidelines moving forward and it sounds like others will as well.
 
 I do this "To monitor actual voltage use PX1, select to show GPU NVVDD sensor on OLED, or use DMM attached to ProbeIt port. Don't forget to uncheck Auto on candytool"
 
my observation prior to tweaking the KPE with RTX of which I own a 2070, 2080 and this KPE was that base clock was 1350 ~ So could it be that during my observation of RTX the setting - NVIDIA control panel\manage 3d settings\power management mode = "prefer maximum performance" was preventing me from seeing the idle 300mhz gpu clock and instead keeping the base clock 1350. Until the driver blew up and I subsequently lost the power management mode value I am accustomed to? If not this specific power management toggle perhaps another anyway
 
I don't know but I can say that I am by no means an expert on any of this stuff. Just an enthusiast that contracted a bad case of overclocks from an AXIA Thunderbird back in the early 2000s and never really found a cure..heheh
 
Honored to play with this beautiful piece hardware, humbled by my ignorance and thankful for everyone's input here on the this forum.
 
 

I've added some details about loadline and how power limiter affects performance in the guide. Check it out.

TiN_EE
I've added some details about loadline and how power limiter affects performance in the guide. Check it out.

Thanks for another awesome update!  Very helpful.  Can't wait for that watercooling section update!

What is the Max Wattage you have seen being used?

Slow incremental climb on the memory is what I have been doing most of today.
I am running +1600 now on the Memory Keeping the GPU @ +178 2220MHz.
https://www.3dmark.com/compare/fs/19196132/fs/19193460
My Last Run and my First Run.
Ending here on Fire Strike 30380 Fire Strike for now. My Test Runs
Core clock 2,220 MHz Memory bus clock 2,150 MHz