New info learned

Subscribed to the thread hoping for some more info on the dip switches. 
Do folks prefer to use the dip switches or to use the classified tool?
My understanding is that the dip switches are the only way to maintain voltage settings between reboots?
Thanks, 
J

TIL: exceeding 350w on the GA102 is instant death of chip and it doesn't matter that the upcoming 3090ti uses 450w in FE guise.

In my experience, I can't say that having your effective clocks match your apparent clocks is as simple as throwing the NV and LL switches.

DIP switches bump your nvvdd by 0.015 and 0.030v. There is no PLL1, there's an LL1 switch which is loadline and that keeps your nvvdd and msvdd from drooping as much. Turning both on could in theory make your effective and apparent clocks match, but only if there's a small discrepancy to begin with and it would achieve this by bumping your voltage by some amount which is technically the same as extending your available VF range beyond 1.1v (unless you're undervolting in which case it just makes it so you're not undervolting as much).

For example, I can set my target clock at 0.950v to 2070 MHz, and with a GPU temp of 47 drop down 2 bins to 2040 MHz. This is reflected in my apparent clocks across all software. If I look at effective clocks under a 100% load, they read 1965 MHz. If I trip the NV switches on, in the same load it bumps up to 2006 MHz effective. If I put the LL switch on as well, it drops back down to 1980 MHz. My thinking is that the LLC doesn't let the voltage droop as much leading to increased internal temps which makes the same clocks less stable. This is all while not power limited so I don't know how that changes things.

Interestingly enough, if I increase the voltages to 0.975 in afterburner at the same requested clocks but with no DIP switches on, I get the same results. Which makes me suspect the DIP switches only help if you're maxing out your voltages and not power limited and then only when you aren't thermally challenged.

Just a bit out of my experience.

homestyle

DemonHunter117
TIL: exceeding 350w on the GA102 is instant death of chip and it doesn't matter that the upcoming 3090ti uses 450w in FE guise.

This is a big time statement. Care to explain further?

It was sarcasm. Someone above alluded how going over the 350w limit of a 3090 will cause long term damage to your chip and thus you should only ever buy FE cards and undervolt factory overclocked cards. This is erroneous because these chips are rated for much higher power limits than stock considering how newer SKUs come out using the same die and have a higher stock power level, like the 3090ti uses the same GA102 chip found in the 3080, 3080ti, and 3090 but its power limit is set to 450w whereas the powerlimit of the 3080 is 320w and the 3090/3080ti is 350w. This goes to follow that running a higher power limit (within reason) on a given chip is perfectly safe so long as you are aware of what you are doing and can manage voltages, amperage, and heat.