Dabadger84
So here's an update, got the new heatsinks in, haven't had a chance to game on them yet, will be doing that shortly, but they seem to be pretty good stuff.  They're light, but the base is fairly thin, so that makes sense.  I scratched one with a screw driver just to make sure they're real aluminum - pretty sure they are:
 

Why not just do a big one?  seems like you could get a 1" tall heatsink maybe 100mm x 140mm.  I think with more surface area it might be better cooling.  I did find that 1000rpm was the best for noise myself. 
 
I can't tell with my back plate if the tiny heatsinks (4mm  high) has done anything... it might just be the effects of the fan I have blowing at it.  couple degrees lower... but seems marginal.

So I redid some of them position wise, and included a heatsink over one VRM area I missed, indicated in this:
 

 
And now, all of my temps are lower than before the heatsinks, with a higher overclock than the pre-heatsinks temps I have a record of.
 

 
Left is with no heatsinks, right is with the latest setup of heatsinks. Crazy that the one iCX Mem temp dropped over 11C.  

Dabadger84
 
Trying to keep the obstruction to a minimum where it can be minimized, only putting them in the areas where there's actually something on the other side of the backplate to transfer heat from, using this as a guide, the ones at the "top" and "bottom" on my picture are over the VRM areas (that are obviously NOT on the back of the card, and I'm unsure if there's thermal pads there either):
 
I'm fixing to run Cyberpunk for a bit & see how the temps shake out with an ambient that's fairly controlled, between 70-71F.

I'm debating whether or not I want to stick thermal pads under the vrm areas with a heatsink on the back plate..  all of this adds up substantially in cost... lol  I'm hoping KPE cards start shipping soon.  I'd honestly be happy with about a 2100mhz clock, as i'm boosting 2000-2050 with my Hybrid, giving me 100+ fps on a CRG9 monitor.  Since this monitor doesn't go above 120hz anyway, I'm just trying to maximize the current capabilities.

tigerlord
That is good news, they should come out before March 2021 then.
I hope so as I on the unofficial list 26 now  as all the rest of the parts for my new computer are just gathering dust from last year. such as my kingpin mother board 

alpharius194
 
Did you replace the thermal material on the heatsinks or leave them as they came? I need a few odds and ends from ppcs and may try a pack or two out while I wait for the water block. Jacob said hopefully end of the month they'll be up. Fingers crossed it's not another auto-notify with no actual date in mind. When I do that I'll actually put nice thermal pads on, I used alphacool 11W/mk on my XI hero monoblock which went well so I plan to use a mix of 11 and 17 on the kingpin.

I just ran them with the thermal tape that came on them.  I might replace it later, but it seems to work pretty well.  At one point testing with the side of my case off I grounded myself then touched one of the heatsinks that isn't covered completely by the fan (still has air blowing through it) while in game, it was pretty warm so they're definitely doing somethin' ^_^ not too warm to touch though.

alpharius194
I'm assuming you used aluminum over copper so that it does not interact with the backplate? Or just personal preference? PPCS has a bunch of both materials in quite a few styles.

Oh I wish I could find copper ones that are the style I wanted (the black ones), couldn't find any.  There's thermal tape between the heatsink & the backplate so I doubt it would be a problem anyway, but the ones I'm using are all aluminum yes.
 
So eVGA CS said 2mm on the backplate for the VRAM?  That's crazy thicc.

TechJessica87
Have you considered getting a thermal imager application for your phone, or something simple like a flir c3?

My msi seahawk 1080 heatsink has lived at 165 degrees since 2016 but openhardware never shows it go over 125 135 140 type numbers. Usually 125 130.

Given that eVGA iCX sensors have been verified by Gamers Nexus independent testing to be very accurate, I don't see the point in doing that, other than not knowing for sure where said sensors are exactly... I think they said which one is roughly where in their video, I'll have to watch it again. 
That's something unique to eVGA cards though, so it's less accurate on cards that either do not have iCX sensors (pretty sure most XC3s do not have the iCX sensors), or that aren't eVGA cards. 

Dabadger84

TechJessica87
Have you considered getting a thermal imager application for your phone, or something simple like a flir c3?

My msi seahawk 1080 heatsink has lived at 165 degrees since 2016 but openhardware never shows it go over 125 135 140 type numbers. Usually 125 130.

Given that eVGA iCX sensors have been verified by Gamers Nexus independent testing to be very accurate, I don't see the point in doing that, other than not knowing for sure where said sensors are exactly... I think they said which one is roughly where in their video, I'll have to watch it again. 
That's something unique to eVGA cards though, so it's less accurate on cards that either do not have iCX sensors (pretty sure most XC3s do not have the iCX sensors), or that aren't eVGA cards. 

I have been watching the pictures of the aluminium heatsinks to put on top of my Kingpin back-plate.

I saw people covering the areas of the vrams specially.

What if I cover almost the entire backplate with a 20mm tall aluminum heatsink and a 120mm pressure fan on take the heat from the heatsink?

Psychor
I'm running the latest version of Precision X1 and have push any available firmware updates to my card using that app. None of the firmware updates failed.

I don't normally look over at the video card's OLED, but I did so recently and have noticed that the GPU clock displayed on the OLED remains at 210 Mhz even though I know the card is clocking much higher during gameplay.
 
Has anyone else noticed this?

I keep mine pretty much only displaying board power readout all the time, so can't comment.

Dabadger84

Psychor
I'm running the latest version of Precision X1 and have push any available firmware updates to my card using that app. None of the firmware updates failed.

I don't normally look over at the video card's OLED, but I did so recently and have noticed that the GPU clock displayed on the OLED remains at 210 Mhz even though I know the card is clocking much higher during gameplay.
 
Has anyone else noticed this?

I keep mine pretty much only displaying board power readout all the time, so can't comment.

I figure it out. You have to keep Precision X1 running in the background even if you have the OSD button unchecked in the app.

I mean, you're only tickling power limit, you're not seeing a hard green bar or anything, so that's not really that bad.
 
You should be able to push way higher than +25 though.  Keep in mind, the Ampere cards (I think all of them) go in +15 increments, so you're either running +15 or +30 actually... from the fact that your clock is showing 2085, that's +30 - since default boost clocks are in the 2025-2055 range depending on temperature under load.
 
I'd say go straight for the gut if I were you, if you want power efficiency & unlikely to hit power limit, 2100MHz @ 1025mV should be stable, I'd be shocked if it's not... that's a +60 effective clock, with lower than stock voltage, should result in lower temps as well as less power draw.
Or you can go bonkers & try to do what I'm running for funsies, 2160MHz @ 1068mV, with a +1000 memory overclock.  No clue if that's something most cards can do, I have a feeling it's not, especially the 2160MHz @ 1068mV part - that's basically a +120 offset, with the second lowest stepping of stock voltage (1062mV-1100mV).  For me, that still runs cooler than stock, while giving better FPS, and still drawing less power (most of the time).
 
Trying to push raw clocks probably isn't going to get much, or maybe not any, further that going for a more optimized overclock/undervolt.  Just my opinion.  I haven't really messed with voltages in Classified Tool either - I'm unsure of what voltages are and aren't "safe" for 24/7 slash full-time usage, so I haven't bothered with it at all outside of the one time I messed with another BIOS.
 
For reference, this is what I saw when running the 2160/+1000 settings, in a pressure run of Cyberpunk 2077 with all the settings maxed, including RT stuff being fully on with the last setting on "Psycho":
 

The voltage says 1.094v because I had to reset to stock for a moment when the core was boosting to 2175MHz because I didn't want that to cause a crash.
Screenshot at my resolution with those settings, the game can definitely bring the 3090 Kingpin even at those clocks to it's knees:

alpharius194
If you are having issues at default clocks or +15-30 when corrected and your psu is strong enough to power everything you may have an issue on the card. Like badger said, the card steps in 15mhz increments, any changes that are not 15 will either clock up or down to the next cycle. But do realize the classified tool is 95% of the reason to get this card. You really are going to struggle to push the 520w limit without adjusting the voltages under classified based on what the card defaults to and I'm not sure any of us actually push 520w consistently in game without the XOC bios. The kingpin cards OC guide lists 1.4 FBVDD (memory) for the waterblock card so I would assume that's safe if you can keep it cool enough like those heatsinks can. I tend to run my MSVDD at 1.0375 when I remember to go into classified and load a profile, my clocks still all run at default but I gain some performance out of loading my profile. Vince said in the first ln2 showdown he ran really high Voltages on the MSVDD if I remember correctly and was trying to get Jay to try it at more like 1.2. At some point I will try undervolting the card again but PX1 one or two versions ago was having trouble saving and loading my curves so I gave up and just use Auto NVVDD out of laziness.  

I read some posts ago the "safe settings" to use with the Classified Tool. Used them and I saw some improvement of performance.
"1.125 nvvdd, 1.4 fbvdd, & 1.2 msvdd. Leave both load lines on level 1 and disable both ocp."

But even that they are considerate safe they mention not to use for daily gaming. Just for benchmarking. That's the only time I got close to 520W but I haven't used since I read about the changes of breaking the card with time (even with those safe settings). And the warranty...

I use stock cooling with aluminum heatsinks in the backplate. My GPU temp is around 50°C most of the time with fans at 80%.

Most of users here with stock cooling use the Classified Tool?

I should’ve thrown an asterisk on mine. Vince was saying that about benchmarking, so I figured a MSVDD under 1.1V would be okay and at auto NV I didn’t see much improvement over higher voltages on MS. My temps are all really nice, usually under 50C and 70C on the hottest memory (my vertical gpu has the power cables kind of blocking the backplate). It’s new home is currently being assembled now and hopefully I’ll get more airflow over the backplate on it since the power cables exit a different way 🤞

Dabadger84

alpharius194
I should’ve thrown an asterisk on mine. Vince was saying that about benchmarking, so I figured a MSVDD under 1.1V would be okay and at auto NV I didn’t see much improvement over higher voltages on MS. My temps are all really nice, usually under 50C and 70C on the hottest memory (my vertical gpu has the power cables kind of blocking the backplate). It’s new home is currently being assembled now and hopefully I’ll get more airflow over the backplate on it since the power cables exit a different way 🤞

I do wonder if 1.2V for NVVDD would actually be "okay" for regular usage IF you keep it cool though.  Technically speaking that's "only" 100mV over maximum stock voltage since you can get the GPU to run 1100mV with the GPU voltage slider maxed in Afterburner (though it usually won't stick at 1100mV, it caps out there), so like 9% over "max" stock voltage.  I can't imagine that is high enough to actually do damage with prolonged use... maybe if you're hammering the GPU 24/7 with Mining or Folding at Home, but for someone like me that games sometimes not at all one day then 8 hours the next day, I find it hard to believe that 1200mV would be "unsafe" for regular usage, if it runs cool enough on the VRMs & core.
 
I may try it out once I get the thermal pads in between the backplate & the VRM's backsides & GPU die backside as well, see what I can push with that voltage/if the extra clock is even worth the extra heat.
 
I think that's the real question, what kind of clocks can you squeeze out of the extra voltage.  Given that I've already seen quite a bit of "diminishing returns" in terms of clock capability per voltage increase, at least on my particular card, we shall see what that extra 100mV nets.
 
Going from 2100MHz stable at 1025mV to only be able to do 2160MHz at 1068mV gives me doubts the clock increase I can squeeze out of 1200mV would be worth the extra heat etc.  But who knows until ya try, right?  Personally I'm still impressed that 2160MHz is "stable" enough to game on for hours at 1068mV.  I've tried 2175MHz but that will occasionally cause graphical issues or crashes at that voltage, and I can't seem to get Afterburner to make the voltage stick any higher than that for seeing if I can actually get 2190 or so stable enough to use 24/7 without messing with Classified Tool.

After playing with afterburner and the classified tool I found that my best will be:

GPU 2160mhz @1.043v
VRAM 1164mhz
Max power draw is 460W.
Port Royal 14725.

No classified too cause I hit power limit easy and get random crashes.

Is it ok to leave 1.043v constant while playing? Also, is it ok 1.067v constant as well so I can try to push to GPU 2000mhz?

Dabadger84
I found that 2160MHz will run benchmarks as low as 1043mV for me but it wasn't totally stable in games like Cyberpunk until I went up to 1068mV.  Either of those is fine for constant usage, I've been running mine at 2160MHz @ 1068mV for hours and hours in Cyberpunk with zero issues - but the caveat of that is I also have good airflow & those heatsinks, which means my temps are well in check.
 
1043mV is nothin' to worry about for sure, neither is 1068mV.  Remember these cards run up to 1100mV "stock" without Classified Tool, just maxing the voltage slider in Afterburner gives you that bump from 1093mV to 1100mV being the max.
 
I think it will vary card to card a fair bit where people can stabilize, but I think the majority of cards should be able to hit at least 2130-2160MHz actually stable, if they can figure out what voltage to stick it at.

where does one get the classified tool if they have a kingpin?
 
edit: when i try to change the voltage in afterburner using the OC scan it gives me an error.