Screenshot of the second score/highest I ended up getting in the session tonight.
 
PCB heatsoak on the Kingpin, as Vince has talked about in live streams with OC sessions with GN etc, is a real thing.  My card ran a few good runs at those settings then the score started going back down despite me letting the card cool down between runs.  My GPU clock randomly started going down despite the same settings.
 
This run was done with ONLY a 69.6F ambient in the room, fans at max on the system, nothing special at all.  And CPU was at STOCK no PBO no nothing.  Insane.  Next time I'm going to try running my curve optimizer settings that boost to 5.1GHz in the low loads like Port Royal and only run 1 CCD instead of both. 
 
I tried +180 and it failed about halfway through the benchmark, and I'm not really comfortable going past where I'm at in voltage, especially because of how I'm hitting 500-551W load on the OLED power readout - I'm not running the 1K BIOS so I'm hitting power limit during the benchmark I think, which means any higher voltage would actually not help improve anything.
 
I also can't get my VRAM to run any higher than 1400 even with the memory voltage bumped to 1.4125V - I tried 1500 & it actually ARTIFACTED in the benchmark then crashed out of it.  I definitely got the bad end of the stick when it comes to my memory chips not hitting speeds other people can hit.
 
Worth noting, the +165MHz offset combined with the low ambient & high speed fans, average temp was 43C... not sure what my peak was because all background processes were closed & I did not look at the graphs after the result.

Yes, any 3090s should SLI together.
 

w_louie

Spurious_ECG
One KP Hybrid + one FTW3 Ultra HC: https://www.3dmark.com/pr/1026718
On stock paste and bios as well. Can't wait to see if I can make top 10 by the end of 2021.
 
Here's what my mess of a setup looks like.
 

 
Kingpin and FTW3 Ultra can NVLink together? :O

magillaman
Antsy waiting for my KP HC block, so I ran some benches to get a baseline... in the process, finally cracked 16k!

PR - 16075

Can't wait to compare/try to smash that score with the HC

WOW! Good job.

magillaman
Antsy waiting for my KP HC block, so I ran some benches to get a baseline... in the process, finally cracked 16k!

PR - 16075

Can't wait to compare/try to smash that score with the HC

Your average temp was 36C, are you using AC or something to cool down the radiator?

Liquid metal is something I would consider fancy/unnecessary/dangerous.  Still, nice score.  You're hitting around 2225 through most of the test from the looks of it.  What offset are you using that it's showing 2280 at the start, and are you messing with voltages in Classified Tool or not yet?

magillaman
LM is quite simple, and safe if you’re comfortable with it.. Conformal coating on the components around the die, mount the card in the pcie slot, never vertically on a riser, and you’re fine.

Using VF curve in Afterburner, 2280@1.1mv, also using Classy Tool. My chip doesn’t require much tho.. giving it 1.16875 nvdd (LL 7), 1.0875 msvdd (LL 6), and 1.36250 for the mem.

Ah. I haven't even went past 1.14375V yet on NVVDD, that explains how you're hitting a higher speed than I can so far.  Next time I'm doing an OC session I'll give'er a bump on that & see how it goes.

Anyone know how much power you can push through the 3080 ftw3 hc? Wanted to know how it compares to the 3090 in power usage since kp can push 520-550
Edit-Oops asked wrong thread 😬

howe10888
@Dabadger84  Thanks for the advice and I broke 15.6K last night...
Port Royal 15640 - https://www.3dmark.com/pr/1037914
Time Spy Extreme 11753 - https://www.3dmark.com/spy/16692392

Are you running any form of special benchmark-only Windows or just normal with background processes closed?  That is crazy that you got that good of a score with an architecture that is that old.  Was there a 3DMark update recently that fixed the scoring issues people were experiencing where almost everyone's system was suddenly scoring 500-700 points less than previous at the same clocks/same drivers?

Dabadger84

howe10888
@Dabadger84  Thanks for the advice and I broke 15.6K last night...
Port Royal 15640 - https://www.3dmark.com/pr/1037914
Time Spy Extreme 11753 - https://www.3dmark.com/spy/16692392

Are you running any form of special benchmark-only Windows or just normal with background processes closed?  That is crazy that you got that good of a score with an architecture that is that old.  Was there a 3DMark update recently that fixed the scoring issues people were experiencing where almost everyone's system was suddenly scoring 500-700 points less than previous at the same clocks/same drivers?

Just normal windows 10 20H1, ultimate power plan, core +165, memory +1200, 100% fan, max slider on voltage, power level and temp.  LN2 Bios with rebar enabled.  Back plate heat sink and fan.  NVVOD 1.15, FBVOD 1.3875, MSVOD 1.10, Both LL1 and OCP Disabled.  Nothing special and that showed the Intel system(I built it in 2017) is still pretty good even compared to the current systems.

howe10888

Dabadger84

howe10888
@Dabadger84  Thanks for the advice and I broke 15.6K last night...
Port Royal 15640 - https://www.3dmark.com/pr/1037914
Time Spy Extreme 11753 - https://www.3dmark.com/spy/16692392

Are you running any form of special benchmark-only Windows or just normal with background processes closed?  That is crazy that you got that good of a score with an architecture that is that old.  Was there a 3DMark update recently that fixed the scoring issues people were experiencing where almost everyone's system was suddenly scoring 500-700 points less than previous at the same clocks/same drivers?

Just normal windows 10 20H1, ultimate power plan, core +165, memory +1200, 100% fan, max slider on voltage, power level and temp.  LN2 Bios with rebar enabled.  Back plate heat sink and fan.  NVVOD 1.15, FBVOD 1.3875, MSVOD 1.10, Both LL1 and OCP Disabled.  Nothing special and that showed the Intel system(I built it in 2017) is still pretty good even compared to the current systems.

Yeah that's a really good score for an architecture that most people would consider antiquated.  I approve 

Got my Hydro Copper in!! Swapped from AIO to block, used liquid metal and 12.8w/mk pads. On it's own loop w/360 60mm rad in push + pull. Classified tool used, and on OLD version of 3dmark before anyone asks. 
 
Think I have a nice piece of silicon here! :)
 
3dmark.com - PR - 16,312

magillaman
Got my Hydro Copper in!! Swapped from AIO to block, used liquid metal and 12.8w/mk pads. On it's own loop w/360 60mm rad in push + pull. Classified tool used, and on OLD version of 3dmark before anyone asks. 
 
Think I have a nice piece of silicon here! :)
 
3dmark.com - PR - 16,312

Nice score and temps. What thermal pads thickness did you use for front and backplate? Thank you

SoldierRBT
Nice score and temps. What thermal pads thickness did you use for front and backplate? Thank you

Ty! 2mm on vram for both sides.
+ another 2mm pad on top of the backplate with big aluminum heatsink and fan.

Dabadger84
What'd you push voltage to to hit that clock stable enough to bench?  That's an impressive score indeed.

Same NVVDD, LL, and Mem as before, only increased MSVDD to 1.09375 and the vram is much cooler with the block + new pads so I was able to push it to +1750

magillaman

SoldierRBT
Nice score and temps. What thermal pads thickness did you use for front and backplate? Thank you

Ty! 2mm on vram for both sides.
+ another 2mm pad on top of the backplate with big aluminum heatsink and fan.

Dabadger84
What'd you push voltage to to hit that clock stable enough to bench?  That's an impressive score indeed.

Same NVVDD, LL, and Mem as before, only increased MSVDD to 1.09375 and the vram is much cooler with the block + new pads so I was able to push it to +1750

I swear I got the worst Memory on any Kingpin that's reported performance numbers on these forums LOL I couldn't do more than +1250 in SOME benchmarks on the AIO cooler, now it's 50/50 on whether or not I can get anything over 1300 to pass at stock voltage, without going over 1.4V Mem voltage it won't go a tick higher.  I did get 1400 to pass Port Royal with 1.4125V, but it only passed it 2 outta 4 times, so either PCB heatsoak was kicking in, or it was not very stable.  My core is pretty good, memory is KekW.
 
 

magillaman
Got my Hydro Copper in!! Swapped from AIO to block, used liquid metal and 12.8w/mk pads. On it's own loop w/360 60mm rad in push + pull. Classified tool used, and on OLD version of 3dmark before anyone asks. 
 
Think I have a nice piece of silicon here! :)
 
3dmark.com - PR - 16,312

Wow! that is crazy good, What was your core max before swap?
 
I got my block in, but not sure I would gain much with it. So debating just keeping it around in case AIO fails. My effective core max is like 2205.