xxplatoontekkxx
 

Got you Emmettt
https://www.3dmark.com/3dm/62492055?

Very Nice! Keep going!
 
I just got a second kingpin, and it's not so good compared to my first..
 
figures.
 
Memory tops at +900.. +1000 is "Iffy"... and +1100 is a no go, and core is crap too, oh well.
 
And it like 75 degrees in my lair too.. and no windows.

http://www.3dmark.com/pr/1074591
added 100 to mem, required a bump to 1.425 FBVDD
I really think its my CPU holding me back, sigh
Edit- also forgot to close google chrome for this pass lol

Couldn't get that setup stable again, might have been heatsoak and being on the edge already lol
I need to get that third radiator and some fittings and extra tubing, so I can redo my loop and add my second pump/res
This should drop temps down overall and decrease time needed to reach equilibrium at idle

CptT95
Couldn't get that setup stable again, might have been heatsoak and being on the edge already lol
I need to get that third radiator and some fittings and extra tubing, so I can redo my loop and add my second pump/res
This should drop temps down overall and decrease time needed to reach equilibrium at idle

I've found that PCB heatsoak/settings that ran stable resulting in lower scores happens after around 4-5 runs if I'm really pushing the card, I would assume that's what it is, heatsoak, and the only way to really alleviate that would be to either power the system down & let it cool off for like 20 minutes, or set to default/stock and let it cool down longer.
 
Leaving the voltages at their 'pushing it' benchmark settings makes the card run warm continuously, especially if you have the 'real' LN2 BIOS on the card, because then the memory is running full clocks (and voltage) at all times.  That's a significant heat increase and increase to the amount of heat the PCB is "heatsoaking" because of the memory chips being double-stacked on either side of the PCB in the same place (and the fact that the Memory chips are flip-chip, meaning the 'core' of the memory chip is closest to the PCB, not the cooling elements). 
 
Even if you lower voltages to stock between runs, heatsoak will still slow ya down after a while.  
 
I'm looking forward to my next session, whenever that is, aiming to get closer to the 16K mark if I can manage it.

*Edit - Nevermind you guys don't need to be subject to my ****ing & moaning.

Dabadger84

CptT95
Couldn't get that setup stable again, might have been heatsoak and being on the edge already lol
I need to get that third radiator and some fittings and extra tubing, so I can redo my loop and add my second pump/res
This should drop temps down overall and decrease time needed to reach equilibrium at idle

I've found that PCB heatsoak/settings that ran stable resulting in lower scores happens after around 4-5 runs if I'm really pushing the card, I would assume that's what it is, heatsoak, and the only way to really alleviate that would be to either power the system down & let it cool off for like 20 minutes, or set to default/stock and let it cool down longer.
 
Leaving the voltages at their 'pushing it' benchmark settings makes the card run warm continuously, especially if you have the 'real' LN2 BIOS on the card, because then the memory is running full clocks (and voltage) at all times.  That's a significant heat increase and increase to the amount of heat the PCB is "heatsoaking" because of the memory chips being double-stacked on either side of the PCB in the same place (and the fact that the Memory chips are flip-chip, meaning the 'core' of the memory chip is closest to the PCB, not the cooling elements). 
 
Even if you lower voltages to stock between runs, heatsoak will still slow ya down after a while.  
 
I'm looking forward to my next session, whenever that is, aiming to get closer to the 16K mark if I can manage it.

I put the 1k bios on the middle OC position, since that bios stock is completely useless
What I want to know, is why I am hitting power limit at 110% with stock voltage on stock ln2 bios, aka 460-470w power 🤔

nice

magillaman

CptT95
I was monitoring with MSI AB and GPUz in the background, using only one 60z 1080p monitor hooked up 🤔

Monitor refresh has nothing to do with it as you’re supposed to run with vsync off, but 1080p does help your score as opposed to a higher resolution.

GPUZ and AB only display the “core clock”. With ampere cards there is a secondary clock called “effective clock”.. you can monitor this using Hwinfo or Thermspy.

The closer core and effective clocks are, the more accurate your benchmark scores and overall performance.. That’s essentially what you’re doing when you increase NVVD and MSVDD with Classified tool.

For example here’s a PR run from a few weeks back @2190 with the clocks matching:

https://www.3dmark.com/pr/1045936

Would you mind posting your score with Both cards in, but SLI OFF and just a stock run with Porte Royal, thanks.
 
I noticed we have same Motherboard and CPU, curious about something.

magillaman

CptT95
I was monitoring with MSI AB and GPUz in the background, using only one 60z 1080p monitor hooked up 🤔

Monitor refresh has nothing to do with it as you’re supposed to run with vsync off, but 1080p does help your score as opposed to a higher resolution.

GPUZ and AB only display the “core clock”. With ampere cards there is a secondary clock called “effective clock”.. you can monitor this using Hwinfo or Thermspy.

The closer core and effective clocks are, the more accurate your benchmark scores and overall performance.. That’s essentially what you’re doing when you increase NVVD and MSVDD with Classified tool.

For example here’s a PR run from a few weeks back @2190 with the clocks matching:

https://www.3dmark.com/pr/1045936

Thought that the average clock is what let's you know effective clocks roughly, since clocks automatically fluctuate depending on available power, voltages, clock settings, thermals, etc
Will try again with hwinfo on when I get back home later today
Edit- iirc I saved those 2.2ghz settings as preset under 1k bios and for classified profile

Is the kingpin worth it (over a fe card) for someone just using it for gaming? No intention (or skill in) overclocking. Thanks

Krdonnel
Is the kingpin worth it (over a fe card) for someone just using it for gaming? No intention (or skill in) overclocking. Thanks

The founders is an excellent card, I have an afterburner preset for 0.975 @2100 and that runs most stuff, a few things I need 1.025.
and with a 60% fan speed it mostly stays in the upper 40's-lower 50's on temps. And although it's maxxed at 400W I see 420W all the time.
I think NVIDIA saves power efficient chips for the founders.

CptT95
I dont know what my issue is with my Port royal scores, is it my temps hitting 50C the issue?
I see someone with a 2205 clock set with average 2172 according to 3dmark, 41c average temps mem clock 1422(1219) and gets 15786 score
I cant even break 15600 at similar clock -_-
Edit-I feel like having my GPU near end of loop is not helping me with temps

What CPU and clock was that users score based on?

@magillaman
 
Is the 16K score with 520W or 1K BIOS?

CptT95
I put the 1k bios on the middle OC position, since that bios stock is completely useless
What I want to know, is why I am hitting power limit at 110% with stock voltage on stock ln2 bios, aka 460-470w power 🤔

Yeah that's the BIOS position I use for any testing of any BIOS I'm messing with because the stock OC BIOS is le garbageo.
 

Krdonnel
Is the kingpin worth it (over a fe card) for someone just using it for gaming? No intention (or skill in) overclocking. Thanks

The main advantages the Kingpin has over a regular 3090 are pretty well detailed by Cpt above, but the better power target, and higher stock clocks, along with better cooling, do make it a much better card for high resolution gaming than a regular 3090 in my opinion, but for JUST gaming, with no intention of overclocking, you'd be better off with a regular Hybrid 3090 or 3080 Ti (or 3080, depending on your resolution).  Keep in mind, the 3090 isn't actually 'meant' for gaming in the first place.
 

magillaman

CptT95
I dont know what my issue is with my Port royal scores, is it my temps hitting 50C the issue?
I see someone with a 2205 clock set with average 2172 according to 3dmark, 41c average temps mem clock 1422(1219) and gets 15786 score
I cant even break 15600 at similar clock -_-
Edit-I feel like having my GPU near end of loop is not helping me with temps

Your problem is your clocks are not accurate..
Look at these screenshots..
 
First one is when I open Afterburner and lock my OC to 2205@1.1v with VF curve.
You can see AB and GPU Z see 2205Mhz, but HWinfo shows the Effective clock is wayyy off.. 

 
In the second screenshot, I used Classified tool.
Disabled "Auto" and added to the NVVDD and MSVDD.. and now you can see all the clocks match.

 
Temps are not your problem, check this run I did where I hit 55c. (In the case with panels on, 21c ambient room temp)
PR 16,017
 
I believe this is the guy you're referring to: PR 15,786
His clocks are lower than yours, but the effective clock is probably more accurate, and he has more thermal head room since only hitting 41c. 
- He can probably score higher with more voltage/better tuning.
 
TLDR: Check and make sure your CLOCKS MATCH, then run your benches.. 

This is something I'm still figuring out on mine I.E. what voltages are needed to get it to actually run the clocks it's set to.  I'm a lot closer at the last ones I ran compared to the previous LN2 BIOS settings (which run 30-60MHz slower effective than what they're set at a lot of the time).

I get effective close to set clock and still either scores low or crashes on me 🤔