UndesiredSanity

tresnugget
Are you running precision while benching? Some things on the oled only work while precision is running.

That's the 890w Asus XOC bios
https://www.techpowerup.com/vgabios/245448/245448

No, I don't I close PX1 because it robs resources, FPS and hurts the score a bit. I always close PX1 or MSI afterburning when benching. So, you're telling me there is no real standalone controller driving this OLED when it comes to updating the clock? PWR, temp, fan speed and other parameters seems to change (not really accurate as mylittlepwny2 has pointed out). Something must be driving these parameters. That's why I think it's the firmware of whatever is driving this OLED (has to be some micro controller like a M0+ or M4/M33 driving it. 
 

I don't know about the Ti but on the 3090 the power still worked, the temp probe still worked, and I think the fan speed but I might not be remembering it right.

Some things worked and some things didn't. I assume the Ti is the same but I can't remember.

UndesiredSanity

tresnugget
Are you running precision while benching? Some things on the oled only work while precision is running.

That's the 890w Asus XOC bios
https://www.techpowerup.com/vgabios/245448/245448

No, I don't I close PX1 because it robs resources, FPS and hurts the score a bit. I always close PX1 or MSI afterburning when benching. So, you're telling me there is no real standalone controller driving this OLED when it comes to updating the clock? PWR, temp, fan speed and other parameters seems to change (not really accurate as mylittlepwny2 has pointed out). Something must be driving these parameters. That's why I think it's the firmware of whatever is driving this OLED (has to be some micro controller like a M0+ or M4/M33 driving it. 
 

PX1 must be open and minimized for most of the monitoring to function. A few of the options may run without it but most wont. It doesn't rob enough resources to hurt 3dmark scores. Any differences youve seen are simply variations within the margin of error. For an ultra memory intensive benchmark like Aida or geekbench I would shut it down completely but for 3dmark most pros simply minimize px1 and then run the bench. That's also what I do.

mylittlepwny2
I didn't try pushing voltage with any of the 3 stock bios. They all are the same as far as I can tell. 516w limits. If you hook up the voltage read point and do the math off of it up youll be able to calculate actual power draw. The oled isn't very accurate. My oled constantly says my card is pulling 510-530w in port royal. But I can see that my clocks are dropping (and not due to gpu boost) when running stock bios.

Not all cards pull equal power. Take both my 3090s for instance. My original launch day 3090 KP pulls well in excess of 500w and even when gaming is always hitting off the 520w power limit. Itll pull around 550w in timespy. My 2nd card (Jan 2021) barely pulls 440w at the same clocks. My launch card scales better with voltage and cold but is a "worse" card to daily drive as it runs 1-2C hotter and always drops clocks unless I'm using the xoc bios. It has everything to do with leakage and asic quality.

It’s not exact but it’s close within reason ime. I only have 1 pmd and the 3 8pins draw between 135-145w each under load. Does that translate to what the card is pulling across all 6? Doubt. But the oled voltage increase is relative to my nvvdd/LLC selections so it is definitely not 520w

The PMD itself is not 100% accurate. It's using software to estimate power draw rather than something like a c clamp. However it's likely more accurate than the oled itself if you could connect all the 8 pins AND the pcie express slot.

Cards don't evenly draw power across all the connectors. Unfortunately the best you can do is make an educated guess. You don't actually need to plug in all 6 8 pin cables. As long as both 16 pins are plugged in you can run the card just fine as long as 3-4 8 pin cables are plugged in. I've benched my card with just 4 8 pins connected. If running ln2 I would suggest using all 6 but for the kinds of power draw were using on water we won't need more than 3. The card was designed to run on 3 8 pins just like the 3090. Nvidia was the one thar forced EVGA to adopt the dual 16 pin setup as that's what they're using on 3090 Tis and likely all future 40 series cards.

I don’t blame them. Hopefully it makes for a solid 40 series launch with most of the kinks worked out

I wasn’t going to order a second pmd but maybe I will. Could use a fan controller anyway

mylittlepwny2
The PMD itself is not 100% accurate. It's using software to estimate power draw rather than something like a c clamp. However it's likely more accurate than the oled itself if you could connect all the 8 pins AND the pcie express slot.

Cards don't evenly draw power across all the connectors. Unfortunately the best you can do is make an educated guess. You don't actually need to plug in all 6 8 pin cables. As long as both 16 pins are plugged in you can run the card just fine as long as 3-4 8 pin cables are plugged in. I've benched my card with just 4 8 pins connected. If running ln2 I would suggest using all 6 but for the kinds of power draw were using on water we won't need more than 3. The card was designed to run on 3 8 pins just like the 3090. Nvidia was the one thar forced EVGA to adopt the dual 16 pin setup as that's what they're using on 3090 Tis and likely all future 40 series cards.

They seemingly forced them to do the 16 pin, but dual 16 pins seems a little much, even for XOC. 3 8 pins should be safe up to like 750w as long as you're using 18 awg or thicker cables.

I know most post & test PR. but I just hit an new PB with this 90ti KPE and jump 110pt more than my previous PB.  Still on stock BIOS. Reduced memory offset a bit. That’s what changed.
 
http://www.3dmark.com/spy/29439100
 
 Also Took #5 spot  for my setup. 
 
https://www.3dmark.com/search?_ga=2.46657255.1708983173.1657973917-1015559352.1657434225#advanced?test=spy%20P&cpuId=2897&gpuId=1463&gpuCount=1&gpuType=ALL&deviceType=ALL&storageModel=ALL&memoryChannels=0&country=&scoreType=overallScore&hofMode=false&showInvalidResults=false&freeParams=&minGpuCoreClock=&maxGpuCoreClock=&minGpuMemClock=&maxGpuMemClock=&minCpuClock=&maxCpuClock=

Nice score. Ur card seems real strong! What's ur stock clock at normal ambient temps if you hit the boost lock button on PX1? 2085? 2100?

Anyone try two of the PerFE 12 cables with the Ti Kingpin? Just curious as they'd make my cable management much easier.

You mean zero offsets and just hit the Boost Lock? I haven’t used that button yet with this card. My past testing I get more crashes when OC’ing when I start a run Boost Locking and warming the chip up when the test is loading.

Let me do a run this way. Stand by.