Hey guys,
seems like the higher GPU utilization enabled by Martin’s trick is tied to a voltage drop of 40 to 50mV. I monitored the voltage and it’s infinitely reproducable, on Nvidia drivers 425, PiTool .129
P.S.: MSI 2080ti Gaming X Trio here.
Hope that helps!
Best regards
JoCool
(aka Fenris Wolf from IL-2)
That’s a very useful bit of info - thanks!
Infinitely reproducable as in it happens every time, or infinitely reproducable as in you can trigger the utilization trick by manually changing the voltage every time?
I’ll try undervolting the card slightly tonight and see what comes of it. This is nice info. It means we are getting to the bottom of what might be causing it. I hope anyway.
Wait, so what are you saying exactly?
Is the low utilization due to lower voltage supplied by the driver?
Or are you just saying that the voltage goes up when we do the trick?
If I have understood that correctly, he has twisted 2 things: 40mv to 50mv is not a drop, but a rise.
And this increase is not the cause of better gpu usage, but the result. Or did I just misunderstand everything? If the usage increases, of course the power consumption also increases as a result.
Perhaps the GPU frees up more usage if the voltage is dropped that little amount. It might be holding back utilisation if it senses the voltage being too high. That’s how I understand the OP. I am going to try reducing power limit and see if it holds back voltage slightly. I can access the voltage slider in Afterburner, but I am not sure if increasing it will give worse results, given the information above. Not even sure if it has any effect at all as Nvidia lock the voltage on newer cards. Also, this must be tied to PiTool somehow, as this only happens with select VR titles, and while numerous, certainly isn’t normal behaviour in standard 2D games and even some VR titles.
Good info. Keeping my eye on this!
I’ve been boosting my GPU using MSI afterburner only slightly but I set the power limit to be full to the right so maybe that’s actually a bad idea… hmm
OP needs to come back and clarify. What he posted is not good info, it’s garbage info.
“he has twisted 2 things: 40mv to 50mv is not a drop, but a rise.” - kellerbach
“OP needs to come back and clarify. What he posted is not good info, it’s garbage info.” - Milopapa
Wow, getting flamed in these forums for attempting to help. Some of the replies are really… something 
, whiney brats. Maybe some of you shouldn’t attempt anything yourself if you can’t even properly understand my initial post… looking at Milopapa and kellerbach!
Hint: The MSI 2080ti’s core runs at >1045mV. Hands off your GPUs you two, before you do something silly, fry them, and blame me in the end.
Good night.
P.S. Anton_Chez, Matt, and the other guys understood right away! Anyway, hope this observation on the behavior of the chipset can help fix the issue. It might not even be tied to PiTool exclusively, it might be tied to the instruction sets, necessary for certain VR titles, or something else. It’s clearly in conjunction with the driver, who contains the sets. So please don’t kneejerk in blaming Pimax right away - they can only access the APIs and work to that gate, not further!
I don’t know why you’re feeling attacked by my posting. Please refrain from such insults as “whiney brats”. I’ve just tried to understand your posting and additionally written that I could be wrong too. Just be less aggressive.
All right kellerbach.
Here is some interesting article on the basics of undervolting btw., but it’s in German, be advised!
Hey, no problem. So you think that there is a protection mechanism that lowers the GPU load in case of too much voltage? I only know the power limit, so if there is too much wattage, the clock will be lowered. I only have a 1080ti and don’t know anything about the new generation.
Honestly, I get better clocks on my 1080TI FE by undervolting because of the heat loads.
I’ve tinkered a lot with my 1080ti. It is even a silly asus turbo which is basically a founders-edition-like card. I have an arctic III on it but with liquid metal which worked even better than the ok-ish MX paste.
Honestly I never found much fps increase by OCing the memory. So I fiddled even with downclocking memory.
I got the sweetspot by lowering the memory by 50 bringing the boost clock of the GPU to a solid and always stable 2060 MHz at a safe 69 degree in order for the heat throttle not to kick in. I know 69 may be considered early but I found that throtteling kicks in at around 74 in my case so I stay on the safe side.
Just wanted to point out that there are many ways to do OC on your GPU, I got there by try and error though. Don’t hate me for this beeing only semi-related 
So I played around with both the power limit and the now unlocked voltage slider in Afterburner (I’m not sure if the voltage slider even does anything but I’m able to move it now, so who knows). I couldn’t get it to do anything different to what normally happens.
I even went back to very old PiTools to see if they ever exhibited the low usage as well. They do, the ones that actually work with the new Pimax firmware. So I’m not sure what else I as an end user am able to do with this information. But I know for a fact that when I get the trick to work I can get a locked 90hz in rFactor 2 when, with the exact same settings, I can only manage 75-80hz with PiTool operating normally, with the ‘bug’.
So how do you explain the fact that once the trick is working, you can select your previous OC profile which will raise the voltage again, yet the trick still works?
I miss the days when overclocking was straightforward