Could you give a synopsis of what the *problem* is since you did title your post as if there is a 'problem'. I tried to watch but too much er babbling throughout to make any sense of it.

I typically get almost the same results no matter what I do. This is the worst overclocking scores I have ever seen.

When I play with overclocking my 1080 TI and even my 1080 SC, I noticed a very low ceiling on where the card would just max out and head south.  On my 1080 TI, it will go as far as 2050 whether overclocked +140/+150/+160, etc.  The harder I push it the lower the scores become.  I found the sweet spot at +145 and my card will max at 2066.  On my older 1080SC, I was seeing a roll off at 2088.  I always set the power limit and temp target to the max, even when not overclocking and find more stable fps, etc.   What would be interesting is if either EVGA or Unwinder could build in a core counter into the monitoring of the software.  That might answer a few questions.

Great video, the more I learn about overclocking this thing the it reminds me how the controls in the Airbus work as opposed to a Boeing, you move the stick and the computer moves the controls, sliding all the setting around this card yields similar results. 

HeavyHemi
Could you give a synopsis of what the *problem* is since you did title your post as if there is a 'problem'. I tried to watch but too much er babbling throughout to make any sense of it.

The title of the video is "The Pasal Problem" and the title is a quote as such. Basically, Sajin didnt title the post, he shared the title of the video.

He states that he has modified it in the past and was testing it at the beginning of the video. My guess is, he has other videos that you can watch and see what he modifies, since he is one of the most referenced people in the videos from Gamer Nexus and such, as they talk about where to solder on caps and where the mods need to be placed to get control of the voltage.  I wouldn't want another hour of video attached to this one to explain what is available in other videos.  30 minutes is plenty, especially when he is running live benchmark on the card and trying to figure out what is causing the clock fluctuation.  
 
I would think that this would require a custom bios to get around anything NVidia has done, and since the test is done with just the voltage modification, then the stock bios wouldn't allow users to do much after voiding their warranty.
 
Also, removed the second post.  I found that the only time it double posts is when it doesn't register the first submission, and then you click it again and then it doubles. 
 
 
PCB breakdown of the card he uses in "The Pascal Problem"  I don't think he modifies it in this video, just talks about the PCB.

His hair makes it hard for me to take him seriously.

Way too much rambling in his video...So if I'm getting this right, just don't undervolt the GPU and you should be fine? What's the problem then? Is it not apparent that dropping the voltage to 0.6V would obviously cause the card to pull some tricks behind the scenes to keep itself from crashing from lack of power? It seems completely obvious to me that the performance at 0.6V regardless of core clock would be worse than at stock or higher voltage...So I fail to see the issue he's pointing out.

mike406
Way too much rambling in his video...So if I'm getting this right, just don't undervolt the GPU and you should be fine? What's the problem then? Is it not apparent that dropping the voltage to 0.6V would obviously cause the card to pull some tricks behind the scenes to keep itself from crashing from lack of power? It seems completely obvious to me that the performance at 0.6V regardless of core clock would be worse than at stock or higher voltage...So I fail to see the issue he's pointing out.

The point was, with overvolting, it is just as bad. The overclock drops down and the scores dont improve.

mike406
His hair makes it hard for me to take him seriously.

yea i know :) lol

I'm not sure what kind of problem this really is. It's probably a sanity check of some sort. If the chips gets too many bad calculations from cores it probably disabled them. Same problem with under volting, the chip doesn't get enough power for the cores to complete the calculations so the chip shuts them down or doesn't use them. Either way it lowers the score.
His argument makes no sense in technical terms. Even if a sensor in the chip could tell the incoming voltage, there's no good to reduce it. I'd say it resembles Intel's system for monitoring the state of the chip. Intel can selectively shut down parts of the chip or selectively send more power to them if that's needed.