Sorry - but the naming of a number of the profiles means nothing.
You have to be very specific with the naming to enable a 3rd party to make sense of them.
There are three 'post-cal' profiles, one pre-cal, and one called just 'New', as well as two LUTs, one called 'direct-lg, and the other 'New LUT'...

I can kinda guess what is what, but you really need to be much more specific and managed with your testing!

I assume the 'New' profile is via Resole, and if you look at the profile data it shows you have a black patch that has measured way higher than many of the grey scale patches, as well as having RGB primary patches that are totally wrong...

That immediately tells you there is a serious issue with the Resolve signal path, and as per the User Guides, it is likely to be an issue with not enough 'Extra Delay'.
It may even be that Resolve is 'stuttering' on playback, or even freezing.
How powerful is your Resolve PC?

But regardless, all the profile data shows you have issues with Resolve as the TPG.

Steve

I'll work on getting some new files today. I did notice one thing with ColourSpace when starting profiling using Resolve ITG. It would skip very fast from the first (0,0,0) to second (8,8,8) patch sometimes. So I usually stop the profiling, restart it, then it goes slowly and reads the first patch slower, like it should. Since I have "intelligent Int" turned on. This is reproducible almost every time, I have to restart the reading a second time. I've also tried i1D3 Integration times up to 1.0sec.

My laptop and PC are both updated and powerful and fast enough including my 1Gb network connection and router. So that can be ruled out.

I did find this in the LG EP950 Best Practices Guide. It seems there are limitations with RGB 4:4:4 at 4k. After calibration when I profile the monitor for accuracy, I am still in RGB 4:4:4 10-bit mode, and that may be the issue. Because #3 says "RGB is only possible at 8-bit at UHD".

3. The maximum bandwidth of HDMI 2.0 is 18 Gbps. UHD resolution at > 30p with 4:4:4 YCbCr subsampling / RGB is only possible at 8 bits per component within this bandwidth. If 10 (or 12) bits per component is a requirement, a lower resolution, frame-rate, or a different interface must be used; UHD 24p 10 bpc RGB is possible over HDMI 2.0.

4. A maximum bandwidth of 10.2 Gbps on HDMI limits UHD resolution to < 30p, 4:2:0 YCbCr subsampling and 8 bits per component. The use of UHD inputs is not recommended when this limitation is set.

Please see the comment about 'Extra Delay'.
That should be your first stop in assessing what may be wrong.

Steve

ok I think I found the issue and now everything appears better. Probably not perfect for an i1D3 probe with no offset, but better than the native gamut of the display.

Here is my latest results, hope these are renamed better for you to make it easier for you.
I have a Pre, Post, and the LUT files. I used Grey, Primary & Secondary Ramp+. I used Full Range, 10-bit color, had to increase probe integration time to 1.5sec, Period, Intelligent Int = 1.000.

I will have to watch more of your videos to see if it can be improved further.

Attached file: 10-1 calibration files

Things do look a lot better.
But as stated, what is your Extra Delay value?
Did you use Auto to set it correctly?
It is entirely possible the extended Integration time - 1.5 seconds is very long - is just masking the issue.
And 10 bits is totally irrelevant.
It may cause incorrect rounding within Resolve compared to using 8 bits as per the Resolve User Guide.

Steve

I did not use Auto. No one mentioned to me it was absolutely necessary to use "auto". All I know is it is an option to use. I'll try it again today when I get time. This display has been taking too much of my time, where I could be up and running. I've had to switch back to my trusty Eizo CG2420 which calibrates very easily and accurately in the meantime. The LG just plain will not calibrate in 8-bit mode. I know it probably should, but why don't you talk to TED about it, because he specifically says in the LGG post that you must have this LG set to 10-bit and Colourspace set to 10bit before profiling. So who is right out of both of you?

All I know is 10-bit works. 8-bit did not calibrate properly with this display, and I had the whole chain set to 8-bit, with Deep Color OFF. B&H Photo is closed until September 9th. So I have until that time to make a decision to keep it. If FSI actually had reasonably priced 4k displays, I would've bought from them. But at this time, I can't pay $6k+ for a 4k display, since I have no work coming in and just want a good setup to get started learning Resolve more.

You changed multiple things at the same time - and likely making Integration Time so long has masked the real issue.
Extra Delay.
The potential requirement for Extra Delay is a standard setting, and is mentioned in the various User Guides.
And I've mentioned it specifically more than once.

Bit depth is never an issue, unless you are using a TPG that can actually do 10-bit, or greater.
Resolve cannot - something we have repeatedly asked BMD to fix, along with BG colour, but unfortunately they have not been able to find the required development time.

Steve

Okay here are four different runs. 1080p & 2160p in 8bit, and in 10-bit. As you can see, 1080p 8-bit and 10-bit both work fine. Although when I was in 1080p 8bit, the LG still told me it was in 10-bit mode. Which may be a bug in fw 3.08. I was unable to ever get it to show 1080p in 8-bit. But it still gave good results.

On the other hand, the 2160p 8bit and 10bit results are terrible. I double checked mini monitor 4k settings and Resolve, and the HDMI Deep Color and LG settings. All were set correctly.

I know you are going to say that it is still something before the monitor that is in the path. But that is not very helpful to me. The only thing in the Resolve TPG path, is the Decklink Mini Monitor 4k. HDMI out to LG HDMI in. I tried different cables, all certified for actually 8k 120Hz. Perhaps the Mini Monitor has some limitation over HDMI to display 2160p30 10bit? Otherwise what else could it be? I suppose I could go buy a Ultrastudio 4k Mini and try that, to rule out the Mini Monitor 4k.

There are no other settings in Resolve that need to be changed, nor on the LG. I followed the guides exactly for Resolve in Full level. It is late and I'm going to bed here in Texas. I spent all day again trying to get this working in 2160p but could not. Thanks in advance! Have a good night/day.

Attached file: new calibration files

As I have asked a number of times, what is your Extra Delay setting?
I see exactly the same issue I have described before.

You may even need different Extra Delay values for different Resolve settings, such as resolution...?

Steve

The delay for these were 0.50. I tried the Auto option many times and it never suggested any additional delay. It remained zero no matter the resolution.

There is very much a different with the higher resolution - did you run Extra Delay with that?

As per the link provided previously, you can easily see any visual delay on the ColourSpave laptop floating patch window, and the Resolve connected monitor.

Steve

Yes, I did run it yesterday at 2160p/8bit and 10bit. Both showed zero extra delay needed. I just ran it now on both bit depths again. It remains at zero.

I also opened the patch window on my laptop and looked at it , resolve, and the LG patches flash simultaneously. I don't see any delay at all with my eyes. If there is a delay, it is very fast. The white patches flashed up instantly on all 3 displays. I can try and add a manual delay of say 1 second again, like I did in the past and test it out again.

screenshot

If Auto Delay says none is needed, none is needed.
But, there is something that shows a different reading when you have the higher resolutions selected.

It could be that Resolve is having issues maintaining real-time at the higher resolutions, but it is random, and not being 'caught' by the Auto Delay function.

The measured values are 'close enough' the issue is not a total incompatibility, but more subtle - such as random playback delay.

If you run a small Cube patch set (5^3 for example) you will see any such issue easier when Anisometric, not Sequential, is selected.

Swapping between Sequential and Anisometric, with a Cube profile, will also help define such playback issues.

Steve

I added a delay under the "extra delay" of .50 and it seems this worked at least once. I will continue testing when I get home from work tonight. I'll try a shorter value to see if I can use .25

Attached file: test with extra .50 delay