For reference:
https://wiki.jriver.com/index.php/Logging
In summary:
- Turn on Logging in the Help menu
- Reproduce the issue
- Use "Report problem.." in the Help -> Logging dialog to generate a zip package with the log and attach it here or mail to hendrik at jriver
Thanks Hendrik.
I've done the following on my Dell XPS 17, as that's where the 3090 is currently installed, in a Razor Core X E-GPU:
- Disconnect from internet
- Uninstall all nVidia drivers with DDU
- Clean install of the latest studio drivers (537.42) without GeForce experience for the 3060 (E-GPU off)
- Power the E-GPU on
- Install of the same drivers for the 3090
- Internet switched back on
That's the standard procedure to make sure there is no conflict between the 3060 dGPU and the 3090 E-GPU.
The chain for the 3090 is:
3090 HDMI OUT >> Denon X8500HA HDMI 2.1 in >> HDMI 2.1 Out to HDFury VRROOM >> JVC NZ8 (RS 3100) or LG 4K Monitor.
I did lots of tests and it doesn't matter whether I use the LG or the JVCs, when the drop outs happen, they happen with either.
I've also tested without the HDFury VRROOM and it makes no difference re Atmos drop outs as it's after the AVR (so not part of the Audio chain).
From that point, I've changed the following settings in the nVidia Cpanel:
- Vsync on (necessary to get rid of glitches and frame drops with the 3090)
- Power: Prefer max performance (not strictly necessary, but just to err on the side of caution)
- Prefer nVidia GPU
- Reference desktop settings
- 4K23 RGB 8bits (I didn't use 12bits due to the nVidia driver issue, it's present in this configuration even though the e-GPU is connected to the HDMI 2.1 input on the X8500HA).
I used the latest version of JRiver MC 31 for my tests (build 54) and I selected JRVR Red October in the video settings, to rule out madVR and any of the custom settings I normally use. I selected the nVidia Audio device with WASAPI and bitstreaming enabled.
I played the film (John Wick 3, as I used it for my previous tests) from MC itself, I didn't launch it from CMC as I usually do, to rule out another potential (though unlikely) factor).
I used BD Menus because that's what I usually use. I also get the Atmos drop outs with direct play, so not a factor.
I waited for at least a minute after starting the film to make sure the the initial frame drops we get with JRiver are reset, so from minute 1 I had reset the usual initial frame drops and had 0/0 dropped repeated frames. I also checked that only the 3090 was used, which was the case, both the Intel iGPU and the 3060 dGPU had zero use.
I got typical results with the 3090 (it would be similar in the HTPC):
- after about 10 minutes, I get 2 frame drops and 1 repeated frame (visible).
- about one hour in, I get one additional frame drop.
- About one hour and 15 minutes in, I got 20 additional frame drops. It caused a freeze of the picture for almost a second. That's the main Atmos drop out, the one that I would not experience with the 1080ti and exactly the same settings, GPU driver and software. Usually, with madVR, I get more dropped frames than that. The drop out happened during a silent moment, so I saw the picture freeze for almost a second, but I didn't hear the Atmos drop out the way I would have if it had been a busier moment in the soundtrack. In any case, 20 frame drops at once is not normal or even typical.
- About 2 hours in, I got one more drop.
As you can see, there are no regular frame drops. There shouldn't be any as the refresh rate is exactly 23.976, but for some reason I always get a couple, which I can live with as long as there is no big Atmos drop out in the middle of the film.
With the 1080ti, I usually get a max of 2/3 frame drops per film, and I never get an Atmos drop out or a one second freeze. The only time when I get a repeat is with Pacific Rim, chapter 2 (one of my favorite stress tests as it's 1.78) , it's always at the same frame and it's caused by the content as it's entirely repeatable.
At the end of the film, the total was an unacceptable 24/1. I took a screenshot during the end credits (which caused another drop, so ignore that last one which brought the total to 25/1), see below.
As you might remember, when I did my initial tests, I had a few Atmos drop outs in the first 30 minutes of JW3, always at the same place. These do not happen anymore. I assume thanks to the fix you made to MC recently.
When I get an Atmos drop out, it's usually only one, and most often around the one hour mark. It can be a bit earlier (as early as 30 minutes in) or a bit later (as with this test, around 1 hour and 15 minutes). It doesn't always happen exactly at the same spot, unlike the other (corrected) issue.
I've sent you the ZIP with the log files by email, as requested.
Please don't hesitate to ask if you have any questions, and thank you for investigating this.