INTERACT FORUM

More => Old Versions => JRiver Media Center 29 for Windows => Topic started by: jmone on January 25, 2022, 11:52:37 pm

Title: JRVR MC29 Testing [Pre-release / Old]
Post by: jmone on January 25, 2022, 11:52:37 pm
[Edit by JimH -- This thread is from the early development of JRVR in MC29.  It has been superseded by a new thread that began about the time MC29 was first released.]

jmone said:

I've done some initial testing (all HDR Passthrough on Windows 11):

- 3090 / 1660Ti:  No issues with all the "best" settings on material up to UHD HDR 59.94
- NUC (iGPU):  Now this is pretty amazing.  :o  This NUC can also play UHD HDR 59.94fps when "Allow HW decoder direct rendering on mismatched size" is checked, without dropping frames!  ... but I get a thick blue line (and it turns green if windowed).  With "Allow HW decoder direct rendering on mismatched size" unchecked then it can play up to UHD HDR 50fps (and there is no blue or green line), but UHD HDR 59.94fps will drop frames.  I know that subtitles are still a work in progress but every sub displayed results in a heap of dropped frames and the VSYNC jumps all over the place (from say 25 to 125hz).

Hoping to get time to look at the tone mapping stuff on the PJ tonight.

Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 12:36:22 am
The transfer of DV to HDR10 seems to work well and appears not to be too costly (well on my 3090).  I'm using this file from Dolby to test http://media.developer.dolby.com/DolbyVision_Atmos/mp4/P81_GlassBlowing2_3840x2160%4059.94fps_15200kbps_fmp4.mp4

One question, however is I can not seem to "see" the DV info on any of my BD Rips (all to a folder structure) either though Full Menu or Title Playback.  Is DV Supported in either mode?

Thanks
Nathan
Title: Re: JRVR MC29 Testing
Post by: Hendrik on January 26, 2022, 12:55:39 am
... but I get a thick blue line (and it turns green if windowed).

Are you using the bundled LAV Filters? Its supposed to color it black. And I've also never seen it that large, curious.
(Edit: I think I found the reason for the large border, but with LAV 0.76 its still supposed to be black)
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 01:05:30 am
Bundled LAV as part of the fresh MC29 install (with MC28 uninstalled).... and yup, it is not subtle.
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 01:06:45 am
Is there an "easy" way to see what version of LAV it is using?
Title: Re: JRVR MC29 Testing
Post by: Hendrik on January 26, 2022, 01:08:04 am
One question, however is I can not seem to "see" the DV info on any of my BD Rips (all to a folder structure) either though Full Menu or Title Playback.  Is DV Supported in either mode?


Blu-ray DV is not supported, because it's extra special. My primary goal was playing DV Profile 5, because without support it's rendered in false colors, making it unwatchable. Profile 8.1/8.4 should work just fine, but all they do is slightly mess with the image a bit, which may or may not result in visible changes. Profile 7 (Blu-ray) is not supported, because it's Dual Layer and would require combining the two layers first, but maybe in the future (but then again, it also only slightly messes with the image, after learning how it all works, most of DV really is just marketing)
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 01:09:12 am
Also, would it be possible to allow changes to the JRVR settings (like you can do with madVR) from the right click menu when playing?  It would be easier to do comparisons between settings while the video is playing (or do setting changes need the graph to be re-built?).
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 01:10:35 am
Thanks on the DV info.... I'm not really that fussed with the whole proprietary HDR thing that is DV anyway.
Title: Re: JRVR MC29 Testing
Post by: Hendrik on January 26, 2022, 02:34:10 am
Are you using the bundled LAV Filters? Its supposed to color it black. And I've also never seen it that large, curious.
(Edit: I think I found the reason for the large border, but with LAV 0.76 its still supposed to be black)

I would appreciate if someone has a AMD GPU to test this with. It appears Intel doesn't like something about it, which I managed to reproduce, but while I can test easily on a NUC for Intel matters, I don't currently have a AMD system running.
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 02:35:33 am
Only nvidia and Intel here.  Just upgraded the other NUC and same thing.
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 04:02:52 am
I've started testing using Auto Select Tone Mapping, and it looks GREAT, but I'm getting some issues. 
- What is the Default for Auto Select?
- I'm seeing Brightness Pulsing after a scene change
- Is it more expensive to tone map a 4000nit video (UHD 23.976 Fury Road) than 1000nit (UHD 59.94 Gemini Man)?  Fury Road is dropping frames and is almost twice the render time on my 1660Ti than Gemini Man.... or is nvidia doing some power mgt thing?
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 04:30:28 am
Forget the performance question.  A reboot fixed the issue with frame drops on Fury Road.... and render time is now half Gemini Man (which is what I'd expect).  Got to love Windows!

On the Brightness Pulsing, it is related to Peak Detection.  It pulses when it is Checked and Does not pulse when it is unchecked.
Title: Re: JRVR MC29 Testing
Post by: Hendrik on January 26, 2022, 04:36:39 am
If you have delayed peak detection on, turn it off. Otherwise it's just what it does, if the image brightness changes, it compensates, which in extreme changes can be more visible then otherwise.
But with the new tonemapping, peak detection should be needed much less now.
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 05:06:00 am
Delayed Peak Detection was OFF.  With Peak Detection ON (and Delayed Peak Protection OFF) I'd see a double pulse after some scene changes as it seemed to rapidly adjust one way then back again.  Anyway, I'll leave it OFF for now.
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 05:30:15 am
I'm very impressed on what JRVR can do in such a short period of time... striking a great balance of performance and quality.  Who would have thought a lowly iGPU in a NUC could handled HDR Passthrough on UHD 59.94 content, or that a 1660Ti could have all the goodies at Max, Tonemap ... and still have plenty of headroom?!

The other benefit that should not be underestimated is that it "works out of the box" with little need to tweak 100's of combinations/permeations of settings with madVR.  While I'm sure hard core videophiles will still prefer madVR to extract the very best results for a particular setup, the fact is that JRVR is far more efficient, hence runs on a much wider range of HW, and yet manages to produce a great image.   It is an enormous step up from EVR, and for most of us, indistinguishable to madVR without the need to fine tune. 

Well done.  I don't see a reason why JRVR should not be the default for RO. 
Title: Re: JRVR MC29 Testing
Post by: JimH on January 26, 2022, 06:55:11 am
Great news!  Thanks for all the feedback, jmone.
Title: Re: JRVR MC29 Testing
Post by: BryanC on January 26, 2022, 12:02:57 pm
I'm still getting some occasional frame drops on high-bitrate HDR passthrough on my Vega 8 w/ all quality settings off, but it has improved in MC29. Whereas I may have had a frame drop every 15-20 seconds on a certain file, now it's every 30-45 seconds. So that's a big improvement and I don't see any obvious side-effects either.

Performance: 5.8 ms/ 7.8ms
direct sampling x2 is enabled

Looking forward to testing out the N5105 to see if this is vendor/hardware related, should be here Feb. 3.
Title: Re: JRVR MC29 Testing
Post by: Hendrik on January 26, 2022, 01:52:16 pm
I'm still getting some occasional frame drops on high-bitrate HDR passthrough on my Vega 8 w/ all quality settings off

"quality settings off" means what exactly? :) In trade quality for performance, you should have "Allow direct rendering on mismatched size" ON, as that saves quite a bunch of memory bandwidth.
There are a few bugs with that option currently (non-performance related), but we'll get those squared away.
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 03:51:05 pm
Once the bugs in "Allow direct rendering on mismatched size" is sorted, is there any real downside to having it on even on more powerful GPUs?
Title: Re: JRVR MC29 Testing
Post by: Hendrik on January 26, 2022, 04:09:55 pm
Quote
Once the bugs in "Allow direct rendering on mismatched size" is sorted, is there any real downside to having it on even on more powerful GPUs?
There might be very minimal image changes due to it, which might result in a slightly darkened row of pixels at the bottom of the screen, but it's on by default and likely going to stay that way.
This would be entirely invisible if you have any black bars of course, and even on fullscreen content its an extremely minimal effect.
Title: Re: JRVR MC29 Testing
Post by: BryanC on January 26, 2022, 05:10:08 pm
"quality settings off" means what exactly? :) In trade quality for performance, you should have "Allow direct rendering on mismatched size" ON, as that saves quite a bunch of memory bandwidth.
There is a few bugs with that option currently (non-performance related), but we'll get those squared away.

Sorry for the confusion, for my low anchor both performance options are enabled w/ bilinear scaling, sigmoidal light off, dithering off (basically the barebone minimum of JRVR). For high anchor I use Lanczos 3 taps, sigmoidal light on, blue noise dithering and both performance options ON (since I don't tonemap and the new feature is tailored for iGPUs) as low bit-rate 2160p and 1080p content plays fine on the high anchor but nothing more intensive than that.
Title: Re: JRVR MC29 Testing
Post by: jmone on January 26, 2022, 06:01:09 pm
There might be very minimal image changes due to it, which might result in a slightly darkened row of pixels at the bottom of the screen, but it's on by default and likely going to stay that way.
This would be entirely invisible if you have any black bars of course, and even on fullscreen content its an extremely minimal effect.

Thanks - I'll turn it on for all my setups as it gives a good performance bump
Title: Re: JRVR MC29 Testing
Post by: jmone on January 27, 2022, 04:05:40 pm
Quote
29.0.2 (1/27/2022)
3. NEW: Added an option to JRVR for toggling reporting the selected primaries/gamut to the display.
6. Fixed: Enabling direct mismatched size rendering in JRVR could result in a larger border then expected.
7. Changed: Disabled JRVR Peak Detection by default due to its performance cost and potential for brightness shifts, and moved the option into the Advanced section.

#6.  Tested on the NUC and it is fixed and looking good.  Still can't believe I don't have to replace these old NUCs anymore.  Saved a bunch of $ and Time!
#7.  I'll leave at off as suggested

#3.  Not sure about this one, or how it functions to meaningfully test?  Does it only impact when Tonemapping is on, or also do something with Passthrough?  Also, could the Ctrl+J OSD display this info on what is being sent?
Title: Re: JRVR MC29 Testing
Post by: jmone on January 27, 2022, 04:14:30 pm
On the feature request side for JRVR, most of what I'd like for HDR is looking good as all my screens are 16:9 (so I'm not a candidate for some of the dedicated Theater Room requests).  There are a couple of things that may be worth considering on the SDR front:
- HW Deinterlacing:  As well as legacy video, we still have interlaced DVB-T, and HW Deinterlacing would be a noticeable improvement.  In my testing (deprecated) CUVID is still unrivalled on this sort of content. 
- SDR to HDR Tonemapping:  This may fall into the law of diminishing returns given how Win11 does this in the background, but a JRVR SDR to HDR may be worth looking at as the Win11 version gets only one adjustment (brightness?). 
- Subtitle Improvement:  I know you are all over this one!
Title: Re: JRVR MC29 Testing
Post by: Hendrik on January 27, 2022, 04:36:31 pm
- HW Deinterlacing:  As well as legacy video, we still have interlaced DVB-T, and HW Deinterlacing would be a noticeable improvement.

I'm not sure what you are asking for. TV or file playback all gets the same deinterlacing, as long as its using Red October for playback.

There are no plans for SDR to HDR at this time.
Title: Re: JRVR MC29 Testing
Post by: jmone on January 27, 2022, 06:13:34 pm
No probs on the SDR to HDR, just an idea.

On the De-Interlacing, isn't JRVR using SW (I presume BWDIF) rather than HW (say as with CUVID)?
Title: Re: JRVR MC29 Testing
Post by: jmone on January 28, 2022, 12:21:29 am
Another random ask.  Is HEVC 10-Bit 4:4:4 HW acceleration support somewhere on the roadmap (not a must have now)? 

Probs only of interest to me but I shoot a bunch of AVC UHD 4:2:2 10-Bit 500-950Mbit content that I currently then encode at the same sort of spec as UHD BD (80mbps HEVC UHD 4:2:0 HDR), but it irks to throw away the extra chroma info to have to recreate it later with a scaler during playback.  Davinci Resolve has a nvidia profile for HEVC 10-Bit 4:4:4 HDR (uses HW acceleration) which plays fine (and looks great) in MC but hammers even a top end CPU whereas the GPU is just twiddling it's thumbs.
Title: Re: JRVR MC29 Testing
Post by: Hendrik on January 28, 2022, 12:35:42 am
On the De-Interlacing, isn't JRVR using SW (I presume BWDIF) rather than HW (say as with CUVID)?

On Windows it uses HW deinterlacing through D3D11, at least as long as its running through Red October.
Title: Re: JRVR MC29 Testing
Post by: jmone on January 28, 2022, 01:14:19 am
Perfect!
Title: Re: JRVR MC29 Testing
Post by: jmone on January 28, 2022, 01:18:26 am
...and Night Time is coming, any Tonemapping testing you want done on a PJ?
Title: Re: JRVR MC29 Testing
Post by: BryanC on January 31, 2022, 05:00:02 pm
I received the itty bitty ZX01 N5105 (10W) mini PC (JRMark 2705) w/ Intel UHD graphics:

(https://i.ibb.co/tZy4xSC/IMG-20220131-134814.jpg) (https://ibb.co/SQqPd2v)

(https://i.ibb.co/VVXFhLQ/IMG-20220131-134829.jpg) (https://ibb.co/cJzWpyC)

(https://i.ibb.co/FntD96g/IMG-20220131-134821.jpg) (https://ibb.co/N7zyXxZ)

I should've added something for scale but it's less than 3" on every side, about the size of a baseball. For some reason I thought this thing would be fanless but it's not, although the fan is very quiet (quieter than any laptop or mini PC I've owned, even quieter than the NanoPi R2S mini router). It needs 2.5A 12V USB-C for power (so ~30W headroom) and includes a 3A 12V USB-C AC adapter and HDMI cable. Surprisingly it has W11 Pro pre-installed as well.

It plays everything in my library using bilinear scaling, no frame drops even w/ HDR passthrough! So it appears that the problems with JRVR and HDR passthrough are relegated to integrated AMD devices. This thing doesn't have quite the GPU horsepower as the Vega 8 (so I can't run Lanczos on non-HDR content) but all I care about is no frame drops, so I'm happy. Time to sell my AMD stuff and enjoy some movies, finally!

With direct rendering enabled I do see a fairly prominent blue line along the bottom of the display so I have turned it off for now.
Title: Re: JRVR MC29 Testing
Post by: Hendrik on January 31, 2022, 05:02:28 pm
The remaining blue line will be fixed in an upcoming update, but it requires an update to external components, so it might be a few days.

Glad to hear i'ts performing well. In the long term we plan on changes/improvements to the frame presentation to allow pre-rendering, which might help to manage the performance expectations of such AMD devices as well, but its hard to predict how it'll impact any system, and of course I also don't have access to every combination to run tests - and AMD hardware in particular is pretty rare in my collection, unfortunately. (Mostly because their video history isn't stellar)
Title: Re: JRVR MC29 Testing
Post by: JimH on January 31, 2022, 05:02:59 pm
Nice!  Thanks for the report and the pictures!
Title: Re: JRVR MC29 Testing
Post by: JimH on January 31, 2022, 05:22:43 pm
Is this where you got it?
https://www.aliexpress.com/i/1005001921619318.html

How was it configured?

I may move this to the hardware board.
Title: Re: JRVR MC29 Testing
Post by: BryanC on January 31, 2022, 05:39:56 pm
Is this where you got it?
https://www.aliexpress.com/i/1005001921619318.html

How was it configured?

I may move this to the hardware board.

Yep, that's the store. I went with 8GB/128GB as my HTPC is basically a "thin" MC client. If I were recording TV on the client I would go with a bigger disk (only ~90GB free after updating Windows, installing Intel driver packages, and running Disk Cleanup) but for my purposes it's more than enough as I record on the server.
Title: Re: JRVR MC29 Testing
Post by: jmone on January 31, 2022, 05:59:40 pm
The is a short write up of these n5105 based mini's -  https://liliputing.com/2022/01/this-2-8-inch-mini-pc-has-an-intel-celeron-n5105-processor-dual-hdmi-ports.html  Much better than the older J4125 based one.  I'm also waiting to see what the new NUC 12's will bring but they will be much more expensive and JRVR seems to be working well on these less powerful chipsets.
Title: Re: JRVR MC29 Testing
Post by: jmone on February 02, 2022, 03:00:58 pm
Looks like Intel has not given up on low end/priced NUCs, https://www.fanlesstech.com/2022/01/jasper-lake-nuc-official.html  with a N5105 barebones retailing local to me for A$250 (no stock).  Also, no IR like other NUCs (if that matters to you).
Title: Re: JRVR MC29 Testing
Post by: JimH on February 02, 2022, 03:40:05 pm
At Intel:  https://ark.intel.com/content/www/us/en/ark/products/217668/intel-nuc-11-essential-kit-nuc11atkpe.html
Title: Re: JRVR MC29 Testing
Post by: jmone on February 09, 2022, 06:23:25 pm
I started testing the new Subtitle Engine, and it looks good and all the subtitle controls work as expected.  I'm getting dropped frames however when subtitles are on.  I see there is a rendering time increase from 6ish to 8ish ms on my 3090 (Gemini Man UHD 59.94fps), but this should not cause it.

Do you want any frame logs etc?
Title: Re: JRVR MC29 Testing
Post by: jmone on February 09, 2022, 07:24:08 pm
Problem solved.  Had to reset
- nvidia control panel --> manage 3D Settings --> Vertical Sync --> On
- nvidia control panel --> manage 3D Settings --> Power Management Mode --> Prefer Maximum Performance
Title: Re: JRVR MC29 Testing
Post by: jmone on February 09, 2022, 10:12:43 pm
Did some performance testing on the NUC.  Subtitles adds about 5ms to the (max) render queue times and hence tips the NUC over the edge with anything above 23.976 (eg older 1080/50i Tv shows).  It's fine on Std BD's however which is great.
Title: Re: JRVR MC29 Testing
Post by: Hendrik on February 10, 2022, 02:55:20 am
Judging from previous comments on subtitles, thats already a lot better then it used to be, correct?

The topic is not done yet, there are important optimizations left to do, the work already done moved subtitles to a different thread and added a lot of flexibility to how its managed to allow us doing the coming optimizations.
Title: Re: JRVR MC29 Testing
Post by: jmone on February 10, 2022, 03:10:05 am
Correct.  Subs are better, don't drop frames and just work.... well, now that I (re)fixed the NVidia driver settings.  In fact on my 3090 UHD 59.94 with all the goodies ON + Subs check out the rendering times!

I should get to check a 1660Ti with HDR Tonemapping shortly. 

The lowly NUC is just what it is.  I'm still surprised how well it does, but there are limits.
Title: Re: JRVR MC29 Testing
Post by: jmone on February 10, 2022, 04:12:34 am
I did have one issue with the 1660Ti in that subtitles were not being displayed and when checking the right click menu no option was selected (not even Off).  While I could select a sub stream, nothing was being displayed and on checking, again nothing was ticked.  I've no idea what caused this and could not repeat it as a closed and re-started MC and it was then all fine.

Once that was sorted, there was no issues to report.  It all played well with Gemini Man rendering in 5-6ms on the 1660Ti with all the goodies ON + Tonemapping + Subs.  The only frame drops I saw was the odd one from not being able to use VideoClock as this setup bitstreams audio for 7.4.2ch).  An old request, but it would be great if one day we can detect Atmos/DTS-X and have the option to just bitstream these formats (but then again, more and more material has these anyway).

The only thing that stressed the 1660Ti into dropping frames was 1080 60/50i with FSRCNNX 16.  I had to drop it to FSRCNNX 8 for this material (and good use case for Profiles).
Title: Re: JRVR MC29 Testing
Post by: Hendrik on February 10, 2022, 11:11:16 am
The next version will have another subtitle improvement - instead of rendering a fullscreen overlay, only small rectangles will now be created, which drastically reduces the amount of data being handled.

PS:
May not apply to certain bitmap subtitles, as they can span the entire screen with transparency, which we cannot easily detect without defeating any performance gains. Varies on a disc by disc basis.
Future improvements will instead let the GPU do the scaling of the bitmap subtitles, instead of doing it in software, further improving performance (at least on high resolution).

PPS:
Same improvement for ASS subtitles coming a build after.
Title: Re: JRVR MC29 Testing
Post by: jmone on February 10, 2022, 01:56:28 pm
Nice!
Title: Re: JRVR MC29 Testing
Post by: jmone on February 10, 2022, 02:35:19 pm
Out of interest, is there anyway to "tell" if the Subtitles span an entire screen or not (most of my subs are BD PGS)?
Title: Re: JRVR MC29 Testing
Post by: Hendrik on February 10, 2022, 03:15:59 pm
Not really, but luckily most discs don't. But I have observed it on a bunch. There are also weird cases sometimes, eg. on the Avatar BD, the normal subtitles are small section, but the forced alien subtitles are covering the full frame.

The other optimization coming will help quite a bit in either case though if you are playing on 4K, because PGS subs are always 1080p, even on UHD BDs, as far as I know, so if we don't upscale them anymore before the video renderer, its quarter of the size to handle and send to the GPU.
While I'm in there, I noticed we don't treat UHD BD subs as HDR10 like it should be, so thats something I need to address somehow.
Title: Re: JRVR MC29 Testing
Post by: jmone on February 10, 2022, 03:28:31 pm
Ahhh, nothing like a consistent approach to mastering!
Title: Re: JRVR MC29 Testing
Post by: jmone on February 11, 2022, 04:21:58 pm
Quote
1. Fixed: Subtitles in JRVR could stop working after seeking backwards.
2. Changed: Subtitles in JRVR are now handled in small rectangles, instead of the full frame, for faster rendering.

#1 - Explains the issues I had on the 1660Ti as I was seeking backwards to get to the same point for testing.  Seemed fixed
#2 - Well, you have done it again.  Tested on the NUC7i5 with HDR Passthrough and White Noise (Balanced is right on the edge)..... It now plays both Gemini Man and 1080/50i material without dropping frames even though the "Max" render time spikes above the required interval when displaying subs.  Did you also change the buffers by any chance? 

Anyway check out these screen shots of the NUC and the 3090 (which now apparently is so bored it can render frames in 1ms with all the goodies on)!
Title: Re: JRVR MC29 Testing
Post by: jmone on February 11, 2022, 04:32:32 pm
...also is there a simple way to revert JRVR settings back to "Default"?
Title: Re: JRVR MC29 Testing
Post by: Hendrik on February 11, 2022, 05:09:24 pm
And bitmap subtitles like on Blu-ray are going to get even faster, too (at least if you play them above 1080p screen resolution)
Also the second overlay entry is from the OSD, so it inflates your summary max as well on the NUC.
Title: Re: JRVR MC29 Testing
Post by: jmone on February 11, 2022, 05:40:28 pm
Right, so even better performance when the OSD is off :) nice...  I'll need to retest the NUC with Balanced.  It would be good to have something like Ctrl+R to reset the stats to see what frames are being dropped/repeated after the startup.... or maybe a version of the Log that only writes out the dropped frames?

Apart from the refinement of performance (which is always good, esp on the low end), are there any development features on your to do list in the short term?  I know of:
- Profiles
- cross platform stuff
- ?

Title: Re: JRVR MC29 Testing
Post by: jmone on February 11, 2022, 06:31:15 pm
Retested on the the NUC with Blue Noise (Recommended) and it was fine even with the OSD up.
Title: Re: JRVR MC29 Testing
Post by: JimH on February 11, 2022, 06:38:43 pm
- Profiles
- cross platform stuff
- ?
Cross platform and subtitle toys first.

Does madVR use ctrl-R?  We should probably conform to what they do, when we can.
Title: Re: JRVR MC29 Testing
Post by: jmone on February 13, 2022, 03:15:59 am
Yes, madVR uses Ctrl+R to reset.  It's good to have the option for a couple of reasons:
1: You can eliminate the start-up Dropped / Repeated frames after playback settles down so you can then see if there is issues during "normal" playback
2: It looks good in screen shots posted to the forum!  Odd reason I know but .... perception's matters and I hate posting screen shots showing JRVR frame drops that are not occurring post start up.
Title: Re: JRVR MC29 Testing
Post by: jmone on February 15, 2022, 11:25:10 pm
V10 Testing

Quote
2. Changed: Bitmap subtitles in JRVR will now be rendered at their original resolution and resized by JRVR (faster!).

Did a quick test on a "Std" BD ('71) I've been using for testing on the NUC7.  The sub seems to be about the same (at least on this disc), though the overall rendering time looks better... So I'm not sure how to interpret the OSD info for this one.


Title: Re: JRVR MC29 Testing
Post by: Hendrik on February 16, 2022, 01:19:01 am
The biggest impact .10 has is probably a reduction in CPU usage, which will help overall playback stability.
Title: Re: JRVR MC29 Testing
Post by: jmone on February 16, 2022, 01:33:53 am
Gotcha
Title: Re: JRVR MC29 Testing
Post by: jmone on February 22, 2022, 09:45:58 pm
I think I saw this on a list somewhere, but JRVR does not (currently) honour the Rotation Tag so video played back is the "wrong way around" (in this case 180 degrees out).  madVR is fine.
Title: Re: JRVR MC29 Testing
Post by: Hendrik on February 23, 2022, 02:37:08 am
Rotation is in for the next batch of features.
Title: Re: JRVR MC29 Testing
Post by: jmone on February 23, 2022, 04:10:31 am
Thanks
Title: Re: JRVR MC29 Testing
Post by: Hendrik on March 01, 2022, 09:42:30 am
With direct rendering enabled I do see a fairly prominent blue line along the bottom of the display so I have turned it off for now.

The blue line should hopefully be gone now with 29.0.14
Title: Re: JRVR MC29 Testing
Post by: jmone on March 02, 2022, 02:43:28 am
Rotation is in for the next batch of features.

29.0.14 - Rotated video is playing correctly for me!


...not tested the changes for Intel GPU
Title: Re: JRVR MC29 Testing
Post by: BryanC on March 03, 2022, 02:55:56 pm
I went back and retested my content on the N5105 with the "Allow hardware decoder direct rendering" enabled now that the blue line is gone.

I can play back all of my <30fps media (720p, 1080p, 2160p w/ HDR passthrough) using Lanczos 3-taps upscaling, pretty incredible! That's basically ROHQ quality on a 10W CPU ;D.

It starts bottlenecking on 50fps European content, but I only have a few of those files so I'm just sticking with Lanczos 3 taps unless there are any additional crazy performance improvements on the way. I can get away with 4 taps on native 4K content but it drops frames on 720p 30fps so not worth the headache of switching back and forth until performance profiles are ready.
Title: Re: JRVR MC29 Testing
Post by: Hendrik on March 03, 2022, 03:46:16 pm
An interesting tidbit: On some GPUs Jinc can actually be faster then Lanczos, because Jinc is hyper-optimized to run in a single-pass and uses compute shaders, and Lanczos needs to do horizontal and vertical separately.