I already have an older version on Github, will update to latest eventually.
OTOH such dl links have it’s advantages, which also include a simple but very efficient version history. Don’t want to cancel this atm…
2 Likes
I’ve been using the -fast variant for a couple days now. (It looks identical to the normal preset for me, and it’s also faster than geom-deluxe.) It’s pretty good! The -ntsc variant gives a nice analog look to the image. But it’s rather heavy on GPU. Any plans on having -ntsc-fast and -ntsc-fastest preset to complement the normal ones?
3 Likes
Hey there! 
Fortunately NTSC Fast version is not this hard to implement, it can be also very nice for high resolution content. ‘Fastest’ version uses some tricks which makes it impossible to merge with ntsc shaders. The speed gain is quite good. The way this new ntsc version works is also a bit different, but should be fine.
As always, feedback is welcome…
New Release Version (2022-01-14-r1):
Notable changes:
- ntsc - fast version added
- new bloom variation (as negative value of bloom)
- some small fixes and changes
Download link:
https://mega.nz/file/ooYEDTiR#PA2PYpxTmYvMQW2ByqTJFWq2gDFXhryr8ZO6dQ_f3MY
Edit:
If somebody likes a ‘messy’ preset: (put where preset’s are saved) 
shaders = "12"
shader0 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/jinc2-dedither.slang"
filter_linear0 = "false"
wrap_mode0 = "clamp_to_border"
mipmap_input0 = "false"
alias0 = ""
float_framebuffer0 = "false"
srgb_framebuffer0 = "false"
scale_type_x0 = "source"
scale_x0 = "1.000000"
scale_type_y0 = "source"
scale_y0 = "1.000000"
shader1 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/stock.slang"
filter_linear1 = "false"
wrap_mode1 = "clamp_to_border"
mipmap_input1 = "false"
alias1 = "StockPass"
float_framebuffer1 = "false"
srgb_framebuffer1 = "false"
scale_type_x1 = "source"
scale_x1 = "1.000000"
scale_type_y1 = "source"
scale_y1 = "1.000000"
shader2 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/pre-shaders.slang"
filter_linear2 = "true"
wrap_mode2 = "clamp_to_border"
mipmap_input2 = "false"
alias2 = "PrePass"
float_framebuffer2 = "false"
srgb_framebuffer2 = "false"
scale_type_x2 = "source"
scale_x2 = "1.000000"
scale_type_y2 = "source"
scale_y2 = "1.000000"
shader3 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/ntsc/ntsc-pass1.slang"
filter_linear3 = "false"
wrap_mode3 = "clamp_to_border"
frame_count_mod3 = "2"
mipmap_input3 = "false"
alias3 = ""
float_framebuffer3 = "true"
srgb_framebuffer3 = "false"
scale_type_x3 = "source"
scale_x3 = "4.000000"
scale_type_y3 = "source"
scale_y3 = "1.000000"
shader4 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/ntsc/ntsc-pass2.slang"
filter_linear4 = "true"
wrap_mode4 = "clamp_to_border"
mipmap_input4 = "false"
alias4 = ""
float_framebuffer4 = "false"
srgb_framebuffer4 = "false"
scale_type_x4 = "source"
scale_x4 = "0.500000"
scale_type_y4 = "source"
scale_y4 = "1.000000"
shader5 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/convert-ntsc.slang"
filter_linear5 = "true"
wrap_mode5 = "clamp_to_border"
mipmap_input5 = "false"
alias5 = "PrePass1"
float_framebuffer5 = "false"
srgb_framebuffer5 = "false"
scale_type_x5 = "source"
scale_x5 = "0.500000"
scale_type_y5 = "source"
scale_y5 = "1.000000"
shader6 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/linearize-ntsc.slang"
filter_linear6 = "true"
wrap_mode6 = "clamp_to_border"
mipmap_input6 = "false"
alias6 = "LinearizePass"
float_framebuffer6 = "true"
srgb_framebuffer6 = "false"
scale_type_x6 = "source"
scale_x6 = "1.000000"
scale_type_y6 = "source"
scale_y6 = "1.000000"
shader7 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/crt-guest-advanced-ntsc-pass1.slang"
filter_linear7 = "true"
wrap_mode7 = "clamp_to_border"
mipmap_input7 = "false"
alias7 = "Pass1"
float_framebuffer7 = "true"
srgb_framebuffer7 = "false"
scale_type_x7 = "viewport"
scale_x7 = "1.000000"
scale_type_y7 = "source"
scale_y7 = "1.000000"
shader8 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/bloom_horizontal.slang"
filter_linear8 = "true"
wrap_mode8 = "clamp_to_border"
mipmap_input8 = "false"
alias8 = ""
float_framebuffer8 = "true"
srgb_framebuffer8 = "false"
scale_type_x8 = "absolute"
scale_x8 = "640"
scale_type_y8 = "source"
scale_y8 = "1.000000"
shader9 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/bloom_vertical.slang"
filter_linear9 = "true"
wrap_mode9 = "clamp_to_border"
mipmap_input9 = "false"
alias9 = "BloomPass"
float_framebuffer9 = "true"
srgb_framebuffer9 = "false"
scale_type_x9 = "source"
scale_x9 = "1.000000"
scale_type_y9 = "absolute"
scale_y9 = "480"
shader10 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/crt-guest-advanced-ntsc-pass2.slang"
filter_linear10 = "true"
wrap_mode10 = "clamp_to_border"
mipmap_input10 = "false"
alias10 = ""
float_framebuffer10 = "false"
srgb_framebuffer10 = "false"
scale_type_x10 = "viewport"
scale_x10 = "1.000000"
scale_type_y10 = "viewport"
scale_y10 = "1.000000"
shader11 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/deconvergence-f.slang"
filter_linear11 = "true"
wrap_mode11 = "clamp_to_border"
mipmap_input11 = "false"
alias11 = ""
float_framebuffer11 = "false"
srgb_framebuffer11 = "false"
scale_type_x11 = "viewport"
scale_x11 = "1.000000"
scale_type_y11 = "viewport"
scale_y11 = "1.000000"
quality = "2.000000"
prescalex = "3.250000"
warpX = "0.040000"
warpY = "0.050000"
csize = "0.050000"
bloom = "-1.100000"
gamma_c = "1.200000"
shadowMask = "3.000000"
maskDark = "0.300000"
maskLight = "1.200000"
slotmask = "0.500000"
slotmask1 = "0.700000"
slotwidth = "1.000000"
slotms = "2.000000"
deconrr = "1.500000"
deconrb = "-1.500000"
deconrry = "1.500000"
deconrby = "-1.500000"
decons = "-1.000000"
textures = "SamplerLUT1;SamplerLUT2;SamplerLUT3;SamplerLUT4"
SamplerLUT1 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/lut/trinitron-lut.png"
SamplerLUT1_linear = "true"
SamplerLUT1_wrap_mode = "clamp_to_border"
SamplerLUT1_mipmap = "false"
SamplerLUT2 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/lut/inv-trinitron-lut.png"
SamplerLUT2_linear = "true"
SamplerLUT2_wrap_mode = "clamp_to_border"
SamplerLUT2_mipmap = "false"
SamplerLUT3 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/lut/nec-lut.png"
SamplerLUT3_linear = "true"
SamplerLUT3_wrap_mode = "clamp_to_border"
SamplerLUT3_mipmap = "false"
SamplerLUT4 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/lut/ntsc-lut.png"
SamplerLUT4_linear = "true"
SamplerLUT4_wrap_mode = "clamp_to_border"
SamplerLUT4_mipmap = "false"
7 Likes
Thanks! Performance is about doubled with ntsc-fast. It does look quite different though. Is the NTSC implementation also altered? crt-guest-advanced.slangp and crt-guest-advanced-fast.slangp look identical here. After I tweaked the normal preset to my liking:
#reference "../../shaders/crt-guest-advanced/crt-guest-advanced.slangp"
interm = "0.000000"
bloom = "0.200000"
brightboost1 = "1.550000"
beam_max = "0.800000
Simply swapping #reference to crt-guest-advanced-fast.slangp produces the exact same image here.
This is not true with -ntsc vs ntsc-fast. It seems slightly blurrier and washes out shadow details. My naive mind would think that if crt-guest-advanced and crt-guest-advanced-fast look identical, then adding the same NTSC implementation to them should also produce the same image.
I suppose it’s not possible to have the same quality NTSC implementation but with just the CRT shaders being the faster implementation?
Looks like I found my Atari 2600 preset 
2 Likes
A notable difference may happen because the first shader in the ntsc-fast chain is the jinc-dedither, which also allows sharper settings and still resolves dithering patterns. I included it because it does a very good job there while still impacting the rest of the image in a very modest way.
Perhaps replacing this pass with stock.slang will bring the results closer.
Second notable change in the default ntsc blend mode. While with normal version it’s 1.0, the fast goes with a sharper 0.0.
Lastly the fast version lacks some of the slower code, like scanline gamma, blend mode 2.0…but otherwise it’s still very versatile.
3 Likes
I believe this is not necessary. One of the main reasons I use NTSC shaders is because they get rid of dithering. A special de-dither shader seems only useful when not using an NTSC shader.
Do you think it’s a good idea to have the full quality NTSC in the ntsc-fast preset, but just based on top of -fast.slangp, and then have the current ntsc-fast preset renamed to ntsc-fastest?
Edit:
I tested performance of crt-guest-advanced-ntsc.slangp between the three NTSC blend modes 0, 1 and 2. They all perform identical for me (121FPS). Same for ntsc-fast.slangp, where modes 0 and 1 also perform exactly the same (230FPS.)
Edit 2:
Indeed. I did that, and at least with blend mode 0, the results are almost identical, except for a minor color temperature difference.
The current preset:
shader0 = shaders/guest/fast/jinc2-dedither.slang
filter_linear0 = false
scale_type0 = source
scale0 = 1.0
Gives this:
Replacing it with:
shader0 = shaders/guest/fast/stock.slang
filter_linear0 = false
scale_type0 = source
scale0 = 1.0
Gives this:
Almost identical, except for the slight color shift.
2 Likes
Oh, btw, nothing changes if I completely remove the first stock pass. What’s the reason of having two stock passes?
1 Like
This can be a bit tricky if certain blending modes are used or ntsc resolution scaling is increased. Some games don’t use dithering, but i guess many users still like a touch of artifacting and fringing while looking at a sharper image. The default ntsc effect looks quite soft though. Having both options is a nice bonus.
This can depend on the compiler too, since blending mode 2.0 brings lots of vec3 pow calculations at viewport resolution. The code paths are if/else separated, in many cases the slowdown can be general though.
Nevertheless, thanks for the feedback, i think i can still improve some features/issues.
Uhh, for simpler adding of additional shader passes, like xbr(z), grade, dedither… Standard presets need at least one stock pass, or custom pre-passes are virtually impossible because the pre-shader must access frame history buffers, which are immune to changes.
1 Like
New Release Version (2022-01-15-r1):
Notable changes:
- NTSC - fast version changes only
- jinc2 dedithering is now a ‘parameter’ option
- basic sharpening added
- NTSC Blend Mode 2.0 added
- Edit: small but notable fix for ‘lut colors’ with fast versions
Download link:
https://mega.nz/file/pgRwBB5S#Q6ZwWY4JK0MV57xcJjLCrAJVlmKOyQqZb2lGKvPKuX8
9 Likes
Thanks. Especially for ntsc blend mode 2 in the fast preset! Looks great, especially in Sega MD/Genesis games.
Btw, disabling Jinc2 dedithering in the parameters doesn’t give any perf benefit. Deleting the shader pass from the preset does, however. Not a big one though, just 1% difference (like 336FPS vs 332FPS,) so it doesn’t even really matter. Perhaps this is due to my weird setup though. I don’t have a weak GPU, it’s just that I have locked the core and memory clocks of my 980 Ti to very low values in order to keep it passively cooled
Great work on this shader though! ntsc-fast looks even better than presets/crt-geom-deluxe-ntsc-adaptive.slangp (which I’ve been using for a while now) but is faster.
Is there any documentation for the options somewhere? For example Display Gamut, CRT Profile, LUT Colors, LUT Size, and Sega Brightness Fix.
Also some of the options have more possible values than the label indicates. Like CRT Profile and CRT Mask.
1 Like
I’m still working/slacking on a new ‘readme’ for the new versions, but the older description describes some color options quite nicely:
Otherwise some more reading on the mentioned topics:
LUT Colors
Using of lut’s basically enables a different color palette. But i didn’t want to make it too ‘interesting’, so the current lineup consists of pushing trinitron phosphors, old lut for trinitron phosphors, a nec multisync display look (was very popular on the amiga scene) and a lut for an older NTSC standard.
LUT Size
‘3D’ luts can be of different sizes/qualities. Current lut png’s are 32 pix wide, but if you find/create a custom lut you want to use it could be sized differently. Endorsing the best quality the size of a lut of 64 is also possible.
Sega Brightness Fix
Sega MD games (?) usually don’t push max RGB component brightness to 255, but it usually stops with 239. Enabling the Sega Brightness Fix basically fixes the brightness range to the full potential. It can be very useful if someone creates a preset, which works nicely with his display and Sega games could look too dark.
5 Likes
I’m adding a nice ntsc-fast version preset, which i can be a starting point to some directions. Should work nice for 16 bit systems…
shaders = "12"
shader0 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/custom-jinc2-dedither.slang"
filter_linear0 = "false"
wrap_mode0 = "clamp_to_border"
mipmap_input0 = "false"
alias0 = ""
float_framebuffer0 = "false"
srgb_framebuffer0 = "false"
scale_type_x0 = "source"
scale_x0 = "1.000000"
scale_type_y0 = "source"
scale_y0 = "1.000000"
shader1 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/stock.slang"
filter_linear1 = "false"
wrap_mode1 = "clamp_to_border"
mipmap_input1 = "false"
alias1 = "StockPass"
float_framebuffer1 = "false"
srgb_framebuffer1 = "false"
scale_type_x1 = "source"
scale_x1 = "1.000000"
scale_type_y1 = "source"
scale_y1 = "1.000000"
shader2 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/pre-shaders.slang"
filter_linear2 = "true"
wrap_mode2 = "clamp_to_border"
mipmap_input2 = "false"
alias2 = "PrePass"
float_framebuffer2 = "false"
srgb_framebuffer2 = "false"
scale_type_x2 = "source"
scale_x2 = "1.000000"
scale_type_y2 = "source"
scale_y2 = "1.000000"
shader3 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/ntsc/ntsc-pass1.slang"
filter_linear3 = "false"
wrap_mode3 = "clamp_to_border"
frame_count_mod3 = "2"
mipmap_input3 = "false"
alias3 = ""
float_framebuffer3 = "true"
srgb_framebuffer3 = "false"
scale_type_x3 = "source"
scale_x3 = "4.000000"
scale_type_y3 = "source"
scale_y3 = "1.000000"
shader4 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/ntsc/ntsc-pass2.slang"
filter_linear4 = "true"
wrap_mode4 = "clamp_to_border"
mipmap_input4 = "false"
alias4 = ""
float_framebuffer4 = "false"
srgb_framebuffer4 = "false"
scale_type_x4 = "source"
scale_x4 = "0.500000"
scale_type_y4 = "source"
scale_y4 = "1.000000"
shader5 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/convert-ntsc.slang"
filter_linear5 = "true"
wrap_mode5 = "clamp_to_border"
mipmap_input5 = "false"
alias5 = "PrePass1"
float_framebuffer5 = "false"
srgb_framebuffer5 = "false"
scale_type_x5 = "source"
scale_x5 = "1.000000"
scale_type_y5 = "source"
scale_y5 = "1.000000"
shader6 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/linearize-ntsc.slang"
filter_linear6 = "true"
wrap_mode6 = "clamp_to_border"
mipmap_input6 = "false"
alias6 = "LinearizePass"
float_framebuffer6 = "true"
srgb_framebuffer6 = "false"
scale_type_x6 = "source"
scale_x6 = "1.000000"
scale_type_y6 = "source"
scale_y6 = "1.000000"
shader7 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/crt-guest-advanced-ntsc-pass1.slang"
filter_linear7 = "true"
wrap_mode7 = "clamp_to_border"
mipmap_input7 = "false"
alias7 = "Pass1"
float_framebuffer7 = "true"
srgb_framebuffer7 = "false"
scale_type_x7 = "viewport"
scale_x7 = "1.000000"
scale_type_y7 = "source"
scale_y7 = "1.000000"
shader8 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/bloom_horizontal.slang"
filter_linear8 = "true"
wrap_mode8 = "clamp_to_border"
mipmap_input8 = "false"
alias8 = ""
float_framebuffer8 = "true"
srgb_framebuffer8 = "false"
scale_type_x8 = "absolute"
scale_x8 = "640"
scale_type_y8 = "source"
scale_y8 = "1.000000"
shader9 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/bloom_vertical.slang"
filter_linear9 = "true"
wrap_mode9 = "clamp_to_border"
mipmap_input9 = "false"
alias9 = "BloomPass"
float_framebuffer9 = "true"
srgb_framebuffer9 = "false"
scale_type_x9 = "source"
scale_x9 = "1.000000"
scale_type_y9 = "absolute"
scale_y9 = "480"
shader10 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/crt-guest-advanced-ntsc-pass2.slang"
filter_linear10 = "true"
wrap_mode10 = "clamp_to_border"
mipmap_input10 = "false"
alias10 = ""
float_framebuffer10 = "false"
srgb_framebuffer10 = "false"
scale_type_x10 = "viewport"
scale_x10 = "1.000000"
scale_type_y10 = "viewport"
scale_y10 = "1.000000"
shader11 = "shaders_slang/crt-guest-advanced/shaders/guest/fast/deconvergence-f.slang"
filter_linear11 = "true"
wrap_mode11 = "clamp_to_border"
mipmap_input11 = "false"
alias11 = ""
float_framebuffer11 = "false"
srgb_framebuffer11 = "false"
scale_type_x11 = "viewport"
scale_x11 = "1.000000"
scale_type_y11 = "viewport"
scale_y11 = "1.000000"
do_dedither = "1.000000"
CS = "2.000000"
TNTC = "4.000000"
WP = "-20.000000"
quality = "-1.000000"
ntsc_phase = "3.000000"
ntsc_scale = "1.749999"
cust_fringing = "1.000000"
cust_artifacting = "1.500000"
nsharpen = "0.250000"
blendMode = "2.000000"
prescalex = "1.500000"
bloom = "-0.500000"
halation = "0.100000"
gamma_c = "1.200000"
shadowMask = "7.000000"
maskstr = "0.550000"
slotmask = "0.550000"
slotmask1 = "0.750000"
textures = "SamplerLUT1;SamplerLUT2;SamplerLUT3;SamplerLUT4"
SamplerLUT1 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/lut/trinitron-lut.png"
SamplerLUT1_linear = "true"
SamplerLUT1_wrap_mode = "clamp_to_border"
SamplerLUT1_mipmap = "false"
SamplerLUT2 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/lut/inv-trinitron-lut.png"
SamplerLUT2_linear = "true"
SamplerLUT2_wrap_mode = "clamp_to_border"
SamplerLUT2_mipmap = "false"
SamplerLUT3 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/lut/nec-lut.png"
SamplerLUT3_linear = "true"
SamplerLUT3_wrap_mode = "clamp_to_border"
SamplerLUT3_mipmap = "false"
SamplerLUT4 = "shaders_slang/crt-guest-advanced/shaders/guest/advanced/lut/ntsc-lut.png"
SamplerLUT4_linear = "true"
SamplerLUT4_wrap_mode = "clamp_to_border"
SamplerLUT4_mipmap = "false"
Screenie:
8 Likes
Currently i’m finishing a new version called ‘HD’. The main benefit is a custom sized and adjustable filter, which can filter any input resolution to both, horizontal and vertical direction. It’s main advantage is that the filter size can be very large, up to 32 original pixels.
Screenies:
10 Likes
Greetings @guest.r!
I’ve been having an issue with Shadow of The Beast for TurboGrafx-CD.
During the attract mode several full screen still images are cycled while some awesome CD background music is played, however during the transitions between stills the music pauses briefly and it totally messes up the timing and immersion of this simple but enjoyable part of this game.
This does not occur on real hardware.
I’ve tested it with redump images both CHD as well as uncompressed. Turned CD Cache on and off. With the image on an SSD as well as mechanical hard disk drive, it still occurs no matter what.
During my troubleshooting, I uncovered some interesting information with regards to performance and behaviour of different CRT Shaders including our beloved HSM Mega Bezel Reflection Shader.
Results of testing:
I disabled Video Shaders and it went away.
I use HSM Mega Bezel Reflection Shader but I didn’t suspect this as a contributor due to the fact that my GPU usage maxes out at 84%, while my CPU usage tops out at 7% during this intro sequence. I also have Run Ahead and Frame Delay enabled so I’m going to disable those or reset them to default and see what happens.
Disabling Run Ahead with the Shader running didn’t make a difference. Neither did Disabling Run Ahead plus resetting Frame Delay to 0 and turning off Automatic Frame Delay.
Going to try the Performance Shader Presets now.
With the Performance Preset in use, my maximum GPU usage drops to 76% (but it’s mostly tops out at 69%) while my CPU usage tops out at 4%
This is with Run Ahead and Frame Delay still disabled.
When I switched to HSM Mega Bezel Reflection Shader’s Potato base preset the issue is about 80% gone. There are still some signs of dropouts but they’re a lot shorter.
Max GPU usage is at 31% while CPU usage tops out at 5% (but it’s mostly at 1 to 2%)
After a while of looping its not as noticeable but I can still see the Frametime spikes and FPS drops and hear the slight pauses in the music during the attract mode scene transitions occasionally.
Tried Newpixie-CRT, no issues with dropouts whatsoever, smooth Frametime and framerate graph with max GPU usage at 17% and CPU usage topping out at 3%.
When I load CRT-GUEST-ADVANCE it’s probably a little better than when I tried the HSM Potato Preset. GPU usage maxing out at 27% and CPU at 2%.
CRT Royale-Composite seemed to have slightly less dropouts than CRT-GUEST-ADVANCE despite having similar GPU and CPU usage.
Lastly CRT-Geom Deluxe had a more consistent GPU usage figure of 40% with my GPU ClockSpeed remaining at a more or less constant 2000MHz and CPU usage at 2% Max and the issue was basically non-existent.
Even after playing around with some of the Shader Parameters in CRT-Geom Deluxe which caused my GPU Usage to increase to 59% Max and my CPU Usage to top out at 3%, the intro music in Shadow Of The Beast still played on without skipping a beat!
I tried setting my GPU to Prefer Maximum Performance but the problem was still there when I repeated using my HSM Mega Bezel Reflection Shader presets.
All of these tests were run with CD Cache Enabled in Beetle PCE Fast’s Core Options.
All of the test I ran besides the Potato and my initial tests using my presets were performed with the stock versions of the Shaders/Presets available in the Shaders_Slang folder unless otherwise indicated. I’ll test with another system. I’m running a mere Gefore GTX 1070.
After updating my Realtek Audio Drivers from the Gigabyte Website (since for the longest while Realtek seems to no longer be supplying their latest drivers to end users), I remembered that I don’t use Realtek Audio at all. I use nVIDIA HDMI Audio. My main PC is a HT/Gaming/Living Room PC hooked up to a receiver and a 4K TV via HDMI.
My nVIDIA HDMI Audio Drivers are as new as can be. They were probably updated about a day or 2 ago.
Just ran some tests on some more Shaders/Presets. RTSS Frame Limiter was off for these:
Guest Dr. Venom 2 37%, 2% - Audio Dropouts noticeable
Guest Dr. Venom 25%, 2% very slight, hardly noticeable but it’s there. After a few cycles, the FPS drop is about 2fps - the spikes are there but it’s even less noticeable.
Guest Dr. Venom Fast 18% 3%. Spikes can be seen, audio dropouts noticeable.
crt-guest-sm.slangp 23%, 3% No audio hitches for the most part but you can see the tiny blips on the RTSS Frametime and FPS Graphs plus you can hear an occasional audio dip but it’s extremely slight. Flawless after some cycles.
The percentages are max GPU and CPU usage figures during the test sequence respectively.
The problem seems to occur exactly between the image transitions. Something seems to be stalling at the exact moment the screen fades completely to black. Depending on the preset used its enough to desync the audio.
By the way, I switched on CRT-Geom-Deluxe in between and it was flawless, not even a spike. Of course that doesn’t do half as much as what CRT Guest Advanced can do but the GPU usage is around double that of Guest Shaders, yet the audio doesn’t budge.
Here’s a video clip showing the issue:
1 Like
That’s an interesting issue, but i definitely cannot help with it, since the chain of events, from emu core to post processing, including sound, is quite long. I guess that it’s repeatable with other heavier shader setups, like royale, dreamscape etc. Maybe you could try a different sound driver, like dsound or wasapi, see how it goes.
1 Like
I can try but the thing is its very rare. The vast majority of TurboGrafx-CD games don’t have this issue at all. As a matter of fact this is the first and only game that exhibits this issue so far. One thing that might be unique to this game is that each image in the intro sequence is at a different resolution and you can see the mode switching creating some lag in the video I made.
1 Like
that probably means it’s having to recompile the shader at each transition. You might try one of the nnedi3 presets, which also have very long compile times, and see if they have the same issue.
2 Likes
New Release Version (2022-01-18-r1):
Notable changes:
- HD version added
- It’s quite versatile, but it’s put to best use with sources with higher input resolution
- small fix for bloom with fast version
Download link:
https://mega.nz/file/AtgUgShT#Z_IjRDNQam8fVfYabKoNQMegN56GONx9kDGYOGp8dUo
5 Likes
I tested them all. Here are my findings:
Percentages are max GPU usage and CPU usage observed respectively during my test runs.
nnedi3-nns16-2x-luma.slangp
40%, 3%
Not a hitch
nnedi3-nns16-2x-rgb.slangp
34%, 3%
Not a hitch
nnedi3-nns16-4x-luma.slangp
40%, 3%
Not a hitch
nnedi3-nns32-2x-rgb-nns-4x-luma.slangp
49%, 3%
Very slight dips could be noticed in the audio but they stopped almost completely after a couple cycles of the images.
nnedi3-nns32-4x-rgb.slangp
91%, 3%
Hardly any dips or hitches, almost perfect, noticebly better than nnedi3-nns32-2x-rgb-nns-4x-luma.slangp despite the peak GPU usage being so high.
nnedi3-nns64-2x-nns32-4x-nns16-8x-rgb.slangp
100%, 3%
Constant 26fps and 100% GPU usage, everything running slow, garbled audio throughout
nnedi3-nns64-2x-nns32-4x-rgb.slangp
100%, 3%
Many dips to 57 fps depending on the resolution of the image not necessarily during the transitions. Lots of hitching, slowed and garbled audio with some listenable audio in between.
Using a different audio driver didn’t seem to make a difference.
2 Likes
