Log in

View Full Version : madVR - high quality video renderer (GPU assisted)


Pages : 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 [537] 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329

Asmodian
5th July 2014, 08:29
Limited and full are independent of chroma sampling. However, you cannot have chroma sampling and be RGB so your TV must be converting to YCbCr. Weather it converts to limited or full YCbCr will be harder to check.

mysterix
5th July 2014, 13:22
Does exist any utility which can show percent of VideoMemory bandwidth loading on AMD's (particularly, on my HD5450)? I know such utility exists for Nvidia cards definitely, but what about AMD? GPU-Z does not show that parameter for AMD cards...

panetesan2k6
6th July 2014, 00:43
Limited and full are independent of chroma sampling. However, you cannot have chroma sampling and be RGB so your TV must be converting to YCbCr.

Thanks for that, it's the conclussion I was looking for :)

Weather it converts to limited or full YCbCr will be harder to check.

I'll try to check this and will report back. :thanks:

Qaq
6th July 2014, 08:21
Does exist any utility which can show percent of VideoMemory bandwidth loading on AMD's (particularly, on my HD5450)? I know such utility exists for Nvidia cards definitely, but what about AMD? GPU-Z does not show that parameter for AMD cards...
5450 is just very limited in this regard. And too weak for madVR in general. If you prefer fanless AMD cards you may want to upgrade to fanless 7750 (like I did) to get more power for madVR and more sensors for GPU-Z.

xabregas
6th July 2014, 23:59
Hi, i just bought an nvidia gt 750 ti alongside with a smartv 1080p and im having several presentation glitches when i play 1080p movie at 1080p resolution, while i dont see any glitche at all, madvr ctrl j shows many...

The problem dont seem to happen while playing 720p movies upscaled to 1080p wit jinc 3 ar on chroma and luma. Only when i play 1080p movies (which dont need any upscale or downscale) the glitches appear.

I use Full screen exclusive mode...

Guess if the gpu dont do upscale/downscale, it wont be used and the fullscreen exclusive mode needs the gpu to do something???

TIA

huhn
7th July 2014, 01:24
you can disable FSE mode.

even with a 1080p source at a 1080p screen there is a lot to do.
chroma upscaling, dithering.

Anime Viewer
7th July 2014, 04:26
Hi, i just bought an nvidia gt 750 ti alongside with a smartv 1080p and im having several presentation glitches when i play 1080p movie at 1080p resolution, while i dont see any glitche at all, madvr ctrl j shows many...

The problem dont seem to happen while playing 720p movies upscaled to 1080p wit jinc 3 ar on chroma and luma. Only when i play 1080p movies (which dont need any upscale or downscale) the glitches appear.

I use Full screen exclusive mode...

Guess if the gpu dont do upscale/downscale, it wont be used and the fullscreen exclusive mode needs the gpu to do something???

TIA

Even when its not supposed to do upscaling (1080 -> 1080) it still tries to, and uses up resources. I recommend you make rules. For the 1080 rule choose DXVA2 which for all intensive purposes will disable all the other upscaling options, and give you the highest probably of having glitch free playback at that resolution.

Asmodian
7th July 2014, 05:12
Even when its not supposed to do upscaling (1080 -> 1080) it still tries to, and uses up resources. I recommend you make rules. For the 1080 rule choose DXVA2 which for all intensive purposes will disable all the other upscaling options, and give you the highest probably of having glitch free playback at that resolution.

What? This isn't true. DXVA2 will be fast and ok quality but there isn't any "it still tries to, and uses up resources" going on. Chroma upscaling, conversion to RGB, debanding (if enabled), smooth motion (if enabled), and dithering still needs to be done for 1080p displayed at 1920x1080.

@xabregas
The 750 Ti should be find for 1080p at 1920x1080 with Jinc3 AR for chroma scaling. Have you changed the dithering options or smooth motion? How about using Bicubic 75+AR for chroma scaling?

Also try with Full screen exclusive disabled and Windowed Overlay enabled.

Which motherboard do you have, madVR likes bandwidth so if you are on something very old there might be an issue there.

vivan
7th July 2014, 05:35
Even when its not supposed to do upscaling (1080 -> 1080) it still tries to, and uses up resources.What? No. You can easily prove it by choosing heavy upscaling algorithm and checking rendering time at 100% zoom and at higher one (like view -> video frame -> zoom 1),

For the 1080 rule choose DXVA2 which for all intensive purposes will disable all the other upscaling options, and give you the highest probably of having glitch free playback at that resolution.It shouldn't change anything. However even if it did - it would only harm quality, since using DXVA-things drops precision to 8 bit, which leads to banding. I would recommend never using them, unless you're using intel GPU and very limited with resourses.

innocenat
7th July 2014, 09:55
Hello,

Not sure if this is reported: madVR doesn't work (it only renders first frame and died [not crashing]) for me if I set it to use external graphic instead of integrated graphic in optimus setup.

- madVR 0.82.10
- Intel i7-4710HQ (mobile processor)
- nVidia GTX850M

It work fine if I set it to use integrated graphics only, but that would be a waste.

I am on latest nVidia driver (337.88). I used to use Optimus before in i7-2630QM+GT550M and it works fine running in external CPU on same driver/madVR version.

kopija
7th July 2014, 13:56
I have just upgraded from 0.8.6.11 to the newest version and am getting GPU driver chrashes.
I use MadVR primarily because I can watch movies and fold at the same time, thanks to the FSE mode.
Even though F@H consumes 95-98% of my 7870 I could use Jinc3 AR AND SM on 0.8.6.11 while folding simultaneously.
But not so with the newest version: even lowly bicubic causes driver restarts. I do not use image doubling nor error diffusion.
Since F@H uses OpenCL, I suspect that it could be fighting with madVR for use of OpenCL resources somehow.

So, basically, my guestion is: for which madVR functions is OpenCL used? Apart from NNEDI, ofcourse.

Thanks for your answers.

huhn
7th July 2014, 14:15
@kopija:

NNEDI3,ERROR DIFFUSION and under general "use OpenCL to process DXVA NV12 surfaces"

the newer version got some tweaks for optimus system could be a problem with that.

but is works totally fine with out F@H (what ever that is), right?

else: http://bugs.madshi.net/my_view_page.php feel free to add a entry.
Madshi is currently working on other things. so it better to add it there to make sure he will look at it later on.


edit: @innocenat

please try a newer version of MadVR 82.10 is ancient and report back.

generalmx
8th July 2014, 04:39
First I'd like to thank madshi and the community for such an excellent product :)

I have a watercooled R9 290X that's pretty consistent with its clocks (no overclock yet, stable at 1000MHz) and even this beast of a card can't handle everything without the occasional frame drop.

Specs
Windows 8.1 (Update 1) 64-bit (Pro w/ WMC)
Haswell Xeon 3.5GHz (w/ HT) 16GB DDR3 RAM
AMD Radeon R9 290X 4GB GDDR5
AMD Driver version: 14.16 (no difference with 14.14)

madVR version: 0.87.10
Software Used: MPC-HC "Lite" (*), MPC-BE (*), LAV Filters 0.62.0, ReClock 1.8.8.4
* - All internal filters disabled.

madVR & ReClock Configuration

ReClock
- WASAPI Exclusive
- Assume source is 30 FPS or Refresh Rate / 2
- Pre-Buffer: 100ms
- Quality: Best Sinc Interpolation
- Added MPC-BE to ReClock's "load always" since it's not added by default.
(all other defaults)

madVR 0.87.10 Configuration
(If not mentioned, option is disabled.)

Deinterlacing:
- Automatically activate deinterlacing when needed
-- if in doubt, deactive deinterlacing
- Only look at pixels in the frame center.

artifact removal:
- reduce banding artifacts
-- default debanding strength: low
-- strength during fade in/out: medium

Scaling Algorithms (All)
-- image downscaling: Catmull-Rom w/ Anti-Ringing Filter + Scale in Linear Light

Scaling Algorithms: <=360p
-- image upscaling: Jinc, 3 taps w/ Anti-Ringing Filter
-- chroma upscaling: NNEDI3, 256 neurons
-- image doubling: Always Double Luma, 64 neurons + Always Double Chroma, 16 neurons (*)

Scaling Algorithms: 360p<=480p
-- image upscaling: Jinc, 3 taps w/ Anti-Ringing Filter
-- chroma upscaling: NNEDI3, 128 neurons
-- image doubling: Always Double Luma, 32 neurons + Always Double Chroma, 16 neurons (*)

Scaling Algorithms: 480p<=720p
-- image upscaling: Jinc, 3 taps w/ Anti-Ringing Filter
-- chroma upscaling: NNEDI3, 32 neurons
-- image doubling: Always Double Luma, 16 neurons + Always Double Chroma, 16 neurons (*)

Scaling Algorithms: 720p<=1080p
-- image upscaling: DXVA2
-- chroma upscaling: NNEDI3, 16 neurons
-- image doubling: (None)

* Note: AMD interop hack doesn't seem to make a difference for my tests.

General Settings
- Delay playback start until render queue is full + delay playback start after seeking, too.
- Use OpenCL to process DXVA NV12 surfaces.
- Use separate device for DXVA processing.
- CPU/GPU queue size: 24

windowed mode settings:
- 16 frames in advance
- (rest are set to defaults)

trade quality for performance: ALL disabled

Smooth Motion
- Enable smooth motion frame rate conversion
-- ...or if the display refresh rate is an exact multiple of the movie frame rate.

Dithering (< 720p): Ordered Dithering
Dithering (>= 720p): Error Diffusion #2

Monitor
HP LP2465 (1920x1200 S-PVA panel)
- Properties: PC Levels (0-255), 8-bit (or higher)
- Calibration: Disabled + Disable GPU gamma ramps.
- Display Modes: (None)
- Color & Gamma: 0/disabled


AMD Catalyst Configuration
- Enable AMD Video Quality features in selected video player applications: UNCHECKED (disabled)
- Enforce Smooth Video Playback: UNCHECKED (disabled)

720p Hi10P Average Stats:
- Rendering (madVR): 28ms
- Interop (madVR): 10ms
- Present (madVR): 0.10ms
- GPU Utilization (HWiNFO): 58% (*)
- D3D Usage (HWiNFO): 23% (*)
- Time Spent with very high (>90%) GPU Usage: 28% (*)
Note: No dropped frames over this report.

(*) I have a super-fancy spreadsheet I made to calculate all of this.

I have tried a large number of different settings from suggestions online and in this thread, and so far these have shown to be the best -- however there are plenty of different source material (I'm mostly testing on anime) that trips it up. Here's some especially weird stuff:

* My settings don't seem to like ReClock set to 24 FPS (for ~24 FPS source), causing more to significantly more dropped frames. I have two monitors connected in an "Extended" desktop: 24" 1200p S-PVA workstation monitor and a crappy 32" 1080p TN HDTV. Both can support 24Hz natively, especially the 1080p, though I must use CRU to add 24Hz for the 1200p, however, even the 1080p set to 24Hz causes significantly more dropped frames (same with 48Hz). Just to check if it was the case of using Windowed and two monitors running different resolutions, I set both monitors to 1080p@24 and the same problem occurred. But I'm guessing the reason it's totally unplayable if Monitor #2 (1080p, Secondary) is set to 1080p@24 and Monitor #1 (1200p, Focus) is still set to 1200@60 is how Windows handles refresh rate differences like that.

* Full-Screen Exclusive Mode gives more to significantly more dropped frames for me, and can introduce artifact errors. Again, note that I can't get EDID for Monitor #2 (1080p), and it's using some generic PnP driver, and Monitor #1 is a 1200p workstation monitor that doesn't support 24Hz (or 48Hz) without hackery.

Hence I've settled on ReClock with 30 FPS or Refresh Rate / 2, which requires Smooth Motion enabled or else I'll get significant judder on anime pan & scan.

Note on Pan & Scan in Anime: I've found certain sources -- say, HorribleSubs -- often have low bitrate sources that will have problems on pan&scan no matter what you do; only outright frame-doubling really helps here (SVP, CUVID, etc.).

On Windows Power Profile: I'd like to stress that you should be using "High Performance" if you're having any sort of trouble, NOT just changing settings under "Balanced". Process Lasso is an example of a program that will set the power profile for you based on what programs are running.

Anime Viewer
8th July 2014, 06:01
* My settings don't seem to like ReClock set to 24 FPS (for ~24 FPS source), causing more to significantly more dropped frames. I have two monitors connected in an "Extended" desktop: 24" 1200p S-PVA workstation monitor and a crappy 32" 1080p TN HDTV. Both can support 24Hz natively, especially the 1080p, though I must use CRU to add 24Hz for the 1200p, however, even the 1080p set to 24Hz causes significantly more dropped frames.

Hence I've settled on ReClock with 30 FPS or Refresh Rate / 2, which requires Smooth Motion enabled or else I'll get significant judder on anime pan & scan.


I'd noticed the same thing on my system when it would try to play 24fps content. At one time I thought it might be the encoding on the video, or possible variable frame rate in the 24fps videos I'd been watching, but since your noticing it too it could be ReClock related like you suspect.

My work around was to delete the 1080p24 and 1080p23 display modes from the list of supported display modes in my madVR settings. With it set as 1080p60, 1080p59, 1080p30, 1080p29 Reclock goes to the next closest thing (30hz). I proposed a change to reclock setting options based on another type of 24hz flawed playback experience (detailed at the linked page), but nothing ever became of it.
https://forum.slysoft.com/showthread.php?60908-Proposal-for-future-Reclock-setting%28s%29&

generalmx
8th July 2014, 07:18
I'd noticed the same thing on my system when it would try to play 24fps content. At one time I thought it might be the encoding on the video, or possible variable frame rate in the 24fps videos I'd been watching, but since your noticing it too it could be ReClock related like you suspect.

My work around was to delete the 1080p24 and 1080p23 display modes from the list of supported display modes in my madVR settings. With it set as 1080p60, 1080p59, 1080p30, 1080p29 Reclock goes to the next closest thing (30hz). I proposed a change to reclock setting options based on another type of 24hz flawed playback experience (detailed at the linked page), but nothing ever became of it.
https://forum.slysoft.com/showthread.php?60908-Proposal-for-future-Reclock-setting%28s%29&

Hmm, interesting. But I do think I've found a significant source of dropped frames that I think is just related to madVR: blending of high contrast / overexposed and normal contrast scenes. With my settings, every now and then, there will be a big string of dropped frames from rapid transitions between the different contrast levels, which I'm guessing is a side-effect of manipulating luma and chroma. And while one could see this in live-action, you're most likely to see it in more light-hearted anime; where even your typical shounen/action anime uses more basic colouring and lighting.

Unfortunately I haven't yet found a combination of madVR settings that prevents dropped frames in this situation, though I know it's my madVR settings as using a different renderer (with or without ReClock) works fine...which is a shame, since it's these anime that looks absolutely brilliant with NNEDI3, so I've just been rewinding when it happens (it's not consistent either).

ryrynz
8th July 2014, 08:07
Unfortunately I haven't yet found a combination of madVR settings that prevents dropped frames in this situation

If all combinations of scalers aren't changing anything, have a play with the buffer and flush settings (lowering and disabling accordingly) and see what happens.

ryrynz
8th July 2014, 08:18
please try a newer version of MadVR 82.10 is ancient and report back.

He meant 0.87.10 there never was a 0.82.10. He's on the latest version.

michkrol
8th July 2014, 08:31
Hmm, interesting. But I do think I've found a significant source of dropped frames that I think is just related to madVR: blending of high contrast / overexposed and normal contrast scenes. With my settings, every now and then, there will be a big string of dropped frames from rapid transitions between the different contrast levels, which I'm guessing is a side-effect of manipulating luma and chroma.

Try those scenes with disabled debanding. Then try with both "strength" and "fade in/out" set to the same level.

With "don't rerender frames ..." trade quality for performance option disabled, when madvr detects a fade in/out it drops frames and rerenders them using other debanding setting. If this works as intended you don't even notice it, but if you have maxed out settings and one fade in/out after another in the video you rerender frames almost constantly with different settings. That's why some quality trade-offs are enabled by default ;)

In you previous post you have mentioned judder with some sources, are you sure it's not in the source? If your encode or deinterlacing goes wrong you will get duplicate frames, which means judder during pan scenes. Easiest to check by frame-stepping through the panning scene.

QBhd
8th July 2014, 08:55
@ generalmx:

For ReClock settings, I have found that the best way to work with madVR is to have every single option un-checked on both the advanced settings and the audio settings. Also is there a reason you are using WASAPI? DirectSound is all you really need to use, unless there is a specific reason.

Have you tried using 13.12 AMD drivers? Any of the newer drivers cause performance drops with my 270x... maybe something you should look into as well

QB

Asmodian
8th July 2014, 09:40
Hmm, interesting. But I do think I've found a significant source of dropped frames that I think is just related to madVR: blending of high contrast / overexposed and normal contrast scenes. With my settings, every now and then, there will be a big string of dropped frames from rapid transitions between the different contrast levels, which I'm guessing is a side-effect of manipulating luma and chroma. And while one could see this in live-action, you're most likely to see it in more light-hearted anime; where even your typical shounen/action anime uses more basic colouring and lighting.

Unfortunately I haven't yet found a combination of madVR settings that prevents dropped frames in this situation

Have you tried turning on the trade quality for performance option "don't rerender frames when fade in/out is detected"? Your description sounds like it could trigger fade detection and needing to rerender several frames very often could account for your issue. This is also the only option that would depend of the content of the video, as far as I know (besides a minor effect when using 3DLUTs).

edit: missed michkrol's reply - yeah, what he said. ;)

madshi
8th July 2014, 10:27
Instead I should say I don't want the dithering function at the input signal side to affect the HX920's super-resolution engine's processing efficiency.
But are you sure that it does? I don't really know what the super-res engine does inside. Depending on the algorithm it could work better with dithering off, or with dithering on, or it could make no difference. There's no way to know except trying which looks better.

I don't remember: Which value does deintFps have if you activate forced film mode for 60i telecined movie content? Is it 24fps? I guess it would make sense and should then also apply to decimated 60p content.
24fps is what I've seen in my testing if I activate forced film mode for 60i telecined movie content.
Ok, in that case it should also be 24fps for decimated 720p60 content. Is there already an entry in the bug tracker for this? If not, could you please make one? Thx.

I have a weird problem on my Win7 x64 HTPC. Installed X86 MPC-HC and madVR ( install.bat ), but although I see madvr in the list available renderers ( output section ) and I can select it, the renderer I see later in the list during the playback is the default "video renderer" of MPC-HC. MadVR wont load..tried to reinstall everything, but Im done. What could be the cuprit please?
Hard to say. Does madTPG show an image and react to the OSD buttons etc? If so, madVR seems to generally work and the media player or your selected DirectShow splitter/decoder could be causing the problem. If not, your PC doesn't seem to like madVR for some reason. In the latter case try temporarily disabling your anti-virus/firewall just to check whether that's causing the problems or not (of course to be safe feel free to first check madVR via www.virustotal.com).


If my receiver is not passing 4:4:4 from my Radeon 5450 too my display, then I gather setting 4:4:4 in the AMD CCC is useless?
I learnt from reading on here that madVR works by converting Y'CbCr very well to RGB and RGB therefore should always be selected in the video card driver. But if I can't accept 4:4:4 on my display because of my receiver, is setting the CCC to Y'CbCr 4:2:2 a better option?
On my Samsung t.v how can I make sure I am not expecting to receive a 4:4:4 chroma signal? There is an option for Colour Space (AUTO/WIDE). Does this have anything to do with Chroma, or is this only related too limited/full range.
HDMI Black level on my t.v is only available to change when I output Y'CbCr 4:2:2 and not Y'CbCr/RGB 4:4:4. Why is this? If I understand correctly it is this setting that assumes you will be receiving blacks darker than 16, but does it also dictate receiving whites greater than 235?
How does all of this relate to bit-depth? When I have Y'CbCr 4:4:2 or RGB 4:4:4 Limited in CCC my receiver shows its receiving a 30bit signal and is outputting 24bit to my display. With 4:4:4 it shows receiving and outputing 30bit.
When I connect my laptop, or my PVR the receiver instead shows an input of 24bit, why does my laptop/PVR show this but not my PC with the radeon?

Most of these questions are very specific to your hardware, so it's hard for any of us to give an exact answer. madVR always renders to RGB. If you set the GPU to YCbCr output (either 4:4:4 or 4:2:2) the GPU is converting madVR's output behind madVR's back. Otherwise your receiver is doing that. Who does it better? I don't know. I can only advise to try different settings with test patterns and real movies and trust your eyes which setting combination produces the best overall result.

Already tried PC standard(full RGB) on GPU output & 16-135 on madVR output, and yeah, doesn't look nice on games and others. I also notice that black levels & grayscale ramp is a bit off the way it should look.

On the other hand, going Studio(limited RGB) on GPU output & 0-255 on madVR output makes other application look proper. Black levels is also spot on(16 is black and 17 to 235 is visible) & grayscale ramp looks nicer with only few improper gray steps.
Yes, if you want other applications to have proper black and white levels, there are only 2 options for you:

(1) Set the GPU to Studio=limited RGB=16-235. This will produce correct black and white levels, but it might introduce banding artifacts.

(2) Set the GPU to PC Standard=full RGB=0-255. And at the same time set your display to 0-255, too! This will produce correct black and white levels and it will avoid banding artifacts. However, not all displays/TVs support this.

I've mentioned this before, but now have a sample to back it up. Using anti-ringing filter on any of the downscaling algorithms after NNEDI3 causes jagged/aliased text. With the AR filter off text looks much, much better. This happens on any video I play, and can most easily be seen at the beginning of tv shows where it shows the actors names.

Take a look at the ticker at the bottom. Any letter with a diagonal like W, N, X, M, ect looks terrible with AR on.
http://www44.zippyshare.com/v/62692648/file.html
Thanks. Could you please create a bug tracker entry for this (if you haven't already)?

Small idea.
It could be helpful for evaluation of LUTs if one could set 2 LUTs and a shortcut for switching between them for A/B comparison.
Maybe even use both LUTs on one half of the image each.
Yes, features for A/B comparison would be nice. But really, they have low priority compared to some other things. So this won't come soon.

As as side note, I find it incredible how much information can be conveyed with 1bit dithering (or 8 colors) :D
Agreed. Especially the gamma corrected dithering helps a lot at such low bitdepths (thanks Shiandow).

Since a few weeks AMD has realased their new 14.6 beta driver. This driver adds a setting to set display color depth for displays that support it. See screenie below.

I've installed the driver, and can set my TV to 10 bpc. When i then set MPC to D3D output, 10 bit out etc. etc. I actually can get 10 bit out (a2R10G10B10 output, CTRL-J stats screen, had to use EVR custom to confirm). So i can confirm this setting seems to work.
Oh, that's pretty good news - thanks!

So as promised they asked AMD about the copyback nonsense, they replied that their engineering team was aware of the problem but didn't have the time to look at it.
Hmmm... They didn't have time to look at it *yet*? Or they generally don't have time to look at it?

I've just started getting into Madvr but I have a minor GFX clipping problem that has plagued me for years and need help resolving it especially now using a quality renderer, i'm at a brick wall before I can carry on.

I have always used amd cards and using a common pluge pattern I have always had to up bri +2 and lower contrast -2 in the CCC settings to get all bars to show correctly, the consensus seems to be that the GFX driver settings for color should be untouched but this gives me a 19-233 video output, current new card is an amd R9 790, this is exactly the same result!

I have tried my onboard haswell 4600, same result as if something is clipping the video slightly.

With madvr integrated into the chain, my settings for colorspace are amd CCC - dynamic range 0-255 - video hdmi output - 0-255 - madvr 16-235 - display set to 16-235 (Pioneer KRP600M).

This gives the same 19-233 output, any change to the above gives either all bars 0+ and washed out or no bars at all on the black clipping pattern with heavy clipping.

I tried just adjusting madvr bri/cont which required bri +10, cont -40 to get 17/18 and 234/235+ bars to show correctly but the video image was a mess and overblown!

This is really frustrating me and need help :(

EDIT: My display has been calibrated using a bluray player and calman5, so it's not the display clipping.
This sounds quite weird. First of all, you can define a custom output levels configuration in madVR. Try 13-237, that might make things work for you.

However, something is going wrong there. My first guess would be that maybe you have some sort of 1dlut loaded through Windows calibration or something? In the madVR device setup under "calibration" make sure you have "disable GPU gamma ramps" checked. Does that help? If not, my next guess would be that either your receiver or your display is messing things up. It's possible that your Blu-Ray player output wrong levels. Or it's possible that your display behaves differently when receiving RGB input. Your Blu-Ray player probably output YCbCr.

From the feedback other users are giving me, a properly configured HTPC seems to output correct levels using madVR. So if you get the very same 19-233 with two different GPUs, then the issue is very likely to either be a Windows configuration problem (the "disable GPU gamma ramps" option should work around the most likely such issue), or a problem outside of the HTPC.

Contrary to what some people have told you Image Doubling can still be active when playing a 1080p video on a 1080p monitor [...]
No, to the best of my knowledge that's not true.

if I use windows classic theme without desktop composition on Win7.. there's tearing in windowed playback mode.

is there a fix for this?
You're probably using an AMD GPU? They're known to have this problem. At least NVidia users are not reporting this as often as AMD users. Unfortunately I dont have a fix for you, other than either enabling desktop composition or using FSE mode. Or upgrade to Windows 8, which has desktop composition forced on - but on the positive side, Windows 8 has a much better desktop composition implementation compared to Windows 7. The implementation in Windows 8 is good enough that I never felt the need to disable it yet.

I was wondering if there is still hope for a 64bit MadVR version in the foreseeable future
madVR x64 will probably come, but probably not very soon.

Ok, I have lowered the number of "in advance" frames in FSE mode to 4 frames, picture appeared, but madvr crashes after couple of seconds and creates crashslog. The same behaviour with 4,3,2,1.

6+ frames blacks out the FSE mode.
Interesting. Try lowing the GPU queue size. You seem to be running out of GPU RAM, unfortuantely... :(

it blends frame based on the Vsync if two frames have to be shown in the same vsync it will blend them based on the time in % they have in this Vsync.

nearly every TV out there does the same.
AFAIK, TVs with proper multi-refresh-rate abilities simply change the driving method to achieve different refresh rates natively. Which is a much better solution (if done properly) compared to madVR's smooth motion blending.

madshi
8th July 2014, 10:30
If you toggle NNEDI3 on or off you should see the image move slightly, if you don't then NNEDI3 is not working. If you do then it's probably working.
FWIW, the latest madVR build avoids the NNEDI3 0.5 pixel shift - except if you do *exactly* 2.0000x scaling.

While experimenting with shaders I seem to have discovered a bug. MadVR crashes when the total number of shaders becomes "too large" the exact number seems to depend a bit on the shaders and the settings but is usually somewhere between 16 and 28. Anyway I've added a bug report, hopefully you'll be able to fix it sometime.
Yes, I have a fixed array size for the rendering steps, and every custom shader consumes such a rendering step. I will need to dynamically allocate the array. Shouldn't be hard to do. I didn't anticipate that anybody would use so many custom shaders (or passes). My fault... :o

I'm finding that when the PC is being used on the primary monitor (connected via DVI), and video is being played back on the (secondary) HDTV, I get presentation glitches, audio delays, choppy playback and all sorts of other things going on. This doesn't seem to affect any applications being used on the primary monitor.

The PC is used for a variety of things, including Photoshop, normal web browsing, document editing, etc. Nothing specific seems to set off the problem; just moving the mouse on the other screen can cause the issues to occur if video is being played back on the HDTV.

Specs of the PC are:
AMD FX8350 (8-core, 4.0GHz)
8GB Corsair Vengeance 1600MHz RAM
Sapphire Radeon HD7870 GHz Edition
Western Digital Black 2TB storage drive

I've tried playing with the various settings in MadVR in an attempt to resolve the problems (disabling image doubling, choosing different scaling algorithms etc), but no joy so far.

This seems to happen with all media types as well, including avi, mkv and mpg.

Any help would be greatly appreciated!

*edit*
I probably should have mentioned that when the PC is not being used for anything else, playback is flawless.
Are you using FSE mode or windowed mode? Generally windowed mode is more susceptible if the CPU/GPU is busy doing other stuff. Although the new windowed mode in the latest madVR build should be somewhat more reliable. Are you using the latest madVR build? And do you have the "present several frames in advance" option turned on?

Also check the madVR OSD (Ctrl+J) to see if any of the queues (which?) are getting empty when the glitches/issues occur.

Which OS are you using?

I don't know if this is a bug but if I untick "don't rerender frames when fade in/out is detected" I get about 7-10 droppped frames each time there is a black fade going on. Makes some of the files unwatchable when there's alot of fast movement and black scene changes.

This happens in both windowed and fse mode. Max rendering times are around 7ms so that shouldn't be the culprit.

Ps. My gpu is R9 290X so there should be plenty of power to run madvr.
Try increasing the CPU and GPU queue sizes. Personally, I'm currently using 24 frames for CPU and 16 frames for GPU. This should help avoiding those dropped frames when there's a fade occurring. Of course you'll need to have enough GPU RAM for this to work properly.

That's more or less expected behaviour. As far as I understand it takes a few frames before MadVR realises that there is fade in/out so what "rerender frames when fade in/out is detected" does is go back a few frames whenever MadVR detects a fade in/out and rerenders those frames to prevent banding on the first few frames of the fade. If this is causing issues it's probably best to leave it off.
Well, it's working fine on my PC without any frame drops or other issues. I think the queues just need to be set a bit larger.

MadVR works natively in full rgb, but when you select "PC Levels" output it cuts BTB and WTW intentionally (as madshi explains here (http://forum.doom9.org/showthread.php?p=1400157#post1400157)), so best/preferable scenario is to use a display that can handle full rgb. This won't give us BTB and WTW, but that's ok. In case of having a display that can't handle full rgb, the signal must be "squeezed" to 16-235 either by madVR (setting "TV Levels" in MadVR and leaving desktop whites and blacks clipped) or the GPU (setting "RGB Limited" in the driver and leaving MadVR in "PC Levels"), having desktop and video correct values. I hope I got it right finally.
Well, your description is technically not fully correct, but it doesn't matter much. You got the general idea correctly.

Hi, i just bought an nvidia gt 750 ti alongside with a smartv 1080p and im having several presentation glitches when i play 1080p movie at 1080p resolution, while i dont see any glitche at all, madvr ctrl j shows many...

The problem dont seem to happen while playing 720p movies upscaled to 1080p wit jinc 3 ar on chroma and luma. Only when i play 1080p movies (which dont need any upscale or downscale) the glitches appear.

I use Full screen exclusive mode...
Sounds quite strange. Do you have smooth motion FRC frame blending active? Does the problem go away if you disable FRC? What happens if you play a 1080p60 file?

Even when its not supposed to do upscaling (1080 -> 1080) it still tries to, and uses up resources.
Nope, that's not true.

Not sure if this is reported: madVR doesn't work (it only renders first frame and died [not crashing]) for me if I set it to use external graphic instead of integrated graphic in optimus setup.

- madVR 0.82.10
- Intel i7-4710HQ (mobile processor)
- nVidia GTX850M

It work fine if I set it to use integrated graphics only, but that would be a waste.

I am on latest nVidia driver (337.88). I used to use Optimus before in i7-2630QM+GT550M and it works fine running in external CPU on same driver/madVR version.
Hmmmm... Try disabling the options "use a separate device for presentation" and "use a separate device for DXVA processing". Does that help? If not, try reducing the size of the GPU queue and the number of pre-presented frames.

Scaling Algorithms: <=360p
-- image upscaling: Jinc, 3 taps w/ Anti-Ringing Filter
-- chroma upscaling: NNEDI3, 256 neurons
-- image doubling: Always Double Luma, 64 neurons + Always Double Chroma, 16 neurons (*)
FWIW, chroma doubling usually does very little to improve the subjective image quality.

General Settings
- Use OpenCL to process DXVA NV12 surfaces.
With the current state of AMD OpenCL drivers I'd recommend to turn this option off.

* My settings don't seem to like ReClock set to 24 FPS (for ~24 FPS source), causing more to significantly more dropped frames. I have two monitors connected in an "Extended" desktop: 24" 1200p S-PVA workstation monitor and a crappy 32" 1080p TN HDTV. Both can support 24Hz natively, especially the 1080p, though I must use CRU to add 24Hz for the 1200p, however, even the 1080p set to 24Hz causes significantly more dropped frames (same with 48Hz). Just to check if it was the case of using Windowed and two monitors running different resolutions, I set both monitors to 1080p@24 and the same problem occurred. But I'm guessing the reason it's totally unplayable if Monitor #2 (1080p, Secondary) is set to 1080p@24 and Monitor #1 (1200p, Focus) is still set to 1200@60 is how Windows handles refresh rate differences like that.
Does this problem go away if you disable NNEDI3 or if you play a native 1080p24 file which doesn't need scaling? Just as a test to see if this problem is related to high GPU consumption? Have you tried clearing/resetting the Reclock timing database? What happens if you disable Reclock completely?

* Full-Screen Exclusive Mode gives more to significantly more dropped frames for me, and can introduce artifact errors. Again, note that I can't get EDID for Monitor #2 (1080p), and it's using some generic PnP driver, and Monitor #1 is a 1200p workstation monitor that doesn't support 24Hz (or 48Hz) without hackery.
Strange. What kind of artifacts do you get?

Hmm, interesting. But I do think I've found a significant source of dropped frames that I think is just related to madVR: blending of high contrast / overexposed and normal contrast scenes. With my settings, every now and then, there will be a big string of dropped frames from rapid transitions between the different contrast levels, which I'm guessing is a side-effect of manipulating luma and chroma. And while one could see this in live-action, you're most likely to see it in more light-hearted anime; where even your typical shounen/action anime uses more basic colouring and lighting.
Try those scenes with disabled debanding. Then try with both "strength" and "fade in/out" set to the same level.

With "don't rerender frames ..." trade quality for performance option disabled, when madvr detects a fade in/out it drops frames and rerenders them using other debanding setting. If this works as intended you don't even notice it, but if you have maxed out settings and one fade in/out after another in the video you rerender frames almost constantly with different settings. That's why some quality trade-offs are enabled by default ;)
^

michkrol's guess would also have been my best guess, that the frame drops are caused by the re-rendering of the frames when a fade is detected. If the tests suggested by michkrol confirm this, you could also try if increasing the size of the CPU and GPU queues fixes the problem.

===============

I've recently moved to a new home which meant no time for madVR development. I'm now slowly getting back to development. But my commercial projects need some attention first. So don't expect a major new madVR version with new killer features soon.

huhn
8th July 2014, 13:26
I've recently moved to a new home which meant no time for madVR development.

i hope with a proper HT room.

Anime Viewer
8th July 2014, 13:33
Nope, that's not true.


Then is there another explanation for why render times would be higher when a 1080p video is playing on a 1080p monitor full screen with settings like Jinc or NNEDI3 set compared to when that same video is playing on that same monitor at full screen with something like DXVA2 or Bi-linear set for a scaling algorithm like image upscaling?

xabregas
8th July 2014, 13:58
What? This isn't true. DXVA2 will be fast and ok quality but there isn't any "it still tries to, and uses up resources" going on. Chroma upscaling, conversion to RGB, debanding (if enabled), smooth motion (if enabled), and dithering still needs to be done for 1080p displayed at 1920x1080.

@xabregas
The 750 Ti should be find for 1080p at 1920x1080 with Jinc3 AR for chroma scaling. Have you changed the dithering options or smooth motion? How about using Bicubic 75+AR for chroma scaling?

Also try with Full screen exclusive disabled and Windowed Overlay enabled.

Which motherboard do you have, madVR likes bandwidth so if you are on something very old there might be an issue there.

So i did what u said, disabled FSE mode but didnt activate windowed overlay as i had no tearing in FS with FSE mode deactivated so i suppose i dont need windowed overlay right?

Ok so i tried several 1080p videos, VC1 and AVC with high bitrates and no presentation glitches, ofc i put bicubic 75 AR in Chroma also...

Also, i turned Smooth motion on for some TV shows i have, even knowing i have 1080p 24hz, for TV shows is better 60hz, but with movies at 23.976, if i enable 24HZ should i use Smooth Motion? DO i need videoclock or reclock?

TIA

huhn
8th July 2014, 14:13
Then is there another explanation for why render times would be higher when a 1080p video is playing on a 1080p monitor full screen with settings like Jinc or NNEDI3 set compared to when that same video is playing on that same monitor at full screen with something like DXVA2 or Bi-linear set for a scaling algorithm like image upscaling?

do you change chroma too? if yes there is your answer all dvd/bd need a chroma upscaling always.

and powerstates make rendertime unreliable too shouldn't apply to your case but i have normally higher render times with 1080p else with 720p on 1080p the powerstate with 1080p is lowest so the render time look higher.

and are you sure your source is true 1080p and not cropped by 2-6 pixel?

kopija
8th July 2014, 17:26
@kopija:

NNEDI3,ERROR DIFFUSION and under general "use OpenCL to process DXVA NV12 surfaces"

the newer version got some tweaks for optimus system could be a problem with that.

but is works totally fine with out F@H (what ever that is), right?



Thanks, will try it out.
And all of you guys with monster cards should read this:
http://en.wikipedia.org/wiki/Folding@home#Biomedical_research

Another question:
LAV Video decoder is performing random dithering and MadVR is performing ordered dithering by default. So is there some kind of conflict? I canot disable dithering in LAV Video, so should dithering be set to "None" in madVR?

huhn
8th July 2014, 17:38
Another question:
LAV Video decoder is performing random dithering and MadVR is performing ordered dithering by default. So is there some kind of conflict? I canot disable dithering in LAV Video, so should dithering be set to "None" in madVR?

there should be a stick about this.

leave dither in MadVR active this option in lavfilter is not used ignore it. it's only used when YCbCr is transformed to RGB this is not happening with MadVR. and even when it is used MadVR should still dither at the end.

James Freeman
8th July 2014, 17:49
I've recently moved to a new home which meant no time for madVR development. I'm now slowly getting back to development. But my commercial projects need some attention first.
So don't expect a major new madVR version with new killer features soon.
Big congratulations on your new home!

Killer the features will be nevertheless. ;)

Stan
8th July 2014, 17:49
madVR x64 will probably come, but probably not very soon.


Sadface :(

Asmodian
8th July 2014, 19:20
So i did what u said, disabled FSE mode but didnt activate windowed overlay as i had no tearing in FS with FSE mode deactivated so i suppose i dont need windowed overlay right?

On Nvidia windowed overlay is the lowest effort mode but it is a very small difference and there are benefits to the non-overlay modes.

Also, i turned Smooth motion on for some TV shows i have, even knowing i have 1080p 24hz, for TV shows is better 60hz, but with movies at 23.976, if i enable 24HZ should i use Smooth Motion? DO i need videoclock or reclock?

If you leave smooth motion on "only if there would be judder" it will automatically turn off when watching 23.976fps @ 24Hz. Reclock will still help if you want to sync timings exactly but I stopped using it with madVR and I don't notice or measure any dropped or repeated frames (on a 60 or 72 Hz display). This might depend on your exact video clock(s) though.

xabregas
8th July 2014, 20:11
If you leave smooth motion on "only if there would be judder" it will automatically turn off when watching 23.976fps @ 24Hz. Reclock will still help if you want to sync timings exactly but I stopped using it with madVR and I don't notice or measure any dropped or repeated frames (on a 60 or 72 Hz display). This might depend on your exact video clock(s) though.

I notice that in 24Hz smooth motion is off, but i notice some oos issues, not by much. Should i use videoclock or reclock?

The oos issues get worse when i watch 25fps TV shows on 50HZ, smooth motion is off but audio gets oos. :(

Asmodian
8th July 2014, 21:45
Odd, your video clocks must be fairly off? Yes reclock would probably help.

pie1394
9th July 2014, 03:49
But are you sure that it does? I don't really know what the super-res engine does inside. Depending on the algorithm it could work better with dithering off, or with dithering on, or it could make no difference. There's no way to know except trying which looks better.


According to Sony DRC team's chief engineer comment, the material library of thousands video image patterns are built into the XCA7 engine. I don't know the details how the process is done. But I guess it could be pattern-matching process similar to image recognization to identify the image content's characteristsics. If I am not wrong, the traditional ways only rely on temporal multi-image information to restore the lost detailed contents?

Although this TV set has the 960Hz backlit-strobe scanning + 240Hz native panel, however, I sometimes still found this engine's Clear+ motion interpolation handling is not perfect. For very few 24 fps contents moving scenes, back-and-force-flicking or ghost image could be still suddenly observed. :p

dansrfe
9th July 2014, 06:54
Is it possible for madVR to exclusively utilize the discrete GPU instead of the integrated Intel GPU?

subz3ro
9th July 2014, 11:50
Is it only me, or anybody else experience random crashes with madVR and Smooth Motion enabled, when seeking through media in Media Player Classic? This is not happening with the 0.86.x version, only with the 0.87.x one.

ryrynz
9th July 2014, 13:13
Just you.. 0.87.x is rock solid.

Anime Viewer
9th July 2014, 13:34
Is it possible for madVR to exclusively utilize the discrete GPU instead of the integrated Intel GPU?

There are a few things you need to do, and that you can check to make sure madVR is using your dGpu.

If you're using Media Player Classic Home Cinema make sure to rename the executable for it from mpc-hc.exe to something else like mpc-hc2.exe, mpc-hchd.exe, mpc-hcnv.exe (pretty much anything aside from the original name will work).

Go into Nvidia Control Panel (I'm guessing you have an Nvidia/Intel system, but if that's not the case you'll need to go into the driver settings area for whatever GPU you have). Under manage 3d settings switch to the Program Settings tab. Select Add and navigate to your renamed mpc-hc file. Change the area under perferred graphics processor for this program to the High-performance Nvidia processor.

Open a video. You should notice an improvement in your render times. Exit the video and open your registry. Navigate to HKEY_CURRENT_USER\Software\madshi\madVR\OpenCL if it has correctly detected your dGPU it should have it listed as a key inside that key showing its being used for OpenCL processes. If you don't see your gpu listed there post again and let us know, and we can give you instructions for creating a key that will force it to be added.

madshi
9th July 2014, 13:50
i hope with a proper HT room.
Not yet, but there's going to be a dedicated HT front projection room.

Then is there another explanation for why render times would be higher when a 1080p video is playing on a 1080p monitor full screen with settings like Jinc or NNEDI3 set compared to when that same video is playing on that same monitor at full screen with something like DXVA2 or Bi-linear set for a scaling algorithm like image upscaling?
do you change chroma too? if yes there is your answer all dvd/bd need a chroma upscaling always.

and powerstates make rendertime unreliable too shouldn't apply to your case but i have normally higher render times with 1080p else with 720p on 1080p the powerstate with 1080p is lowest so the render time look higher.

and are you sure your source is true 1080p and not cropped by 2-6 pixel?
^

madVR carefully checks the source and target resolutions and only does what is necessary. If no upscaling/doubling is needed, it is not performed at all. If you have higher GPU consumption with Jinc compared to Bilinear with a video that you think needs no scaling, there must be something wrong somewhere. Of course it could be a bug in madVR, but I rather think it's likely something else. Check the OSD to make sure source and target width/height *really* match perfectly.

I notice that in 24Hz smooth motion is off, but i notice some oos issues, not by much. Should i use videoclock or reclock?

The oos issues get worse when i watch 25fps TV shows on 50HZ, smooth motion is off but audio gets oos. :(
The only way you should get oos problems with madVR is if you're constantly dropping lots of frames. If that's not the case and you *still* get oos problems, then it must be something really strange, like your GPU driver or audio renderer messing things up, or something like that. Not sure what to recommend here. Maybe double check with other renderers just to make sure the problem is isolated to madVR or not. Try other audio renders. Try disabling the "present several frames in advance" feature.

According to Sony DRC team's chief engineer comment, the material library of thousands video image patterns are built into the XCA7 engine. I don't know the details how the process is done. But I guess it could be pattern-matching process similar to image recognization to identify the image content's characteristsics. If I am not wrong, the traditional ways only rely on temporal multi-image information to restore the lost detailed contents?
I don't really know. But I rather think dithering should not harm the Sony DRC algorithm more than the quantization noise you'd get from *not* dithering.

nevcairiel
9th July 2014, 17:59
One thing that was never all that clear to me, if I use NNEDI to scale above target (luma only), does it also always double chroma (with the image scaling option) or does it scale to output target directly?

6233638
9th July 2014, 18:43
One thing that was never all that clear to me, if I use NNEDI to scale above target (luma only), does it also always double chroma (with the image scaling option) or does it scale to output target directly?I think what happens is:

1. Chroma is upscaled to the luma resolution using your selected chroma scaling algorithm
2. NNEDI3 image doubling doubles the luma resolution
3. Either NNEDI3 is used to double chroma to match, if it is enabled, or the luma upscaling algorithm is used instead
4. Your selected downscaling algorithm scales the image to fit the output resolution (luma/chroma)

I think this is probably the best way to handle it.

Shiandow
9th July 2014, 19:11
I'm not 100% sure if those last two steps are correct. If you're not using NNEDI chroma doubling then an other possibility is the following:

3. Your selected downscaling algorithm scales only the luma channel to fit the output resolution
4. Use luma upscaling algorithm to scale chroma to the output resolution

This would be faster, and might explain some of the problems discussed in this thread (http://forum.doom9.org/showthread.php?t=170769). It's especially suspicious that the doubled images in that thread don't look as bad as the ones that have been downscaled.

huhn
9th July 2014, 19:27
I'm not 100% sure if those last two steps are correct. If you're not using NNEDI chroma doubling then an other possibility is the following:

3. Your selected downscaling algorithm scales only the luma channel to fit the output resolution
4. Use luma upscaling algorithm to scale chroma to the output resolution

This would be faster, and might explain some of the problems discussed in this thread (http://forum.doom9.org/showthread.php?t=170769). It's especially suspicious that the doubled images in that thread don't look as bad as the ones that have been downscaled.

i think that was the case if i remember correctly.
i think Madshi explained it in avsforum.

edit:
http://www.avsforum.com/forum/26-home-theater-computers/1477339-so-you-ve-built-your-htpc-now-what-next-how-get-ultimate-picture-sound-quality-your-htpc-madvr-svp-xbmc-mediabrowser-jriver-18.html#post24529747

http://www.avsforum.com/forum/26-home-theater-computers/1477339-so-you-ve-built-your-htpc-now-what-next-how-get-ultimate-picture-sound-quality-your-htpc-madvr-svp-xbmc-mediabrowser-jriver-18.html#post24531007

dansrfe
9th July 2014, 21:48
There are a few things you need to do, and that you can check to make sure madVR is using your dGpu.

If you're using Media Player Classic Home Cinema make sure to rename the executable for it from mpc-hc.exe to something else like mpc-hc2.exe, mpc-hchd.exe, mpc-hcnv.exe (pretty much anything aside from the original name will work).

Go into Nvidia Control Panel (I'm guessing you have an Nvidia/Intel system, but if that's not the case you'll need to go into the driver settings area for whatever GPU you have). Under manage 3d settings switch to the Program Settings tab. Select Add and navigate to your renamed mpc-hc file. Change the area under perferred graphics processor for this program to the High-performance Nvidia processor.

Open a video. You should notice an improvement in your render times. Exit the video and open your registry. Navigate to HKEY_CURRENT_USER\Software\madshi\madVR\OpenCL if it has correctly detected your dGPU it should have it listed as a key inside that key showing its being used for OpenCL processes. If you don't see your gpu listed there post again and let us know, and we can give you instructions for creating a key that will force it to be added.

Thanks for replying!

I changed the name of the executable, added it to Nvidia Control Panel and selected "High-performance NVIDIA processor" however MPC-HC now freezes at "Opening" for some reason.

madVR's registry entry doesn't show an OpenCL key either.

Anime Viewer
9th July 2014, 22:51
Thanks for replying!

I changed the name of the executable, added it to Nvidia Control Panel and selected "High-performance NVIDIA processor" however MPC-HC now freezes at "Opening" for some reason.

madVR's registry entry doesn't show an OpenCL key either.

Go into the registry : HKEY_CURRENT_USER\Software\madshi\madVR\ create a key and name it OpenCL, and in that OpenCL key add a new string called forceVendor and give it a value of Nvidia. If it works correctly after the next time you open a video it should add a directory (key) named after your Nvidia chip/card inside the OpenCL one.

XMonarchY
9th July 2014, 23:31
What videocard and generally speaking, what minimum system specs would you people advice for an HT build that could handle OpenCL dithering, and NNEDI3 32 neuron Chroma Upscaling along with NNEDI3 32 neuron Luma Quadrupling without any framerate loss/drop for 1080p@60Hz display without any overclocking?

I assume a budget AMD card would be selected for such a task because AMD outperforms nVidia in OpenCL performance. My 3770K @ 4.8Ghz & GTX 780 Ti (1250Mhz / 7500Mhz OC) can only handle 64 neurons in Chroma Upscaling and Luma Quadrupling, but my friend's 3570K @ 4.4Ghz & R9 290 can easily handle 256 neurons with identical settings & 1440p@60Hz display. Also - which drivers would you recommend for an AMD card? If I remember correctly, the latest Catalyst Beta drivers performed worse then the previous Catalyst drivers, but I could be wrong about that.

madshi,
Do you expect 64-bit madVR build to perform significantly better than the current 32-bit build in terms of rendering performance?

What would a game developer have to do to integrate madVR into its engine to render pre-rendered videos, like BINK for .bik files, as far licensing/permission goes? Would something like that be allowed?

What about a hardware device, like a very powerful BD-Player - could a manufacturer purchase a license / have permission to use madVR's rendering engine? I know it is free for HTPC use, but I figured at some point in the future more powerful BD-Player hardware or even HDTV hardware could handle madVR's advanced dithering and chroma upscaling.

madshi
9th July 2014, 23:33
I'm not 100% sure if those last two steps are correct. If you're not using NNEDI chroma doubling then an other possibility is the following:

3. Your selected downscaling algorithm scales only the luma channel to fit the output resolution
4. Use luma upscaling algorithm to scale chroma to the output resolution
That is correct if the target resolution is lower than 2x source resolution.

This would be faster, and might explain some of the problems discussed in this thread (http://forum.doom9.org/showthread.php?t=170769). It's especially suspicious that the doubled images in that thread don't look as bad as the ones that have been downscaled.
In this situation the luma channel is downscaled with linear light and the chroma channel is upscaled in gamma light. I'm wondering if the shaders do something wrong with specific combination that would explain the artifacts? Not sure...

madshi
9th July 2014, 23:36
Do you expect 64-bit madVR build to perform significantly better than the current 32-bit build in terms of rendering performance?
I expect pretty much zero difference because (with a few exceptions) all the work is done on the GPU and the GPU doesn't care whether the CPU code is running in 32bit or 64bit.

What would a game developer have to do to integrate madVR into its engine to render pre-rendered videos, like BINK for .bik files, as far licensing goes? Would something like that be allowed?
I'm not sure if it would be worth the effort. Video playback usually doesn't have to be perfect quality within games. But if a game developer actually wanted to use madVR, I might be willing to consider the idea.

What about a hardware device, like a very powerful BD-Player - could a manufacturer purchase a license to use madVR's rendering engine? I know it is free for HTPC use, but I figured at some point in the future more powerful BD-Player hardware or even HDTV hardware could handle madVR's advanced dithering and chroma upscaling.
It would have to be a fully blown Windows machine because madVR uses Windows APIs and Direct3D9/11. But generally it would be possible. Somewhat unlikely, though, I think...

Shiandow
10th July 2014, 00:18
In this situation the luma channel is downscaled with linear light and the chroma channel is upscaled in gamma light. I'm wondering if the shaders do something wrong with specific combination that would explain the artifacts? Not sure...

I don't think they are doing something wrong. I think that doing half the processing in linear light for only one of the channels is just asking for trouble. Not only that but hoping that an upscaling algorithm yields the same result as a different upscaling algorithm combined with a downscaling algorithm might just be a bit too optimistic.

The reason that linear light is especially bad could probably be explained by considering the way edges are treated in gamma light and linear light:

To leave edges intact it's probably important that the line running exactly through the middle of an edge should remain at the same location. But in gamma light and linear light there's a difference between what value is halfway between the colours on the opposing sides, in linear light it's close to the lighter side, and you'd expect that the middle of the edge should have a value close to this average.

So if you have an image with white text on top of a red background you'd expect that doing the processing in linear light would make the letters slightly smaller than doing this processing in gamma light (since in linear light the line midway an edge is close to the lighter side, hence the letters seem smaller). This creates a small region around the letters where the image is neither white nor red, hence black.

dansrfe
10th July 2014, 02:14
Go into the registry : HKEY_CURRENT_USER\Software\madshi\madVR\ create a key and name it OpenCL, and in that OpenCL key add a new string called forceVendor and give it a value of Nvidia. If it works correctly after the next time you open a video it should add a directory (key) named after your Nvidia chip/card inside the OpenCL one.

For some reason madVR still doesn't render the video. I did a fresh update of the drivers, added the OpenCL key and forceVendor string exactly as described and MPC-HC just plays the audio when the file is opened.

If I attempt to close the player then MPC-HC and madVR processes freeze and need to be killed.