DI comparison

GJeffrey wrote on 3/9/2017, 8:45 PM

I compare the few different possibilities offered in Vegas to create DI.

Computer Spec: Cpu 3930k - GPU GTX570 (driver 376.33) - 32Gb Ram - Footage on 90% empty spinning HDD.

Footage: Gopro h264 420 8bits footage - 1080-30p

Project settings: 1920x1080-30p - 8bits pixel format

Preview @ Best Full

To compare the result:

  • I save a snapshot to a png file for the original video and each video encoded with the different codec.
  • Import the encoded video snapshot in Photoshop
  • Add a new layer with the original video snapshot
  • Change layer blend to difference
  • Add a new adjustment layer and increase the gamma to +12.

 

Codec used:

  • ProRes 422HQ & 444XQ
  • Cineform 422 & 444 both in filmscan 2 quality
  • XAVC-I 1080-30p template
  • HDCAM SR 422 & 444 1080-30p template
  • DNxHR 8bits, 12bits and 444
  • MagicYUV RGB, YUV422 & YUV444
  • Sony YUV 8 & 10bits

 

Results:

  • ProRes 422HQ

  • ProRes 444XQ

  • Cineform 422

  • Cineform 444

  • XAVC-I

  • DNxHR 8bits

  • DNxHR 12bits

  • DNxHR 444

  • HDCAM SR 422

  • HDCAM SR 444

  • MagicYUV RGB

  • MagicYUV YUV422

  • MagicYUV YUV444

  • Sony YUV 8bits

  • Sony YUV 10bits

  • Original footage

 

File size and preview speed

  • HDCAM SR 444 is buggy. There is a black line at the bottom. Not possible to use it.
  • ProRes encoder has duplicated 1st frame and the end is not quantize to frame. Encoding result is good though with reasonable file size. Not possible to use it until the bug previously mentioned is solved.
  • DNxHR is the worst. Blocking and many artifacts. Use quicktime in the background which will lead to memory issue with many events on the timeline. Not advise to use.
  • Cineform has surprisingly blocking with 422 and problem with the red channel in the midtone.
  • Cineform 444 gives good result.
  • XAVC I is average.
  • MagicYUV RGB is really lossless as advertised (only in 8bits project with Vegas)
  • MagicYUV YUV422 is not lossless and give overall good result with some blocking artifacts.
  • MagicYUV YUV444 is not lossless and gives better result than YUV422 with fewer blocking artifacts.
  • Sony YUV 8bits is lossless. File size is huge. Might need SSD for real time playback.
  • Sony YUV 10bits is very good with few blocking artifacts. Might need SSD for real time playback.

 

I was always using Cineform as DI before I doubt it's the best solution right now.

ProRes seems pretty good (and is better than Cineform with smaller file size) but is buggy.

MagicYUV RGB is perfect but with huge file size.

XAVC-I seems to be a good balance between file size & quality.

What you guys will advise to use a DI?

Comments

CogDiv wrote on 3/9/2017, 8:59 PM

Nice comparison. HDCAM SR 422 looks better than XAVC-Intra but you didn't mention it in your analysis . . . Thoughts?

Maybe I need to look into MagicYUV . . .

Wait a minute, are you ONLY using the GTX570? If that's the case, isn't it only accelerating your timeline and processing OFX? If it is your only GPU in the system then you probably are not achieving the CUDA encoding acceleration.

Musicvid wrote on 3/9/2017, 9:34 PM

Add a new adjustment layer and increase the gamma to +12.

Why? To exaggerate the differences?

The only codec you chose that is not always Y'CbCr is Magic yuv. Begs the question, did you test it as RGB or YUV? It appears you ran it as RGB which of course invalidates your other comparisons. Was there something wrong with including Sony YUV, the gold standard of compressed DIs?

To do this properly, use belle-nuit as your source and DON'T adulterate the results. Then you can use the vectorscope as additional evidence of chroma subsampling inconsistencies, which won't show well with your grey-blue keyed example. Of course, any hw acceleration degrades the results, so leave it completely disabled.

A quantifiable measure of what you are attempting to do is called PSNR. Read about it.

OldSmoke wrote on 3/9/2017, 9:37 PM

If it is your only GPU in the system then you probably are not achieving the CUDA encoding acceleration.

Wrong! Only with the GTX570 will you achieve CUDA encoding!

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

CogDiv wrote on 3/9/2017, 9:40 PM

@OldSmoke So VEGAS drops the T/L acceleration and OFX processing to run the CUDA encoding routines, or all work together on a single GPU?

OldSmoke wrote on 3/9/2017, 9:52 PM

CUDA encoding is ONLY done when rendered to MC AVC with the option CUDA enabled. However, if the GTX570 is the only card in the system, Vegas will use the cards OpenCL capabilities too for applying all the other work. Look at rendering as a two stage work, stage one is the same as timeline preview and it's OpenCL only, part two is rendering and that can be CUDA (Fermi cards and earlier ones), OpenCL (AMD) or CPU only.

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

astar wrote on 3/9/2017, 9:53 PM

Star Wars: Phantom Menace was shot on HDCAM SR 444. I didn't notice the black bar during the movie. hmm.

Musicvid wrote on 3/9/2017, 10:00 PM

Comparisons of software codecs using any hardware acceleration are invalid, as would be introducing any uncontrolled variable.

GJeffrey wrote on 3/9/2017, 10:04 PM

Phantom Menace was shot on HDCAM SR 444. I didn't notice the black bar during the movie

Maybe a Vegas bug?

OldSmoke wrote on 3/9/2017, 10:05 PM

Comparisons of software codecs using any hardware acceleration are invalid, as would be introducing any uncontrolled variable.


I sort of agree with that. Is the OP using SVP13 or VP14?

Last changed by OldSmoke on 3/9/2017, 10:06 PM, changed a total of 1 times.

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

GJeffrey wrote on 3/9/2017, 10:06 PM

HDCAM SR 422 looks better than XAVC-Intra but you didn't mention it in your analysis . . . Thoughts?

HDCAM SR 422 file is bigger than Magicyuv which is lossless. Better use the later

GJeffrey wrote on 3/9/2017, 10:09 PM

Is the OP using SVP13 or VP14?

Obviously VP14 to use ProRes encoding capabilities.

Encoding was made with GPU off. Only turn it on for preview purpose.

OldSmoke wrote on 3/9/2017, 10:11 PM

Just tested HDCAM SR 444 and yes, there is a small bar at the bottom.

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

GJeffrey wrote on 3/9/2017, 10:16 PM

Just tested HDCAM SR 444 and yes, there is a small bar at the bottom

Is it a Vegas bug?

CogDiv wrote on 3/9/2017, 10:18 PM

@OldSmoke I forgot about only one codec utilizing CUDA. I could have sworn I saw an acceleration option on one of the Sony/VEGAS codec preference panels . . .

Musicvid wrote on 3/9/2017, 10:19 PM

Just why all the incessant talk about CUDA?

NVIDIA dumped it some time back for NVENC, which for some pretty good reasons is not GPU-centric.

OldSmoke wrote on 3/9/2017, 10:20 PM

Just tested HDCAM SR 444 and yes, there is a small bar at the bottom

Is it a Vegas bug?


Not sure... would require more testing.

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

GJeffrey wrote on 3/9/2017, 10:31 PM

The only codec you chose that is not always Y'CbCr is Magic yuv

HDCAM SR 444, Cineform 444 and ProRes 444 are also RGB

OldSmoke wrote on 3/9/2017, 10:32 PM

NVENC is actually what broke the Mainconcept encoder.

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

CogDiv wrote on 3/9/2017, 10:33 PM

@Musicvid So why didn't you say something earlier? Thanks for the reference.

john_dennis wrote on 3/9/2017, 10:36 PM

As a budding shade tree scientist, I'm skeptical of all the machinations in Photoshop to amplify the difference in the output files.

"Just why all the incessant talk about CUDA?" 

If I could run a 128 core CPU and one of these...

... I think I'd be much happier.

Musicvid wrote on 3/9/2017, 10:48 PM

Musicvid So why didn't you say something earlier? 

 

Actually, I did say something earlier. And you dismissed it out of hand...

CogDiv wrote on 03/08/2017, 11:11 PM

And BTW, CUDA is not on its way out, as Nvidia has made significant advances...

 

CogDiv wrote on 3/9/2017, 10:53 PM

My input into Vegas is a steady Apple ProRes 4:2:2 HQ due to my recorder. So, I'll still need the accelerated encoder for h.264, but at least now I know I could hardware accelerate conversion to other formats from there if I needed to using FFmpeg. The quality issues I'll have to read more about.

CogDiv wrote on 3/9/2017, 10:55 PM

@Musicvid That's because you keep implying it replaced CUDA, and it doesn't. NVDEC & NVENC are specialized components within the GPU now it seems, but the CUDA cores are still there as well, even if they are not used for video decoding and encoding routines.

GJeffrey wrote on 3/9/2017, 11:11 PM

@Musicvid, Cogdiv,

if you want to fight about cuda or anything else OT, create another post.

Thanks