CloneBD results with video card hardware acceleration||Post HERE

[gameowl]

Saw that the new CloneBD beta was out here on the forums that does video hardware acceleration. So I right away downloaded the new CloneBD Beta and tried it out. Take note I'm using a Mechanical Drive at 7,200RPM SATA 3. and the most important thing is I'm using a GeForce GTX 980TI 6GB video card to transcode.

My results were transcoding a 45.4GB ripped with AnyDVD HD, off the store shelf bluray movie with NON video card hardware acceleration strictly my CPU doing the transcoding was one hour and 19 minutes to fully transcode it to fit on a 25GB bluray disc (results do not include bluray burn to disc time)

Now using my video card hardware acceleration on highest quality took me 31 minutes to transcode same movie. Very impressive. On fastest speed it took with the same movie 24 minutes to transcode. I had no errors or problems with the transcoding process, movie played on PowerDVD 14 Ultra fine. So I'm happy to say, this is a great feature Elby put in for CloneBD, and who knows it might even get faster down the road with future updates. I'm interested to hear everybody else's results, using video card hardware acceleration, please state your exact video card model. What speed/quality method you used to transcode in CloneBD options, what type of format where you encoding and how much data was it before transcoding. And of course your transcoding finish time.

 

[Ch3vr0n]

I posted some of my results in the sticky itself as a reference point for some different output types.

 

[gameowl]

I posted some of my results in the sticky itself as a reference point for some different output types.

yes I saw, thanks for the heads up. CloneBD is amazing. Elby has some really great coders. And RedFox's AnyDVD HD is the king of bluray ripping. WIth those two side by side, it's just an awesome combination. Proud to be a Lifetime CloneBD member, and next year I'm renewing my one year Red Fox AnyDVD HD to lifetime.

 

[Sabertooth]

Testing CBD 1.0.9.Xbeta. Here's my results so far for the same content run through over and over again with various configurations.

In-Common Configuration for all tests:

Hardware:
Intel i7-6950x with 32GB DDR4-3200
Zotac Geforce GTX 1080 Amp! Edition
Source Drive: Intel Series 535 SATAIII 480GB
Destination Drive: Samsung 950 Pro m.2 512GB

Software:
CloneBD 1.0.9.0beta unless indicated as 1.0.9.1beta, 1.0.9.2beta, or 1.0.9.3beta
AnyDVD HD 8.0.6.0
Virtual CloneDrive 5.5.0.0
nVidia Driver 375.86 or 375.95

Content Source Material:
The Legend of Tarzan encrypted ISO (pISO)
mounted with Virtual CloneDrive 5.5.0.0
decrypted on-the-fly by AnyDVD HD 8.0.6.0
transcoded or re-encoded with CloneBD 1.0.9.Xbeta
Main title size: 28.77 GB/158214 frames

CloneBD settings:
Variables:
New settings including
Encoder Quality Bias: I prefer Highest Quality but also tested with Highest Speed for comparison with CBD 1.0.8.8.
Decoder: CUDA or Software
Encoder: CUDA or Software
Copy Mode: MP4 File container h.265/HEVC w/ AAC5.1 or Partial Disk/Folder/ISO to BD-R Folder with keep original audio and direct play or Complete copy to BD9 ISO with downconvert HD audio

The tests:

Test 1 PROBLEM (workaround introduced from 1.0.9.1)
Copy Mode: MP4 File container using Software Decoder and Software Encoder at Highest Quality h.265/HEVC transcode
Result: FAILED (already reported for all version since CBD 1.0.8.4 in July 2016) 0fps

Test 2
Copy Mode: MP4 File container using CUDA Decoder and CUDA Encoder at Highest Quality h.264 transcode
Result: 2.50 GB completed in 07:58 or 331 fps
Test 2 Rerun (1.0.9.3beta/nVidia Driver: 375.95)
Result: 2.34 GB completed in 07:59 or 330 fps

Test 3
CBD 1.0.8.8 comparison test
Copy Mode: Partial Disk/Folder/ISO to Folder using Software Decoder and Software Encoder at Highest Speed re-encode
Result: 23.32 GB completed in 08:55 or 296 fps

Test 4 (Back to CBD 1.0.9.0beta for the rest)
Copy Mode: Partial Disk/Folder/ISO to Folder using Software Decoder and Software Encoder at Highest Speed re-encode
Result: 23.33 GB completed in 08:45 or 301 fps

Test 5
Copy Mode: Partial Disk/Folder/ISO to Folder using CUDA Decoder and CUDA Encoder at Highest Speed re-encode
Result: 23.09 GB completed in 05:27 or 484 fps

Test 6
Copy Mode: Partial Disk/Folder/ISO to Folder using CUDA Decoder and CUDA Encoder at Highest Quality re-encode
Result: 23.00 GB completed in 08:22 or 315 fps
Test 6 Rerun (1.0.9.3beta/nVidia Driver: 375.95)
Result: 23.16 GB completed in 08:23 or 315 fps.

Test 7 Mixed
Copy Mode: Partial Disk/Folder/ISO to Folder using Software Decoder and CUDA Encoder at Highest Quality re-encode
Result: 23.05 GB completed in 08:07 or 325 fps

Test 8 Mixed (1.0.9.1beta)
Copy Mode: Partial Disk/Folder/ISO to Folder using CUDA Decoder and Software Encoder at Highest Quality re-encode
Result: 22.90 GB completed in 03:55:03 or 11 fps. Processor utilization only 33%.

Test 9 (1.0.9.1beta) PROBLEM (log here)
Copy Mode: Partial Disk/Folder/ISO to Folder using Software Decoder and Software Encoder at Highest Quality re-encode
Result: 22.82 GB completed in 03:37:50 or 12 fps. 34% CPU utilization is awful. Doesn't seem right.

Test 10 PROBLEM (workaround introduced from 1.0.9.1)
Copy mode: MP4 File container using CUDA Decoder and CUDA Encoder at Highest Quality h.265/HEVC transcode
Result: FAILED (already reported for all version since CBD 1.0.8.4 in July 2016) 0fps

Test 11 Processor Affinity set to 16 cores due to CBD bug.
Copy mode: MP4 File container using Software Decoder and Software Encoder at Highest Speed h.265/HEVC transcode
Result: 2.28 GB completed in 28:42 or 92 fps.
Test 11 rerun with 1.0.9.1beta, processor affinity setting no longer necessary however processor utilization limited at only 65%
Result: 2.28 GB completed in 28:35 or 92 fps

Test 12 (1.0.9.1beta)
Source material: Torchwood Series 1 D4 pISO
Source Drive: Hitachi 2TB 7200rpm SATA III HDD
Copy Mode: Complete copy to BD9 ISO using CUDA Decoder CUDA Encoder at Highest Quality re-encode
Result: 23.10 GB to 8.16 GB in 13:01 (as this is a complete disk copy, fps measure is not possible)
Test 12 Rerun (1.0.9.2beta) PROBLEM (log here) (fixed under 1.0.9.3beta)
Result: TRANSCODER FAIL
Test 12 Rerun 2 (1.0.9.3beta/nVidia Driver: 375.95)
Result: 23.10 GB to 8.40 Gb in 12:57

Test 13 (1.0.9.2beta)
Copy mode: MP4 File container using CUDA Decoder and CUDA Encoder at Highest Quality h.265/HEVC transcode
Result: 2.50 GB completed in 08:50 or 299 fps. Smokin'!!
Test 13 Rerun (1.0.9.3beta/nVidia Driver: 375.95)
Result: 2.34 GB completed in 08:59 or 294 fps.

Test 14 (1.0.9.2beta) PROBLEM (log here) (fixed under 1.0.9.3beta)
Source material: Torchwood Series 1 D2 pISO
Source Drive: Hitachi 2TB 7200rpm SATA III HDD
Copy Mode: Complete copy to BD9 ISO using CUDA Decoder CUDA Encoder at Highest Quality re-encode
Result: TRANSCODER FAIL
Test 14 Rerun (1.0.9.1beta)
Result: 24.23 Gb to 8.48 GB in 16:19 (definitely a problem introduced with 1.0.9.2beta)

Test 15 (1.0.9.2beta) PROBLEM (log here) (fixed under 1.0.9.3beta)
Source material: Torchwood Series 1 D3 pISO
Source Drive: Hitachi 2TB 7200rpm SATA III HDD
Copy Mode: Complete copy to BD9 ISO using CUDA Decoder CUDA Encoder at Highest Quality re-encode
Result: TRANSCODER FAIL
Test 15 Rerun (1.0.9.1beta)
Result: 22.62 Gb to 8.36 GB in 12:06

Test 16 (1.0.9.2beta) PROBLEM (log here)
Source material: Torchwood Series 1 D4 pISO
Source Drive: Hitachi 2TB 7200rpm SATA III HDD
Copy Mode: Complete copy to BD9 ISO using Software Decoder Software Encoder at Highest Quality re-encode
Result: 23.10 GB to 8.12 GB in 03:23:06 (27% CPU usage)

Test 17 (1.0.9.1beta)
Source Material: Ghostbusters 2016
Copy Mode: PartialDisk/Folder/ISO to Folder Main title (167712 frames) using NO COMPRESSION (minus one subtitle), direct play and keep original audio.
Result: 23.66 to 22.90 completed in 04:10 or 671 fps (max. system output)

Test 18 (1.0.9.3beta/nVidia Driver: 375.95)
Source material: Torchwood Series 1 D1 pISO
Source Drive: Hitachi 2TB 7200rpm SATA III HDD
Copy Mode: Complete copy to Custom (BD9) ISO using CUDA Decoder CUDA Encoder at Highest Quality re-encode
Result: 22.55 GB to 8.36 GB in 13:16

Test 19 (1.0.9.3beta/nVidia Driver: 375.95)
Copy Mode: Complete copy to Custom (BD9) ISO using CUDA Decoder CUDA Encoder at Highest Quality re-encode
Result: 29.94 GB to 8.26 GB in 11:03

The CUDA encoder with the GTX 1080 clearly bests(beats) the Software Encoder even on my i7-6950x 10 core system.

Correction: Under 1.0.9.0beta, H.265 did not complete successfully on my system with the CUDA decoder encoder as originally stated. The settings had reverted to the default h.264 so the encode was h.264 rather than h.265. Workaround introduced under 1.0.9.1 and CUDA HEVC encoding under 1.0.9.2

 

[Adbear]

I'm also finding it a lot quicker. I have my rips on an SSD on my high end laptop. Without GPU acceleration on average they take around 28-30 minutes using a 6700K CPU encoding to an MKV. With the GPU acceleration using a laptop GTX 1060 they take between 7-8 minutes with encoder quality set to highest quality

 

[Adbear]

I also found that if I set it to decode using the CPU it was around 40fps slower so I've set it to decode and encode using the CUDA on the GTX 1060

 

[Sabertooth]

I also found that if I set it to decode using the CPU it was around 40fps slower so I've set it to decode and encode using the CUDA on the GTX 1060

I received an even narrower margin in the other direction with my i7-6950x 10 core (20 thread) CPU of around 15 fps faster when using the Software decoder and the CUDA encoder with a GTX 1080 compared to the CUDA used for both decode and encode. I did note, however, that the CPU utilization was only in the 40% range during processing. Using the Software encoder, however, is a lot slower. I didn't complete the tests yet (maybe tomorrow) but it was looking like over 2 hours for the Highest Quality encoder bias using the Software encoder.

 

[Pete]

I also found that if I set it to decode using the CPU it was around 40fps slower so I've set it to decode and encode using the CUDA on the GTX 1060

Yes, it is important to understand, that when mixing different methods for decode and encode, each picture needs to be scooped out of GPU1 and fed into GPU2.

Similarly, if you use software decoding, you have to feed full decoded pictures from system memory to the graphics card, whereas when transcoding through the same GPU, you only feed the compressed input.
This transferring data from system memory to GPU memory and vice versa costs a lot of time.

Therefore, in the final version, there will be no option to mix adapters or activate acceleration for only decoding or only encoding.
Though it will still happen - HEVC encoding is currently not implemented, and when it is, still many graphics cards won't support it, so CloneBD will have to automatically fall back to sw encoding while decoding with acceleration.

 

[Ch3vr0n]

I merged both topics together and stickied them to simply results for elby. Lets stick to 1 topic.

 

[Sabertooth]

Therefore, in the final version, there will be no option to mix adapters or activate acceleration for only decoding or only encoding.

As my result appears to differ from Adbear's, I would hope to keep the option intact. Though the difference is only small at a 4.5% improvement, it's still an improvement and saved me a quarter of a minute. I've retested several times and the result is consistent.

 

[Ch3vr0n]

The CUDA encoder with the GTX 1080 clearly bests the Software Encoder even on my i7-6950x 10 core system.

euhm no, CUDA is hardware (your GTX)

 

[Sabertooth]

euhm no, CUDA is hardware (your GTX)

Sorry but you're not making sense? Let's substitute CUDA = Hardware, The hardware encoder with the GTX 1080... works fine for me and it was necessary to make it clear what the hardware is.

 

[Ch3vr0n]

nvm, i read "gtx 1080 best software encoder". overlooked that extra s at the end

 

[Sabertooth]

"gtx 1080 best software encoder"

OIC, adjusted for clarity.

 

[Bubbator]

Hi,
I have:
1x nvidea gtx 680 stock clock
cpu 3930k oc to 4.3ghz
mem 4x ddr3 8gB @667mhz
asus x9p79 deluxe
samsung evo 840 1tB
win 10 pro v1607 64 bit

settings: hardware accel on -both
I just processed 2 blu rays to 25gB and iPad2 wifi mp4.
(using 1.0.9.0)The first was Batman: The Killer Joke. AnyDVD image ripper hung at 100% so I just mounted the image with VCD and proceeded anyway. I usually get 135-150 fps. for disc copy and I got 136 for most of the main feature. One featurette blew by at 5000 fps, the rest about 135 fps. BTW The time remaining started at 2 hrs. and change until about 30 minutes in when 90 minutes vanished from time remaining at the end while processing the warning in a zillion langues for about the same 30-40 minutes I would have expected
without accel. The 1.0.9.1 beta was available after I finished. The mp4 created ~150fps 33 minutes again as I would have expected before.
The second disc was The Man Who Knew Infinity which has no extras. I used 1.0.9.1 (too late for Batman). CloneBD of course found corrupt data just to mess us up. The results were similar to Batman--no noticeable speedup.
Sorry I didn't run them both ways but the big difference just wasn't gonna be there.
I will upload Log file for Batman.
I will post again when I get another disc and run it old way and new.
B

 

[Pete]

I will upload Log file for Batman.

Since this is about CloneBD, Please upload a CloneBD log file.
(Press the Alt-Key on the completion page for the log buttons to show up).

We need to see, whether hw acceleration was really active.

 

[antipodes]

with regard to the information required for this thread do you require only the Clone BD log file or additional commentary as well?

 

[Adbear]

I think it was on Bubbator's they needed the logfile for so they can see why it doesn't seem to be accelerated

 

[Bubbator]

working on other movie for second logfile...
B

 

[Pete]

working on other movie for second logfile...
B

Ok, Cuda appears to be active.
Maybe that's all you can get out of that GPU.
The GTX 680 is from 2012 with the Kepler chip set (and was a race horse back then) - quite possible, that it's at its limits here. Gaming speed, btw., is not so relevant.