This forum is for users to exchange information and discuss with other users about a TMPGEnc product.
In case you need official support, please contact TMPG Inc.
TMPGEnc Video Mastering Works 5 BBS [ Sorted by thread creation date ]
==--
Are you guys sure that the encoding metrics are reported properly?
I'm getting 1% CPU, 99% CUDA, but all 6 CPUs are at 100% utilization and my geforce 470 is running at 3% utilization.
I know this is possible in theory, if the vector processors are vastly more efficient than the CPU and they both have to process the same frame at the same time. I assume that some tasks can only be handled by the CPU and the GPU might be spending almost all its time waiting for the CPU to catch up.
But isn't this asymmetry kind of, well, ridiculous?
Maybe it doesn't matter because I'll get my edited porn eventually no matter what the metrics say. But if there's an optimization problem somewhere that prevents my GPU from encoding the video at ultra-speed, I want to to fix it, and anyway, I'd like to think that the numbers I'm looking at are accurate.
Does anyone else notice the CPU being maxed and the GPU almost idle, even though it says CUDA is doing most of the work?
TMPGEnc Video Mastering Works 5, if select CUDA, coding on GPU ONLY(!) FILTERS (and yet not all), x264 ALWAYS on CPU! = no 2 pass encoding on PEGASYS
B.
p.s. ...others SW coding x264 Cuda 2pass no problem
If you want to encode with Cuda, you have to select it as the encoder (default is x264), otherwise it will only be used for filters. You cannot use two encoders at the same time.
No 2-pass for cuda encoding though.
However, there is a 2-pass option in the preferences for the x264 encoder.
I have a NVIDEA CUDA enabled graphics card. I noticed that there are three Video Encoder options for encoding to a MPEG-4(AVC)file. The options are as follows:
- CUDA
- Intel Media SDK Software
- x264
I want to take advantage of improved processing acceleration by using CUDA with my NVIDIA graphics card but I also want the file to be encoded using the H.264/AVC encoder. Therefore if I select CUDA, will the output file still be encoded with the H.264 encoder? It is unclear to me since there are no additional options for selecting the H.264 encoder if I select CUDA.
Based upon the description of the "NVIDIA CUDA H.264 Encoding" option, as described on the PEGASYS website (and as shown below), I am assuming that if I select CUDA the file will still be encoded using the H.264 encoder?
NVIDIA CUDA H.264 Encoding.
Not Just for Filtering Anymore.
In addition to CUDA-enabled filter processing and decoding, TMPGEnc Video Mastering Works 5 supports H.264/AVC file output using the CUDA technology provided by NVIDIA. The CUDA technology makes use of the GPU (graphic board CPU) with multi-core parallel processing to process complex calculation problems in a short time. By splitting the processes into parallel elements, and using the massive processing power, the program can run even more effectively. Using this function with a compatible graphic board improves processing acceleration when compared with CPU-only use in most cases.
I'm not quite sure if I understand what you're asking. Each one of those 3 are H.264/AVC encoders; you are selecting the H.264 encoding engine. If you want to use CUDA for H.264 encoding, then select CUDA and it will use CUDA's H.264 encoder.
Selecting the other two will use them as the H.264 encoder, and CUDA will only be used for filter processing.
From my recent reading about CUDA 2/2013 there is a serious trade-off in image quality for the limited speed gained by using CUDA. If you know of actual image quality improvements, then please post to this thread or start one about it detailing exactly how to do this.
I've yet to read where CUDA is improving image quality at the high-end of HD.
I'll ask the same about filters too... does anyone know which CUDA filters equal or improve image quality over CPU vs GPU/CUDA saving time but getting a lower quality image.
I'll also mention, the CUDA SDK is up to version 5.0+ at the NVIDIA website, but I think TPMG5 is only using version 3.1 SDK CUDA as of 2/2013... anyone know otherwise?
Anyway, please report about image quality improvements using CUDA and saving encoding time. Which filters specifically... which NVIDIA card you need to do it too... I'm currently very skeptical about CUDA having read a few detailed articles of recent CUDA software reviews. That's 3-4 years now of development time without much gain, if any, in high-quality output from the reviews I read.
I have tried convert a AC3 audio file to PCM audio file but when PCM arrive to 4GB Tmpgenc show a error messaje: File Exceeds 4GB... And conversion is finished. Why?.
Standard WAV files are limited to 4GB because they are a 32bit format.
Only 64bit formats can go above 4GB in file size. These 64bit formats are typically only used in the much more expensive professional editing softwares.
New update! New output templates for PS Vita and the new iPad, plus they updated the x264 encoding engine.
It also seems to fix errors I was getting with the latest NVIDIA drivers installed.
at list they fixed those error with nvidia 306.23
My wish is more plugins and please add video stabilization like "MERCALLI", no one make plugins for this software so i have to use another software
I use tmpgenc all the time and I'm always eager to get the latest fixes and features. Unfortunately, pegasys doesn't bother to alert you when an update exists. And for some bizarre reason, you can only check for updates while you're editing a project.
Do not access or download anything from the links posted by this person. These links guides you to malware infected software.
This criminal has been posting these links to tons of forums with claims about free video editing software for the past 6-12 months.
What you actually get is spyware infected files. This spyware records everything you type on your keyboard, such as passwords and credit card infos, and then sends it to him and before you know it, he has hacked your bank account and stolen all your money.
Currently decoding & encoding in TVMW 5.0 can be accelerated by using CUDA. While gaining the speed, CUDA processing is not that great in terms of quality to compare with CPU only. Hardware acceleration with similar to CPU processing quality can be achieved by using Intel Quick Sync technology. There are currently only several programs that can use this technology, like CyberLink Media Espresso 6 etc.
Would be great if TMPGEnc VMW 5.0 is one of those programs too!..
Check Preferences->Input/output format list->Intel Media SDK (Hardware). You should see "+" for Input and Output. If not, make sure the onboard GPU is active (sounds like you are also using a discrete NVIDIA GPU). You should see the "Intel HD Graphics" icon in the system tray. Normally you need to have a monitor plugged into the onboard GPU for this to be active. Motherboards that support Lucidlogix's Virtu technology can activate the onboard GPU with a single monitor connected to the discrete GPU. That said, I was not able to get this working with TVMW5, so I uninstalled the Virtu drivers, and plugged a "VGA Dummy" into the onboard GPU. I purchased one from EBay, but you could probably make one for a few dollars. Hint, you can setup your multiple displays so that the "fake" monitor touches the real ones in a corner, that way you can still do the Windows 7 drag and snap functions.
I purchased TVMW5 for the sole purpose of encoding 1080i STB captures to 720p using QuickSync, and it does a great job IMO. One-hour programs take 30 mins, about 4X faster than the single-pass x264 software encoder, with comparable quality (if you give it enough bandwidth). This is on an i5-3570K under Win7. TVMW5 seems to be finicky on the Intel driver version though, I remember having to re-install an older driver when an Microsoft's auto-update pushed a new driver onto the machine.
Intel needs to do a better job promoting Quick Sync support in video tools, it really is a killer feature in Sandy/Ivy Bridge CPUs, and like you mention, there are few applications that utilize it reliably at the moment.
Ironically, I came to this forum looking for CUDA issues. I just got a GTX 670, and while CUDA is recognized, it is not utilized (during encoding, load is split CPU/GPU: 100%/0%, which wasn't the case before. I am using the GEFORCE 306.23 driver from last week. Reverted to 301.42, same results.
Thank you for a great answer! I have a i7-2600K processor on a Z68 motherboard with a discrete video card GTX 570 (now a single monitor is plugged into it...) I don't know, what format 1080i to what format 720p you are trying to convert... I have a great success using CUDA decoding and encoding only AVCHD files (CPU/CUDA up to 30%/70%). When I deal with MPEG-2, CUDA does not help.
I will try to switch my monitor to the MB video output just to check if it works. Do I have to also adjust BIOS for Quick Sync to become active? Then if it works I might connect a second monitor (I don't have one...) Never heard of VGA-Dummy though...
Most motherboards will by default auto-enable the onboard GPU if a PEG device (PCI Express Graphics) is not found. Since you have a discrete GPU, you might want to force it on in BIOS. Then again, plugging the only monitor into the onboard GPU might override this. The BIOS may also have a RAM setting for the onboard GPU. In my experience, this doesn't have any affect and is not reflected in the Intel GPU driver information, but if you got lots of memory, it doesn't hurt to maximize it.
I mis-read the encoding Status display in TVMW5. The CPU/GPU load shown is only for the filters, not the encoding stage. Since the only filters I have are the default resizer/deinterlacer, neither of which are CUDA-accelerated, the load will always show 0% for CUDA. To test this, I added a Sharpness @100 (to make it obvious), and the status now reads:
... Pre-processing (CPU=15.4%, CUDA=84.6%)
Instead, I tried monitoring GPU load with GPU-Z. With CUDA encoding, it registers about 10% on the GTX 670 (about 4% for the Sharpness filter mentioned above). Tried it again with a relatively weak GT 430, and the GPU load was 60%, which makes me think TVMW5 does not utilize all the CUDA cores available.
My source files are H264 in a MPEG-TS container (HW:Hauppauge HD-PVR, SW:SageTV/MediaPortal), target files are AVC/H264 in either MP4 or MKV containers. I think all the hardware encoders only come into play during H264 compression, probably because CPUs have become fast enough to deal with MPEG-2 with reasonable speed.
I agree, for a given bitrate, QuickSync results in better image quality than CUDA, and is the preferred encoder given a choice.
@rasterX
Thank you for the good explanation. I'm currently evaluating MW5 on my Notebook - equipped with an Intel Core i5-2520M but I cant get QuickSync enabled. MW5 says its available but not usable. My Notebook as no discrete GPU and I've already tested it with external Monitors aconnected.
I have a Sony PMW-EX1 camcorder, which records files in MP4 container using MPEG-2 codec. Currently, if you trying to input those files to TVMW 5.0 only audio part can be seen - video is missing. Also, if the duration of the recording exceeds ~13 min., clips separate due to 4GB limit of FAT32 formatted cards used by the camera.
Would be great, first, to be able to input XDCAM EX MP4 files, and second, merge those separate chunks of 4GB or less in size back into 1 long file seamlessly. This could be achieved with the help of Sony XDCAM Browser (it does the merge) if it is launched within TVMW 5.0.
Is it possible to do a multicam edit with this software. moving from cam to cam files on separate layers. I have only seen demos do only simple things.
If you have two angles, one angle per layer, you'd have to edit the top most layer to remove the sections where you want the lower layer to show.
Easiest way to do this is to split the top clip in the clip editor at the points you want to cut to and from the other angle; it is important to use the split clip function instead of the cut/delete tool. When exiting the editor, select to keep all split sections as clips.
Back in the timeline, remove the sections where you want the lower layer to show.