I would not worry about your system failing to go past region 5 with the XDCAM HQ 1920x1080-60i, 35 Mbps Render template, on my system it hangs around 51% mark in Vegas 15 and 53% mark in Vegas 16
I also use the studio driver. I even keep it in gaming although it is maybe not optimized for specific gaming titles. But I prefer stability and a silent PC above all else, even in gaming!
Yess, but my OBS settings to make that screencapture through QSV is on my laptop configured to record no audio, because IMO to show something on video the audio is not important for others.
BTW. Also rendering to XDCAM EX HQ 1920x1080-60i, 35 Mbps with the Nvidia for accelerated video freezes also in the middle of that section 5, while if there no GPU or the QSV is choosen it renders good: QSV in 70 sec and with no GPU 192 sec.
When I tested last year with my new build I was using the older type intel driver. My Qsv render times then were 15s with a very small OC. With the newer Intel drivers this changes to about 48s, no overclock...The Rhino may have achieved similar but overclocked to 5/5.1?
I use the Intel QSV driver 25.20.100.6373 to achieve 14s on the Red Car Test in combination with one of the newer VEGA 64 drivers. I have monitors connected to both the onboard Intel & Vega 64 and both GPUs show utilization during renders. Apparently Vegas 15 & 16 have not been updated to take full advantage of the newer Intel DCH drivers, so maybe this will be fixed with Vegas 17... My 9900K is overclocked to 5.0 ghz at 1.3V.
AMD's new $750 USD 16-core 3950X, to be released in September, will outperform the 9900K & AMD 1950x/2950x and therefore Threadripper CPU prices have hit bargain-basement prices... For instance, our local Microcenter has the Threadripper 1950X on sale for $329, a CPU that listed for $1000 a year ago... Other Zen2 CPUs will be available starting next week, so we'll start seeing how those perform...
@bitman & @j-v - I'm also not particulary worried by the Render, but it does identifying an issue. More concerning to me is that I can't Preview in anything HIGHER than Preview Half or Preview Good, without failing. Now THAT's the issue. Also, the CPU and GPU don't get above 6%. Also switching off GPU gets me staggery Preview.
Note - If you update or revert to older GPU drivers, you need to do a "clean install" of the driver which totally gets rid of the current drivers before installing the new... Also, try setting your Vegas RAM preview to 200 and then 0 to see if either helps... IMO putting together the hardware to build a new PC is the easy part. Installing the correct drivers, applications, updates, etc. is the hard part. I backup my entire hard drive as images throughout the installation process so that I have know working restore points to return to if something later messes things up...
I overlooked that I hadn't the latest studio driver for my PC ... 430.86 I installed it now and the render still freezes using XDCAM template, now at 54%. So if J-V is rendering XDCAM aok, its not the Nvidia Studio graphics driver anyway since i’m using the same one.
...... its not the Nvidia Studio graphics driver anyway since i’m using the same one.
When I use the NVidia for hardwareacceleration on rendering it freezes, when I use there no GPU or my Intell (QSV) it renders good with different times.
Former user
wrote on 7/6/2019, 9:47 AM
Ok, the plot thickens, but good news, @bitman@j-v. I now have both PC and laptop rendering out fully the XDCAM red car. What I did was remove HW Acc first, rendered out ok, then put back HW Acc. and it then rendered out ok! Same action and results for both machines, its as if doing the one render without HW Acc. fixed it?
My worry when I read these reports is that my 9 year old clunker (OK, has modest RX580 GPU upgrade) renders in a similar league - i.e., approx. 30 seconds for XDCAM 35mbit - and preview in this 1080 P project scrubs completely smoothly.
So, opinions?
1) Is the new hardware really just not all that much better?
2) Is this an NVIDIA issue, or Intel QSV issue?
Don't want to go through the hassle and expense just to get "meh" performance improvements.
PS Have you guys agreed to a standard benchmark for 4k projects?
Sorry, I have no problems besides when I render to a for me total not interesting codec as XDCAM EX HQ 1920x1080-60i, 35 Mbps . My renders are always for every use I have in MAVC with the NVEnc help and there it goes very fast for my laptop hardware with 27 sec
“Don't want to go through the hassle and expense just to get "meh" performance improvements.”
Thats a fair point. For 4K you really do get a nice improvement especially in playback, jumping around a timeline etc. When you go greater than 4 cores, everything seems just easier. I recently got Neatvideo 5, and without its HW acc. it would probably bring most older systems to their knees.
IMHO The Red Car project converted to 4K might be the simplest way of going for a 4K benchmark. But it needs to be made available as a downloadable project. Just saying that I converted this and that media to 4K, in the Red Car project, and if you do the same then we can compare times, may only lead to inconsistency.
It would be a large download, but only has to be done once. The render templates to use, ideally just one, should be clearly laid out also, with variations, Cpu, vce, nvenc, qsv. HW Acc on or off.
Former user
wrote on 7/6/2019, 1:12 PM
I mentioned this previously... going forward, as they say, with a 4K benchmark, it would probably be best to start a new thread called say benchmarks. As users post results, with a very specific format, they would be entered into a ss, or chart, or table as soon as practicable. The format would be say user name, render times, PC specs, not all, just a sufficient number of items, gpu, cpu, not more than is necessary. No point in adding in items that would have no or very marginal effect on test results. Memory size and speed, motherboard type, probably not too relevant.
Someone would need to keep this info updated as results come in, not necessarily immediately.
It contains the original project file and pdf with the media file transcoded to a number of 4k formats. Takes a little longer to render but doesn't seem to yield much insight into 4k issues... probably need a real hour-long project for that which would be difficult to package for download.
A lot talks how fast their system are. But I don't see any kind numbers, specifically the Red car 4K benchmark. the 4K benchmark should be strictly using .h264/265. Should not use XDCAM to render because it is only a 1080 resolution unless you have to change. Stay strictly at default template h264/265 as a test bed.
Since, no ones want to show their 4K render in AMD VCE/ NVENC. I'll sart with mine.
As far as I can tell. Render in 1080P. No one can touch my 15s in h264/h265 render with AMD VCE that I known of. QVS is a hair faster but as most of knows, the quality in QVS is garbage vs my AMD VCE.
Nobody's fault, but.... man this is a mess! One guy reports using easy to encode 1080p XDCAM, next uses some hardware accelerated version (VCE, NVENC, QSV) then the last guy says everyone should compete for h264/h265 (not the same!) to 4k, but reports 15 seconds 1080p VCE h264 (or h265? - we can't read your mind you know.)
It's like 100 guys running in different directions over hill and dale crying "I'm the fastest, I'm the fastest." 😀
Nobody's fault, but.... man this is a mess! One guy reports using easy to encode 1080p XDCAM, next uses some hardware accelerated version (VCE, NVENC, QSV) then the last guy says everyone should compete for h264/h265 (not the same!) to 4k, but reports 15 seconds 1080p VCE h264 (or h265? - we can't read your mind you know.)
It's like 100 guys running in different directions over hill and dale crying "I'm the fastest, I'm the fastest." 😀
How it is any one fault? I was never thought about h264/265. All every body care about getting fastest render in 1080P. In this case it was XDCAM for me. I was not trying to hide anything. Othervise I would never have listed for every body to see. Beside the point. XDcam is only 1s faster then h264/265 for me any way. What I said and will absolutely stand by it. Show me who on the forum can touch the number I had clearly said. Sure take a 5.0/5.1 ghz with garbage encoding quality known as QVS to get the 13s . Nothing here to brag about. What I want to see is high quality encode with fast speed is every one care about. If you got so much to said about. Why don't you show some number here before you talk like you know it all.
Nobody's fault, but.... man this is a mess! One guy reports using easy to encode 1080p XDCAM, next uses some hardware accelerated version (VCE, NVENC, QSV) then the last guy says everyone should compete for h264/h265 (not the same!) to 4k, but reports 15 seconds 1080p VCE h264 (or h265? - we can't read your mind you know.)
It's like 100 guys running in different directions over hill and dale crying "I'm the fastest, I'm the fastest." 😀
How it is any one fault? I was never thought about h264/265. All every body care about getting fastest render in 1080P. In this case it was XDCAM for me. I was not trying to hide anything. Othervise I would never have listed for every body to see. Beside the point. XDcam is only 1s faster then h264/265 for me any way. What I said and will absolutely stand by it. Show me who on the forum can touch the number I had clearly said. Sure take a 5.0/5.1 ghz with garbage encoding quality known as QVS to get the 13s . Nothing here to brag about. What I want to see is high quality encode with fast speed is every one care about. If you got so much to said about. Why don't you show some number here before you talk like you know it all.
The reason I ran the XDCAM is because the test project started back then by Hulk who originate the data which is including XDCAM for the test. That every body who participate test also known about.
Bruce, Bruce, Bruce... take a deep breath and read my post again. I clearly said "Nobody's fault..." right at the start of my post.
And I did post my time. See above.
And you were NOT clear which codec you were using, thus ADDING to the confusion.
That is my point: UNLESS WE ALL AGREE TO A TEST STANDARD, THEN THE RESULTS ARE NOT ACCURATE.
I say h264/265 meaning Magix AVC . Magix AVC support VCE. hope to clear this up.
Former user
wrote on 7/7/2019, 1:40 AM
Before launching a new thread “Benchmarking” a few points need to be agreed upon. It isn’t necessary to get everything right before the starting whistle blows, but for sure a good portion of them.
1. Is it legitimate to use the Red Car project in 4K modified or any form.
2. Agree on a single output render template, to 2K or to 4K or both?
3. Agree on rendering to say h264 or Hevc or both?
4. Agree on the format of data to use in say a spread sheet.
5. Agree on the format that users will post their results in, should be fixed mostly.
6. Agree that the Red Car 4K modified version is aok, i.e. that it won’t need any further changes.
7. Agree that the data, say a spread sheet would be regularly made available to a downloadable location.
Add more suggestions as needed ...
Item 1above .. If theres an issue with it then an alternative might be converting to 4K the Magix 2k benchmark thats supplied with VP, Magix is unlikely to object.
Item 6above .. A few users need to download the suggested Red Car 4K project and confirm that its 2k to 4k media conversions are correctly done, that no further changes will be needed. For example are the fps settings correct, should the data rate be higher or lower, is the codec used ok for the new 4K media.
Item 4suggestions, order of items not finalised ...
The Machine entry item could be say [OB] or a specific manufacturer model example [HP dv7]. OB is own build.
VPnn example .. VP16.
The Cpu and Gpu items would have say multiple entries each, Gpu .. manufacturer and model, example .. [Nvidia] [Gtx 1070]. Cpu .. manufacturer, model, cores, example .. [Intel] [i7 4790k] [4].
The OC overclock item would be say a .. [yes], [no] or [partially] entry. For example if the cpu wasn’t overclocked but the gpu was then maybe partially covers it? Or the cpu had a mild OC.
Encoding could be two fields .. H264 or Hevc and Cpu Nvenc Vce Qsv Cuda etc. Example [h264] [CPU].
The FPS entry could be say [full] or an actual value. Use the comments field to clarify if necessary. It would be necessary to make sure that proxy files were not in use and that preview was at full/best.
The comments entry could be used to clarify that its a dual cpu, gpu etc.
1) I downloaded the zip. It hangs my laptop and main PC.
I suspect that the problem is that this file "Mercedes Clip 2 AMD VCE 2160p30.mp4" has been converted to variable frame rate, not constant, as in the original Red Car test. VP doesn't play well with VFR so why do this?
2) The 4K data rates are not typical of what you would have coming off an average consumer 4K capable camera, let alone pro 4k cameras. 4K UHD usually weighs in at 100 mbps and upwards. This low data rate will still do fine in say comparing the relative performance of different machines against each other, but what it won't do is give an idea of what it would be like to ingest 100 mbps UHD media into a 4K UHD project and think that well, I have this camera so this machine in the benchmark list plays/renders it aok.
I'm not saying that some 4K cameras media isn't captured at 21.494 kbps or 22,300kbps, (2 of the 3 4k files) but its on the untypically low side.
Of course it may be a case of comprising on the download size, but then its not fulfilling one of the two tasks (B), as I see, it for a benchmark. i.e. … A) comparing machine to machine B) Confirming that a particular specced machine can handle a 4K project.
I redid 2 files to a higher data rate, (the 3rd file was already at a high data rate), one of which had VFR, now constant FR, one to 67,459 kbps and the other to 89,815 kbps. Zip file size is 7.98.
To me that seems to be better than stealing a question of someone else.......😉
Former user
wrote on 7/7/2019, 9:03 AM
I rendered out Red Car 4K project (with mods to two source media as mentioned above) to 4K UHD at a data rate of 24/20. I could only use CPU rendering, attempting either Nvenc or QSV caused a freeze in rendering at 16%. Rendering to FHD using Nvenc was fine, no problem. I think people will mostly want to render out to FHD anyway.
A blank text file, see image to above right, can be used to serve as user data input.
The image to the left is the 4K render template used.
Outputting to FHD 24/20 data rate using Nvenc was 0:31s.
User name :JN_ Machine :OB VP :VP16 CPU :Intel i9 9900K 8 cores GPU :MSI RTX 2080 Ti OC :No Encoding :h264 Nvenc Render time :0:31s FPS :27.566 Comments :fps 95% of the time