XAVC Intra files playback better, but still not 4k @ 29.97fps (I am getting about 26-27 fps).

I looked at GPU-Z, and when playing back 4k XAVC Intra, the GPU has a work load of about 12%.

The strange thing is the video clip that is giving me a problem, is a mixture of screen captures and footage from my FDR-AX100. During the sections of the screen captures the GPU has a load of 12% but during the sections of the footage from my camera, the GPU drops to 2-4% and the playback drops considerably.

Still think it is odd that this machine cannot play XAVC footage (with zero effects) in real time.

perhaps try turning off gpu. from what i've read most nvidia after 58xx series just don't do very much for performance..

With the GPU off the performance issues and playback are the same.

For kicks I tried to import and playback a XAVC Long GOP encoded with Vegas with Final Cut Pro X on my Late 2013 MacBook Pro and Final Cut Pro X could not read the video, only the audio. Which leads me to think that there is an issue with the XAVC Long GOP codec.

I also tried to import an XAVC Intra (also encoded with Vegas) into FCP X, and that file could be read fine. And playback was a bit choppy (as I would expect with a Late 2013 MacBook Pro with integrated graphics).

Playing around with this some more this morning....

Tried to playback the file in Catalyst Edit, the playback starts out choppy but picks up. However if I tried to render (using XAVC Long from Catalyst) the program crashes during the render. What is interesting about this is that when trying to export the same file from Catalyst Prepare, also using XAVC Long), it too crashes.

Catalyst Browse can only export to XAVC Intra (not Long), and it does not crash.

I really think there is something wrong with the XAVC Long files that Sony is creating.

[I]Computer is not really a slouch by any means. 4770k (i7 quad 3.5ghz), 32GB RAM, 3x SSD drives, and a GTX 970.[/I]

Well, 4K does require a lot more then that for smooth editing. Even Sony recommends a six core for 4K editing and as for GPU, 290X is a much better choice. I do quite a bit of 4K editing on my 3930K (4.3GHz) with 2x R9 290 and 32GB and even my machine doesn't fair so well. The XAVC-Intra files from Catalyst work better then the XAVC-S files straight from the camera (AX100). Also the upgrade to Windows 10 has improved timeline performance, I can playback 4K XAVC-Intra in a 32bit float full range project at Best/Full but once I drop some FX or apply transitions the fps drops to around 22.

Been doing some more testing around....

The problem is not the card, it is how Vegas can utilize it.

I take the clip, drop it on the time (set the project settings to match) and playback while watching GPU-Z. During the scenes where the frame rate would nose dive, the GPU would drop to 1-2%. During the scenes where it was just screen captures, the GPU would run around 10% and I would get about 25-26 fps (clip is 29.97 fps).

I then bring the same clip into the new Catalyst apps. Browse and Prepare play the file perfectly smooth, not dropped frames. The GPU would stay around 15-17% throughout. Might drop 1-2% during the tough scenes, but would not affect playback.

Catalyst Edit would play the file perfectly as well, and the GPU was up to 24-26%

This tells me that Vegas just has issues with how it deals with using GPUs - thus a hit and a miss for which card is being used. Where it looks that they got it right with the Catalyst apps.

Maybe it is time for Vegas to get a complete rebuild.

Yes, Catalyst is a newer build but make sure that you set the playback quality correctly, try "play all frames" and see how it fairs. Even that is not a full comparison because there is no audio available with that setting.

Vegas uses only OpenCL for timeline playback which is just much better implemented in AMD cards then Nvidia cards; Nvidia cards are focused on their proprietary CUDA. This has been discussed countless times in this forum.

To me, your results indicate that LGOP content takes more processing time to assemble the images and get them to screen on time, Your system is not capable or optimized to perform this.

The fact Catalyst performs better with higher GPU assist, likely means that Sony is improving on their CUDA support. Catalyst performance is comparing apple and oranges, catalyst has a whole new video engine different from Vegas. You would need to optimize your OpenCL performance, which like OldSmoke, myself and others has posted countless times. AMD holds the crown for OpenCL performance as of this post. Check out Anandtech.com reviews of your card, and look at the "Compute" review section. The stats for Vegas and Luxmark are your best indicators for enhancing your performance.

CPU core count is the next big factor. If you really want to edit in 4K, getting a modern Xeon based system with 8 cores, and not HT cores, would your best bet. Combine that with a 290x or fury X.

Smooth video playback is about having a system that can return the results of complex math faster than real-time or is needed to playback at the speed you are asking for. If you are getting less than that, you need better gear that is tuned properly for the application you want results from.

Sorry guys, going to have to disagree with those assessments.

I don't think poor OpenCL on Nvidia GPUs is 100% as accurate as people claim. If so then standard OpenCL benchmarks like CompuBench would result in poor benchmarks all around for Nvidia cards. The scores I get from the GTX 970 match or beat a lot of the scores from the AMD 290.

What is probably more accurate is that OpenCL in Sony Vegas is not as good as it is in the Catalyst products. In fact my MacBook Pro which has an Intel Iris GPU gets better playback performance with most of my XAVC clips in FInal Cut Pro X (also using OpenCL) than Vegas can (but Catalyst has the smoothest of the three apps so far, I need to test some other apps).

Vegas has been around for a while, I would imagine it would be hard to implement new tech. into perfectly as opposed to adding it on the ground floor of a new application (i.e Catalyst).

[I]The scores I get from the GTX 970 match or beat a lot of the scores from the AMD 290.[/I]

I don't see that.

approx. 150fps for the R9 290.

close to 75fps for the GTX970.

I am getting 120 fps on CompuBench.

Just tried Resolve (free) - played the same clip back with no frames dropped.

[I]I am getting 120 fps on CompuBench.[/I]

Which test did you run?

Correction, it was 111 fps with the video composition test....

https://drive.google.com/open?id=0B5QmcW_8DZ4MRUgwY0JqUS1Bak0

The thing is I want to stay with Vegas, it has been a workhorse for me for many years. But Resolve and the new Catalyst stuff appear to have better broader hardware support (OpenCL support).

So does one choose a card that will work with Vegas, or is it time to think that Vegas has reached it's level in terms of support for new hardware and pick an NLE that is not so specific in terms of which GPU is used.

[I]So does one choose a card that will work with Vegas, or is it time to think that Vegas has reached it's level in terms of support for new hardware and pick an NLE that is not so specific in terms of which GPU is used.[/I]

That is more a personal or work related choice. For me, I choose the card that supports Vegas, didn't go all the way to 290X. But, the R9 290 performs great with Catalyst too. I doubt Nvidia will ever support OpenCL as well as AMD, they have put so much development into CUDA. But, who knows, maybe VP14 will fair better with the latest hardware; if there will ever be such a version.

I got 135fps under CompuBench. Have ever tried the SCS Benchmark project? That is a very good indication for how well your system is balanced.

Noting your increased performance with Windows 10 and 4k, I'd just add that I received a very strong boost with my 280x (Tahiti) moving to Win10. There seems to be evidence that Radeon is handling DirectX 12 very well.

--tom

"Correction, it was 111 fps with the video composition test...."

So in my experience Luxmark represents a closer match to Vegas OpenCL. What scores do you see compared to the lists on :

http://www.anandtech.com/show/9059/the-nvidia-geforce-gtx-titan-x-review/15

Or

What score and GPU utilization do you get rendering the Vegas test file.

Who knows what cinebench video composition is doing, and even if it remotely close to the process that Vegas handles video. The Vegas test file is really the only true benchmark out there. I think the Vegas test file is dated by todays standards and should be upgraded to 4K (4096x2160-60p 10-bit XAVC-I) this would allow people to really see the difference between hardware again. The way it stands now, the test file has peaked.

With LuxMark Hotel I get a score of 1967.

Anybody know where to get the project that anandtech uses to test Vegas render times?

Use the SCS Benchmark project; much better suited for tests and that is what we all did in the past to establish a benchmark for GPUs.

Can you point a brother in the right direction? ;-)

SCS removed the website for that a while ago; sorry, I assumed you had it. I found it on my Google Drive.

Anybody try out the R9 390 yet? I am thinking of exchanging the GTX 970 and wondering is there is any feedback on this card.