Similar to i5 or i7 ?

Only 10% faster ? That is very very low !

I get much better times using Intel HD video card, which is a rubbish video card.

QuickSync only works if:

- your CPU has a video card

- it is activated in BIOS

- you have the driver installed

- motherboard allows both dedicated and on board video cards simultaneously OR you have only the onboard video and no dedicated card

- on board video is set as main display in Windows graphics settings.

I just did some testing, using Teamviewer, with this setup: i5 3570 @ 3.40 GHz, 8 Gb RAM DDR3, Sapphire R7 260X, Windows 7 x64, Sony Vegas 13.

Very disappointing. Very very disappointing.

Using my exact same project, rendering times were:

- CPU: 08m:13s

- GPU, Sony AVC, Render settings: Encoding mode: GPU if available: 07m:18s

- GPU, Mainconcept, Render settings: Encoding mode: OpenCL: more than 10 minutes.

I think Mainconcept does not support OpenCL, but the program still allows selecting the options.

But, here's the problem: GPU-Z: GPU Load shows low usage (6-50%) and very very rarely, hardly ever GPU usage, like some short spikes every 15-30 seconds.

Almost the entire time GPU is not used at all !

When I use Intel QuickSync on my computer, GPU Load is constantly at 80-90%.

GPU load with QS is going to be high since it is using the special aspects of the GPU.

With OpenCL if your GPU usage is low, then your CPU is not keeping up, or the math so lite that you are only seeing when it is needed. The fractional usage could be much higher on the GPU, then it waits for more task from Vegas. GPU-z is an average over the interval period.

If your goal is to have your GPU loaded up during render, use a low end GPU that is more matched to the low end CPU you are using.

Users that complain about how low the GPU usage is when rendering straight up video conversion, do not really understand how OpenCL works to improve things. Run your same tests in 32-bit project mode, and apply a blur or Min/Max effect to the footage. Then report the CPU only to GPU assisted times.

Also choose X series (XTChip) AMD cards as these cards have the most OpenCL compute units and the best bandwidths.

GPU accelerated AVC rendering (chosen in the render settings) is not supported on cards later than the AMD HD 6970 or the Nvidia GTX 580. But GPU acceleration of video processing will still improve render times if the card has good support for OpenCL. More details in this thread. Admittedly that article doesn't cover Intel Quick Sync, but probably should.

Thanks for the article, quite impressive.

Some of the other links I have already read, but lots of new informations.

Unfortunately, I'm even more confused :)

Starting from R9 390, I stepped down to R9 290x (a bit cheaper and almost exact same performance), then even more back to Radeon HD 6970 (cheaper even more and better performance - I think).

Now, based on this article - http://www.hyperactivemusic.com/vegaspro/vegaspro.html - I'm thinking of a GTX 580.

It's even cheaper and almost the best performance - if I read the table corectly: Render Mainconcept = 36 seconds.

Again, I'm only interested in fastest rendering, I don't care about quality and editing/timeline preview.

Also, unfortunately, another problem arose: I need a better PSU, right now I have 450/500 W (can't remember exactly).

So, would anyone recommend the GTX 580, with current i5 3470 and new PSU ?

A better power supply, say 750W and gtx 580 will give you a great render performance increase imo.

The gtx 580 can be purchased 2nd hand at reasonable prices. That's how I got mine. Just be aware that there is the possibility that Magix might just get the finger out in 2017 re: the non-performing issue for rendering with modern gpu's and actually fix it, so maybe wait a while? Unfortunately Magix don't announce their future intentions.

Also, some users may steer you towards say an old AMD card (say amd 6970) of similar era to the gtx 580, that should also do, with about similar iimprovements.

Just be careful while reading advice on the forum of merits of 1 card over the other based on only 1 users testing and experience, including my own. Arriving at a definitive opinion based on only 1 set of tests by only 1 user is not reliable, even though some experienced forum users (who should know better) seem to have fallen for this recently, then again this can be because it fits with previously held assumptions, as a recent thread showed.

Any render testing (or Amd vs Nvidia openCL testing) should be supplied with CPU, OPENCL, CUDA and FPS figures by MORE than 1 user imho, and with a test set thats free and available to all users.

Queensoft, don't get hung up on speed times for one card vs another or one GPU vs another. For any video editor aim for reliability then speed, but for SCS Vegas products, my choices have been - a well cooled Intel i7 CPU (don't over clock) on ASUS MOBO's, with an AMD card (CUDA was faulty IMO), quality low latency memory (16 GIG minimum), and spinning HDD's. You'll save time, money and headaches. From my experience, the R9-390 is a winner (thanks Old Smoke). The Graphics Card is a biggy so only get recommended ones. There are so many factors that affect rendering times that comparing them is chasing the wind. Do you really care if one card can render your video in 37 minutes vs. 42? I have an i7 4770K, 32Gig Crucial RAM, ASUS MOBO and an R9-390. Software on an SDD but I store and write video to HDD's. When I render, I walk away or play games or read email. Rarely a problem. ONLY UPGRADE to a new editor if the old one can't produce what you truly need. Just getting a new editor can set you back. When I went from SVP10 to11, because of the CL language, it obsoleted my NVidia QuadroFX card that cost $3,999. You can buy em used for $25 now. I have other stories but who cares. BTW, SVP products love OPEN CL and barely tolerate CUDA.

Excellent reply, thank you very much.

Yes, I do need to squeeze every bit of rendering time. I have renders almost every other day, if I work harder then every day - and that's just from one customer.

And it looks like 2017 will bring at least one other customer with daily projects. If I'm lucky, there willl be 2 new customers. Not to mention all the other video editing/conversion jobs.

Unfortunately, the above results table does not contain R9 390, it stops at R9 290, which is more than twice slower.

But, I have seen various tests which show some 10% improvement for the 390 or 390x over CPU (various models).

However, this shows big improvements:

I have tested my exact same project, from first post, on another computer: i5 3570, R7 260x, 8 Gb DDR3.

- CPU = 08m:13s (up from 07m:20s)

- AMD GPU = 07m:18s (very slow) - problem is GPU Load was almost always under 15%

Back to R9 390 and 290 (I can find more benchmarks of this card).

They both have STREAM PROCESSING UNITS: Up to 2560 (40 Compute Units), using the Graphics Core Next architecture - which is not supported by Vegas (as I understand).

The older HD 6970 has 1536 Stream Processors, but not using GCN, which is better (?).

If GPU-accelerated AVC rendering is all that matters to you, and you don't have many effects, compositing etc. on the timeline, then get a used Nvidia GTX 580 and render with the MainConcept AVC encoder with CUDA acceleration.

If you want to make it even faster, and you have the budget, put in a card that is good for OpenCL as well as the Nvidia card and use it for GPU acceleration of video processing in your video preferences. You'll need a big power supply, motherboard and case for the 2 cards.

For a good OpenCL card, why not look at newer cards like the AMD RX480/470 rather than those you mention? Check out the links in part 1b of (again) this post.

Yes, as I said, fast rendering is the most important aspect.

Using MainConcept or Sony AVC, doesn't matter which one, quality is just fine anyway.

No budget for 2 video cards.

Projects, usually, 90% consist of one long video or static background, then images/clips/lower third graphics on top of that. Very few (let's say 5-10 for a 40 minute project) titles/color corrector/other minor fx.

Now, back to GTX 580 + PSU and keep current computer (i5 3470), figures look good:

I will try very hard to find time to update and test Vegas Pro 14 today.

Have you tried implementing Vegas2Handbrake? The Vegas rendering engine does not make the best use of available resources, so why not send that task to Handbrake? Start with a constant quality setting of around 20 and set the X264 preset slider to Ultrafast. Rendering should be much faster, quality will be as good if not better than MainConcept or Sony AVC and file size will probably still be smaller than what you are getting now.

Did not try that.

I tried something similar, using ffmpeg (?) or some other encoder, but without success.

I will definitely try it.

I did not read the whole tread but this one is also a possibility to spare time:

If you have more than one project to render, you can open Vegas in more than one instances to render.
So Vegas comes lightly to 100% processor workload.

Last time I did this with 6 instances - with no problems.
 

I don't think that's an option, CPU / GPU Load is already nearing 75-90% with only 1 instance.