(FAQ) Graphics Cards & GPU-Acceleration for VEGAS Pro

Nick Hope wrote on 11/23/2016, 3:44 AM

IMPORTANT! VEGAS Pro 15 introduced the following changes:

  • A new MAGIX AVC/AAC codec significantly accelerates encoding on machines that utilize Intel® Quick Sync Video (QSV) technology
  • A new MAGIX AVC/AAC codec significantly accelerates encoding on machines equipped with modern NVIDIA graphics card GPU technology
  • More efficient handling of AVC files on the timeline through Intel QSV technology

In summary (as I see it for now):

  • The MainConcept AVC/AAC encoder has been replaced with the MAGIX AVC/AAC encoder, which inherits the same functionality and adds rendering acceleration via QSV or NVIDIA NVENC (supported NVIDIA GPUs).
  • QSV is now an option for GPU acceleration of video processing.
  • For VEGAS Pro 15, the general recommendation is now to keep GPU drivers as up-to-date as possible. This differs from previous advice.

This FAQ post will be updated to reflect these changes when their effects are better understood. In the meantime, it remains out of date and applies to VP14 and earlier. Clearly the changes in VP15 may affect recommendations for choice of GPU and CPU.




GPU acceleration was first introduced in Vegas Pro 11 and has not been significantly updated since, but it can still bring significant benefits with the right graphics card.


VEGAS uses OpenCL for GPU acceleration. It uses it to accelerate playback of the timeline and GPU-enabled FX.

1a. Enabling GPU acceleration of video processing

GPU acceleration of video processing is enabled or disabled here: "Options" menu > "Preferences" > "Video" > "GPU acceleration of video processing". You will need to restart VEGAS after changing the setting.

1b. Graphics cards that work well for GPU acceleration of video processing

Although both AMD and NVIDIA GPUs support OpenCL, AMD's implementation is stronger. The following GPUs/graphics cards have been known to work well with VEGAS Pro 11-14:

Presumably similar variants of those cards will also perform well.

1c. Graphics cards that don't work so well

The following GPUs/graphics cards have been known to give disappointing performance with VEGAS Pro 11-14 (of course your decision may be affected by other applications which work well with these cards):

  • NVIDIA GeForce GTX - Except GTX 580 and earlier, specifically for MainConcept AVC GPU rendering (see below)
  • NVIDIA Quadro - Unless you specifically need NVIDIA 3D Vision live preview, which requires a Quadro card, or (debateably) 10-bit montitoring. AMD Radeon gives more acceleration for less expense. Choice of driver may make a huge difference in Quadro performance (example).
  • AMD FirePro - These workstation cards may work OK but most users choose Radeon.

1d. Notes about GPU acceleration of video processing

  • Artifacts may appear with some formats.
  • Some FX and compositing may behave slightly differently with GPU enabled.
  • Windows may update your graphics card driver and introduce problems such as VEGAS not recognizing the card. If so, update your driver from the manufacturer's website or from the NVIDIA or AMD website. In some cases the driver may need to be rolled back to an earlier version. Display Driver Uninstaller can help with a clean uninstall.
  • Some Geforce cards (e.g. GTX 745, GT 730) cannot be recognized by VEGAS with certain drivers and the driver may need to be rolled back.
  • A spreadsheet of "good" and "bad" NVIDIA drivers for VEGAS can be found via this post. Please post your own reports of success or failure in a comment on that post.
  • If your preview window is blank with an NVIDIA card, there are reports that this can be solved by installing a CUDA Toolkit. There is a particular problem with Nvidia driver 378.66.
  • You don't need GPU acceleration. I successfully edit UHD video in VEGAS without it using my Intel i7-5960X processor, and there are other alternatives to speed things up such as proxy editing.
  • AMD Vega is due for release in 2017 and may perform very well with VEGAS.
  • There are some reports of NVIDIA drivers improving in terms of OpenCL performance (example 1, example 2).
  • Some VEGAS FX that were purchased from Velvetmatter cause VEGAS to crash with some AMD drivers more recent than 16.10.1. They may also work better with Nvidia GPUs anyway. The affected FX are:
    • Defocus
    • Fill Light
    • Glint
    • Rays
    • Soft Contrast
    • Starburst


The two AVC/AAC encoders available in VEGAS allow GPU-accelerated rendering, which is a different thing from GPU acceleration of video processing explained above in part 1.

In Vegas Pro 11 to 13, the GPU rendering options were diplayed by default. Since VEGAS Pro 14 build 201, GPU rendering must be enabled here: "Options" menu > "Preferences" > "General" > "Allow legacy GPU rendering".

The term "legacy" is used because the code was optimized for the GPUs that were available when it was written, a number of years ago. The code is not optimized for currently-available graphics cards.

2b. MainConcept AVC/AAC

GPU rendering with the MainConcept AVC/AAC codec is enabled here: "Render As" > "MainConcept AVC/AAC" > "Customize Template" > "Encode mode"

CUDA GPU rendering in MainConcept AVC/AAC is not supported on GTX 600 series or later cards, however it works well on the GTX 400-500 series, which were based on the Fermi architecture. The GeForce 296.10 driver is proven to work well, but is reported not to work after Windows 7. Users have successfully used 334.89 and other later drivers by tweaking the "OpenCL Memory Size Filter" in the internal preferences. The official information from the Support team is that the GeForce 337.88 or Quadro/Tesla 341.05 (an R340 driver) or older driver is required.

If fast rendering of AVC video is critical for you, then consider using CUDA with an NVIDIA GTX 580 card, which was the last and fastest card supported. They are still available on eBay etc.. If your PC case is large enough, you could fit one alongside an AMD card that is better for OpenCL acceleration of video processing and combine the benefits of both. I have had a GTX 580 and an HD 6970 in the same PC without issue.

OpenCL GPU rendering in MainConcept AVC/AAC requires an AMD Radeon HD 4000/5000/6000 series GPU. The latest, fastest card to support it was the AMD Radeon HD 6970. CUDA rendering with the GTX 580 is reported to be faster. Note that you can still select "Render using OpenCL if available" even with later cards that support OpenCL, but doing so will probably make your render slower.

2c. Sony AVC/MVC

GPU rendering with the Sony AVC/MVC codec is enabled here: "Render As" > "Sony AVC/MVC" > "Customize Template" > "Encode mode" > "Render using GPU if available"

OpenCL GPU rendering typically does not significantly accelerate the Sony AVC/MVC encoder and in some cases it can makes it slower.

2d. Notes about GPU-accelerated rendering

  • The rendering result will be different with GPU-acceleration enabled. It is likely to be slightly inferior for the same settings.
  • If "GPU acceleration of video processing" is enabled, a suitable GPU will still accelerate your render even if "Encode mode" is set to "Render using CPU only", because VEGAS still has to process your timeline.


  • GPU acceleration may not work well if your CPU is relatively under-powered or over-powered. The two should be in balance.
  • In some cases GPU acceleration or rendering may actually be slower or less stable than with CPU only.
  • Driver choice can be critical to GPU compatibility, stability and performance. Different versions of Windows may support or prefer different graphics card drivers. Updating or rolling back a driver can make a huge difference.
  • VEGAS does not benefit much (or at all) from NVIDIA SLI or AMD CrossFire. See this discussion.
  • Powerful graphics cards may be extra-long and take up multiple slots. Be sure your motherboard and case have enough room.
  • The guidelines in this post are based on users' experience and may not represent performance in your own system. If in doubt, test the options yourself.



Nick Hope wrote on 11/23/2016, 3:45 AM

Suggestions for corrections and additions via comments or personal messages are welcome but comments here may be deleted later as this is an "FAQ" post. Please start a new post or use an old thread to discuss subjects in depth or to raise individual cases. I will try to keep the original post up to date.

astar wrote on 1/16/2017, 7:11 PM

1b. Graphics cards that work well for GPU acceleration of video processing:

The best bet would to choose cards not on price. Buy AMD GPUs that come with an XT chip / x series chip as these cards have the most compute units enabled for the GPU chip type.

Single point GFLOPS performance is what you are looking for with GPU acceleration.

If NV is finally coming around to support OpenCL better than advocate 1080 or later cards.

No one should be pointed at older NV cards to support MC encoder, as the latest CPUs i7-5xxx or higher are faster at than some old NV card turning an h.264 encoder that was not that good in the 1st place. GPU encode is a passe mode of operation from 6 years ago.


1D - "AMD FirePro - These workstation cards may work OK but most users choose Radeon" Actually these suffer from the same issue above, and should be selected with the XT chip and not lower end versions.


"You don't need GPU acceleration." - sure you can walk across town vs drive. Every CPU of the last 5 years has registers that create OpenCL compute units virtually on the CPU. That means even in "CPU only" mode the opencl calculations are happening on the CPU only, but it is still operating OpenCL. The reason users should solve their GPU stability issues, is that CPUs are massively slower at single point FP calculations than a CPU+GPU OpenCL hybrid operation.

Why gimp your system and then complain about slow renders.

OldSmoke wrote on 1/16/2017, 7:45 PM

No one should be pointed at older NV cards to support MC encoder, as the latest CPUs i7-5xxx or higher are faster at than some old NV card turning an h.264 encoder that was not that good in the 1st place.

I disagree. I recently put my GTX580 back into the system because I need fast MC AVC render for Internet and there is nothing faster out there. Especially the combination of the Fury-X that takes care of the OpenCL frame preparation and the MC AVC CUDA encoder is simply amazing. Granted, at bitrates below 10Mbps there are better codes. Sony's AVC OpenCL codec was never as fast, even with the last working HD6970.

Paul-Fierlinger wrote on 1/16/2017, 8:22 PM

I have AMDFirePro V7900 because I need four monitors and it's been working for me just fine for months (or is it over a year or two by now?)

SphinxRa40 wrote on 1/26/2017, 6:01 AM

Hi Nick,

Shouldn't be there more info about which why GFX works well or not? As far as i learned its all about the architecture NVIDIA is using the last years why it isn't working like for GPU acceleration (Kepler, Maxwell, Fermi, Pascal), i believe it was: Kepler/Maxwell not and Fermi/Pascal are working for GPU Acceleration...Because some older models uses old architect in Quadro's i saw on wiki, i think that would make more sense and shorter explained wich card would work or not for GPU acceleration for new users for now, instead to specify and get discussions about model numbers.

OldSmoke wrote on 1/26/2017, 8:48 AM

SphinxRa40 have you read all of Nick's post and all the links? This topic has been discussed since VP11 came out and Nick has summed it up perfectly.

SphinxRa40 wrote on 1/26/2017, 9:28 AM

Not yet, but i will do now read the links, maybe i read to fast^^

OldSmoke wrote on 1/26/2017, 1:39 PM

I do however see that Nick left out the Radeon Fury X which I can highly recommend from my own experience.

Nick Hope wrote on 1/26/2017, 9:32 PM

I do however see that Nick left out the Radeon Fury X which I can highly recommend from my own experience.

I have added it. It would be useful to get a brief summary of where that card stands compared to the other AMD cards (especially R9 and RX 4**) in terms of age, architecture, power, price and market positioning. Unless you're keeping a close eye on things, the range can be confusing and I know others like yourself are more on top of this than I am. Also, any comparative test figures?

OldSmoke wrote on 1/26/2017, 10:26 PM

The R9 Fury X is based on the Fiji XT GPU and most importantly incorporates HBM instead of GDDR5. More details can be found here. https://en.m.wikipedia.org/wiki/List_of_AMD_graphics_processing_units

Only the upcoming 500 series will have an updated HBM2 with increases memories, clocks and band width surpassing the Fury X.

Nick Hope wrote on 1/26/2017, 10:37 PM

Thanks OldSmoke. Is the upcoming 500 series the same as "Vega"?

(I added 2 new notes to part 1d and 3 new links to part 4.)

OldSmoke wrote on 1/27/2017, 12:20 AM

It may or may not or may again only be available in the higher end cards like the current Fury X. I haven't seen any reports with regards to the Vegas architecture and their model designation. AMD has started a high end segment with the launch of the Fury series. All these new developments however are very much targeted for the gamers market and not necessary are better for NLEs. I decided on the Fury X because of the rather high band width, number of shaders and compute units.

astar wrote on 1/27/2017, 12:56 AM

This might help. I created a chart from the Wiki pages, and combined info from other places like Techpowerup.


The main stats to look for in a GPU for Vegas would seem to be memory bandwidth, compute units, and Single Point GFLOPs. When you change your project settings to 32-bit FP, you start to lean on the GFLOP performance of the CPU+GPU OpenCL combo.

FijiXT is still the current power in the waters even after 18 months past its release date. GPUs like the RX480 are geared more towards other features and energy saving, but do not have the compute performance of the Fury-x.

Even the NV1080 barely surpasses the Fury-x even after having a year to catch up, or NV laid up with card settings that just surpassed AMDs.

Nick Hope wrote on 1/27/2017, 3:43 AM

Thanks astar. Great chart! I've added a link to it from the first post.

qualle wrote on 3/9/2017, 5:37 AM

Thanks Nick for these detailed instructions and thoughts! This encouraged me to plug an older Radeon 7750 card into my system which runs all PC displays from a GeForce GTX 770 card. And I would like to share that this works like a charm: 3 displays plugged into the GeForce card for all software application windows, but GPU acceleration within Vegas 14 redirected to the Radeon 7750. Works stable and seems to be really fast (no benchmark tests, though, just my impression). Moreover I can redirect the preview to BM Intensity Pro 4K as well using these settings. So I can encourage everyone to test configurations like that based on the FAQ article - I never imagined this might work, but it does.

KenzG wrote on 3/15/2017, 5:34 PM

I hope that's Vegas Pro in the future or next time Version, which is support with Newers GPU..

I had GTX 1060 and its yeah still might be problem while rendering,

Jake-O wrote on 5/18/2017, 10:05 PM

hello guys! This has been very helpful. However, I am mostly working on a laptop and looking for gpu with equivalent power as the ones mentioned above.

I have a hard time looking for laptops with amd gpus and would like to seek for advice if the r9 m275 for mobile would be good for fluid playback (considering no effects) and also how about the radeon rx460 that asus has.



EksDi wrote on 9/3/2017, 4:32 AM

I just wish the Encoder in MAGIX AVC Has AMD support since I have an RX470.

campechano wrote on 9/15/2017, 4:18 AM

Warning to use this codec with nvidia graphic card ( my case).

Very bigger pixel distorsion when plays the final render with defaults parameters. Exemple: if use noise default effect and them any video, the disaster appers . Not recomend

williamvv wrote on 9/15/2017, 10:55 PM

Dear Vegas community,


We all know that the GPU acceleration,

Has never been without complication.


When Vegas 15 promised support for the newer GTX, I thought there was hope,

But after trying various codecs for hours, I shook my head and said, “nope”.


Accelerated with my GTX 960, a plain video rendered just fine,

But when filters or fades were added, the result was Frankenstein.


Flicker flacker, full of glitches, even the unsharp mask went awry,

For so much effort getting a paid upgrade to work, the time cost is too high.


With such a common graphics card, I guess I’m not alone,

A graphics program that fails at graphics – how can anyone condone?


Has anyone rendered a complex project with the GTX 960 (or similar) and did succeed?

If you could kindly share your experiences, I’d be thankful to read!