Introduction

On March 17th, 2020 AMD launched its Radeon RX 6700 XT video card with a suggested price tag of $479.  We have evaluated the video card in our full review looking at 1440p performance in eight games plus comparing AMD Smart Access Memory enabled versus disabled.  We compared it against the GeForce RTX 3060 Ti and GeForce RTX 3070.  In that review we found the video card (at its default) dances around the GeForce RTX 3060 Ti in performance, rather than the GeForce RTX 3070, even though it is priced closer to the RTX 3070. 

Therefore, we followed-up with that review with a full overclocking review overclocking the AMD Radeon RX 6700 XT video card.  By overclocking the video card, it now closer equals GeForce RTX 3070 performance most of the time.  The rub is, it took that overclock to be competitive with the RTX 3070 in gaming performance.

There’s another aspect of video cards though that is extremely important nowadays.  We are talking about productivity, content creation, workstation, compute/GPGPU class workloads.  Video cards are used for far more purposes than just gaming, even a gamer may use the video card for render tasks like Blender.  These workloads are more common than ever, and an important factor when deciding what video card is best for you.

In our testing, we found some disturbing data that indicates the Radeon RX 6700 XT is a lot slower than the competition at some of these workloads.  Therefore, we wanted to test it, provide some benchmarks for you, and see what the deal is.  This review today focuses on running some popular workstation/productivity benchmarks and comparing them against the competition.

Our test setup is simple, we are using the exact same system, video cards, and drivers from our launch review and our overclocking review. We do have AMD Smart Access memory enabled on the Radeon RX 6700 XT. We also tested with it disabled, and all the performance results were the same, so it made no difference, but we did leave it enabled.

Blender Open Data Benchmark

The Blender Open Data Benchmark tests real-world Blender renders on GPUs.  The latest version it supports is Blender 2.92 and so that is the version we are testing.  There are several scenes to render, we chose to use the “pavillon_barcelona” and “victor” scenes.  In addition, on GeForce RTX GPUs you have the option to run the renders using CUDA or OPTIX, we tested both.  With AMD GPUs the only option is OpenCL.

pavillon_barcelona

In this first graph, we are comparing the pavillon_barcelona scene and NVIDIA GPUs are running CUDA and AMD GPUs OpenCL.  The first thing to notice is that the Radeon RX 6700 XT does provide a big improvement coming from the previous generation Radeon RX 5700 XT.  However, it is being beaten by the GeForce RTX 3060 Ti by about 9%.  It is being beaten by the GeForce RTX 3070 by an even greater degree, 27%.  If the Radeon RX 6700 XT is priced closer to the RTX 3070, it’s doing a lousy job in Blender, by comparison, being even slower than the RTX 3060 Ti.  And this is without using OPTIX render, which changes the game completely.

In this graph, we are using the same pavillon_barcelona scene, but now we have the special RTX OPTIX render enabled.  This just really changes the game.  There is just no comparison, the RTX 3060 Ti and RTX 3070 are in an entirely different class here, and able to render this scene extremely fast compared to the Radeon RX 6700 XT.  The RTX 3070 is 66% faster than the RX 6700 XT.  That’s huge.

victor

Now in this graph, we are looking at the victor scene.  Once again NVIDIA is using CUDA and AMD OpenCL. The Radeon RX 6700 XT is way faster than the previous generation Radeon RX 5700 XT, so that’s a great thing to see.  It’s 60% faster, which is a huge gain.  But, the GeForce RTX 3060 Ti and GeForce RTX 3070 are faster still, but this time not by as big of a difference. 

Now we have enabled the special RTX OPTIX render mode.  There is just no comparison with OPTIX.  The RTX 3060 Ti is 63% faster than the RX 6700 XT and the RTX 3070 is 66% faster than the RX 6700 XT.

Brent Justice

Brent Justice has been reviewing computer components for 20+ years, educated in the art and method of the computer hardware review he brings experience, knowledge, and hands-on testing with a gamer oriented...

Join the Conversation

9 Comments

  1. I had thought that AMD’s solutions for Professionals and Creators were always solid an competitive. Boy was I wrong! Thanks for the follow-through on another great review Brent!

    nVidia’s OPTIX is really a game-changer it seems with respects to performance. Those Blender times are nuts!

  2. The Vulkan test was interesting – virtually unchanged from the 5700. Makes me wonder if it’s immature drivers causing some of this, as every other test shows at least some architectural improvements. That would be a heck of a driver bug – but hey, it’s AMD, drivers… yeah.

    Or maybe they just heavily optimized gaming, after years of optimizing compute only to lose on gaming benchmarks… possible but doesn’t fit the pattern.

  3. I guess for most people this shouldn’t be an issue. But I’d still get the RTX3070 if only becuase I’ve been waiting for so long… :ROFLMAO: :ROFLMAO:

    nVidia’s OPTIX is really a game-changer it seems with respects to performance. Those Blender times are nuts!

    RTX cores at its best.

  4. Glad you guys decided to look at workstation productivity shiznit, as I know you were thinking about doing such. Very interesting stuff. Great article! Hope to see more like it in the future. Yyeeaahh sadly the way things are, CUDA and OptiX and NVENC are unstoppable. nVidia is kinda the only real choice for these kinds of workloads. Radeons don’t even win with the OpenCL stuff.
  5. The productivity stuff doesn’t mean much to me, but the streaming capabilities and nvidia broadcast tip the scale IMO.

    BTW some photoshop/vegas filter benchmarks would be welcomed.

  6. The productivity stuff doesn’t mean much to me, but the streaming capabilities and nvidia broadcast tip the scale IMO.

    Same; I need to test RTX voice out on something that has RTX though. That’s a big one for me given that it’s pretty hard and pretty inconvenient to get to a place where the ambient noise is below the noise floor of the recording equipment!

  7. As someone who has been using gpu computing in a professional field for a long time, there is a reason we only ever consider buying nvidia gpus.
    And it’s not like the CPU market where AMD can just pull an ace like with the CPU market against intel. In their absence, they allowed nvidia to become so entrenched in the professional gpu market that they stand no chance. Many software uses proprietary nvidia technology that makes AMD a non player even if it had the performance.
  8. As someone who has been using gpu computing in a professional field for a long time, there is a reason we only ever consider buying nvidia gpus.
    And it’s not like the CPU market where AMD can just pull an ace like with the CPU market against intel. In their absence, they allowed nvidia to become so entrenched in the professional gpu market that they stand no chance. Many software uses proprietary nvidia technology that makes AMD a non player even if it had the performance.

    I’ve said this dozens of times, not only do you get great performance hardware with nvidia, but better support and middleware that makes it a much better choice. Cry all you want for OpenCL, CUDA is the de facto standard in many content creation applications. OPTIX is also orders of magnitude faster than anything AMD has to offer and don’t even get me started with AI and inferencing.

    AMD has to invest in middleware if they want to compete on the high end, for years AMD was a compute monster but all that power got to waste in most professional applications.

Leave a comment