So now that Vega Frontier Edition has been out for a few weeks now, what’s the consensus on this card and its performance in Vray RT and 3ds Max viewport? I’m hearing stuff about it being a very nice card for compute. Did anyone test it yet (Chaosgroup?)? I don’t see any benchmark results with vray’s benchmark.
I don’t think it is really out yet. It has had a super limited launch with just a couple of cards in the actual wild, so the chances of someone with vray having one are probably slim.
No, its super available (at the frontier edition price). Newegg is showing them right now
As for RT, I discovered something last night using my ryzen and fury x - its fast, like very fast. Like faster than my 6700k and 980ti using CUDA (I believe). There is something going on because under devices, it doesn’t list opencl or c++ as my 6700k does. Instead, they both list as amd parallel compute something something.
During hybrid rendering, they both use the cpu and gpu at 100% (which was not my experience with 980ti/6700k hybrid). I’ll have to benchmark to do an apples to apples comparison, but I would not expect those results
We still haven’t had the chance to benchmark Vega. But from guys that have used it with V-Ray GPU I know that it should work fine with the latest release on Windows.
Best,
Blago.
Sounds exciting. Did you test it with the Vray benchmark? I don’t see a single benchmark run on a Radeon graphics card in the results page.
Interesting. Any word about the performance compared to a 1080ti ?
Not yet.
Best,
Blago.
I expect performance to be decent… the last two gens of AMD 290x and fury have worked well with vray RT, being about equal to nvidia, but no support for CUDA, so less features.
There are less features if using OpenCL? Is this true and why? I quickly checked the RT help pages and couldn’t find such statement.
It is. First things I can think of are adaptive lights, sss, clipper and aerial perspective.
We have to work on the docs for that. Not sure what would be the best way to write that … docs are already kinda messy.
Best,
Blago.
Oh ok. Then I guess these cards are still out of the question. 
I would have loved to try an AMD card like VEGA for a change. Are there plans to add the same level of features in OpenCL in the near future? If the raw compute performance of VEGA is really as nice as people mention it to be, it would only make sense to port as much RT features to OpenCL as possible for users take advantage of the speed of these cards.
If the performance of Vega is that nice, sure, we can most of those features to the OpenCL version as well (the only thing that will most likely not make is FastSSS).
Best,
Blago.
That’s a pity about the translucency options… but appreciate that the differences with openCL don’t make it very efficient. Maybe it would be possible to use MDL or OSL if the currently supported closures support it? Would that make a difference from writing a native sss shader for CL and is the cpu part with hybrid based on CL or Cuda?
Just saw the cpu hybrid is based on Cuda…
I’m expecting a good option (at least for those with multiple gfx cards) is to use Vega for the display port - but not taking part in rendering - and maybe using Vega to perform the denoise, as the previous AMD cards out performed Nvidia by a large margin. Not sure if removing a gpu from rendering also removes it from denoising?
Food for thought… Ben
V-Ray automatically picks the most suitable device for denoising based on heuristic - it picks the fastest devices, but also tries to avoid devices with monitors attached (unless they seem much faster than the other ones in the system).
If it did not pick the device that you have liked to, there is a way to override its choice, using environment var VRAY_OCL_DENOISER_DEVICES. For example VRAY_OCL_DENOISER_DEVICES=amd will force the denioser to use only AMD devices.
Best,
Blago.
I recalled the FURY X being a Monster for denoising, it was 5 - 10x faster than the titan…
Well the RX vega is out, and it is power hungry and slow in games, but i saw a very promising blender bench mark… I wonder if there is any vray rt benchmarks out for it?
I found this guys:
Which is intersting because I thought vray bench mark didnt support Open CL?
We do support OpenCL for AMD cards in the latest release of the benchmark.
Best regards,
Vlado
So turns out Vega is 10% faster than 1080 ti in Vray Benchmark? That’s quite a surprise. Can you run a Vega alongside a GeForce GPU when rendering with RT and would it be wise to mix up AMD and Nvidia GPUs? I suppose that if they both have to use the same computing language when rendering you’ll be wasting speed since Vega will run slow on CUDA and GeForce will run slow on OpenCL.
I would take these results with a grain of salt as it’s not clear if the CPU OpenCL device was contributing to the result or not (the results were not posted to our benchmark site so we don’t know). We will patch the benchmark to only run the OpenCL code on GPUs. Then again, if CPU OpenCL was enabled, it should have contributed to the LuxMark and ProRender results as well… it’s a bit unclear.
Best regards,
Vlado
Would the benchmark run in hybrid mode?
Alex it is probably not wise to mix gpu types, even if possible… you’d be forced to run open cl on the nvidias in that case. But blago is best to answer you 