The Adobe MAX conference took place earlier this week. It consisted of several keynotes, live streams, workshops, breakout sessions, announcements, and it aligned with an update to several Creative Cloud applications.
One such announcement is that NVIDIA RTX is coming to Adobe Dimension CC.
While the application has not exactly taken off yet, it is interesting to see Adobe and/or NVIDIA put the engineering into bringing their ray-tracing units to it. First, for its audience, the speed boost (and thus increased preview size) should make the experience much better. Second, if NVIDIA helped with the engineering effort, which I suspect they did, then it suggests that they are hoping to bring RTX basically everywhere. I’m curious to see who else gets RTX support. Fingers crossed for an announcement at BlenderCon next week. I shouldn’t hold my breath, but I am.
So, for gamers, RTX content is still pretty-much MIA, as is constantly reported. The same is mostly true for professionals… but that might change soon. We’ll need to see.
this will be interesting to
this will be interesting to see how it develops.
And eager to see how 3d packages like Blender/Maya/Max implement RTX.
Well most 3D graphics
Well most 3D graphics rendering as well as Ray Tracing(Done via the standard Shader cores Ray Tracing method) is not FPS dependent like on gaming workloads where there is a limited amout of gaming Frame Time available(33.33ms for 30FPS and 16.67ms/less for 60FPS/higher) frame rates. So it’s going to be Ray Tracing accelerated via Turing’s RT cores compared to Ray Tracing done via Radeon ProRender(OpenCL based Blender 3D plugin) on Shader cores for Vega/eariler or Blender 3D/CUDA on Shader cores for Pascal/eariler. [Note: Blender 3D’s cycles rendering works also on AMD GPUs via OpenCL and Nvidia GPUs via CUDA)
So for 3d graphics related workloads like animation and 3D rendering and even 2D rendering those workloads are not constrained by any FPS related time frames and most of the time for 3D animation rendering a single frame can take 15-30 minutes or longer to render with the Ray Tracing settings/sample rates in the Graphics software(Blender, Adobe, others) set a the higher/highest Ray Sample Rates.
Most certianly Nvidia’s Turing RT cores will be able to do the BVH and Ray Tracing Calculations but at what cost compared to just doing them on Pascal or Vega variants. Vega still has plenty of shader cores with wihch to accelerate Ray Tracing on the GPU via OpenCL accelerated Ray Tracing on Vega 64’s 4096 shader cores using Radeon ProRender’s Blender 3D plugin while Pascal SKUs offer less amounts of shader cores.
There will need to be some Benchmarks that have to be created that can make us of Turing’s RT cores for Ray Tracing Acceleration with the Ray Tracing Sample rates turned up higher in Blender 3D and Other software and compared to Shader core based Ray Tracing done on both Pascal’s(Vai CUDA) and Vega’s(Via OpenCL) shader cores.
Nvidia’s RTX IP is costly even compared just the Regular Ray Tracing done on Pascal and Vega shader cores and even that AI based technology that Adobe supports can be accelerated vai AI Tensor Flow algorithms implemented in CUDA or OpenCL on Pascal and Vega GPUs and even on older generations of Nvidia and AMD GPUs. Adobe already supports AI ralated filters/filtering on Nvidia’s and AMD’s older generation GPUs
For the Price that Nvidia is charging currently for Turing/RTX most 3D and 2D graphics software can already implement Ray Tracing and any AI based functionality in software. And Ray Tracing can still be done the orginial way on CPU cores also and that’s still used although its not as fast as using a GPU.
So 2 GTX 1080TIs can be had for around the Price of one RTX 2080Ti or even 2 Vega 64s with 8192 shader cores available for Radeon ProRender’s Blender 3D Plugin to accelerate Ray Tracing via OpenCL on Vega 64. There are even some Vega 64 SKUs selling for as low as $475 so that’s 3 Vega 64s and Blender 3D and the Radeon Pro Renderer will work all 3 Vega 64s for Ray Tracing acceleration and other raster workloads.
Also another Feature that Vega offers is Vega’s HBCC/HBC IP for 3D graphics rendering workloads where for 3D animation projects animators make of very high levels of Model Detail(Millions of Polygons Per Model) with very high resolution Textures. And a 3D animations scene’s total polygon counts can be in the billions and that and the high resolution textures will overflow the VRAM and fill up the available DRAM also. So on Vega 64/56, or any Redeon Pro WX variants, Vega’s High Bandwidth Cache Controller can turn the HBM2 into High Bandwidth cache with any additional VRAM paged out to System DRAM or even SSD if needed. So Vega can address a total of up to 512TB of virtual VRAM and only page what is actually needed by the Vega GPU into HBM2/HBC to be worked on.
Also most folks making use of Consumer Gaming GPUs for 3D rendering/render farm usage are looking at total shader compute for Ray Tracing acceleration. So just like the coin miners they will undervolt and maybe even underclock their GPUs to get the best Performnce/Watt and highest average clock rates rather than overclock and overvolt for the highest FPS rates. So those refrence GPU SKUs for Pascal and Vega will work just fine if they can be had at lower pricing if still available but even the Pascal and Vega AIB cards are sometimes selling below MSRP.
Real Time Rendering is not needed for 3D animation and other non gaming graphics workloads so Nvidia’s Turing/RTX SKUs will have to produce enough Rays Traced to compete on the price/performance metric and that inclides against any dual or even Triple/Quad GPU setups where the 3D Software already can make use of Muit-GPU setups with no need to worry about any FPS/Gaming Only constraints.
The noise is real
The noise is real