Importance of raster vs. Ray Tracing performance for the Enscape engine

Hi,

I am building production machines for a client. It is the first time I encounter Enscape as a use-case / requirement.

I am more used to mechanical design CAD use-cases (aerospace, consumer products, etc.), where ray tracing is rarely a factor.

I understand that the Escape engine uses RT on some level ; but to what extent ? The most recent info I could find on this forum is from Demian in 2021 , here :

(source : RE: Enscape Engine question , emphasis is mine)

So clearly, 4 years ago the Enscape engine was already planned to evolve with a much bigger focus on RT.

Where does it stand, as of now, early 2025 ?

For typical day-to-day use-cases in Enscape, should I favour pure rasterization performance or RT performance , for maximum client satisfaction ?

For example in the following scenario :

  • let’s pretend 2 comparable-tier GPUs on the market, GPU A and GPU B

  • GPU A wins by 20% in raster

  • GPU B wins by 20% in RT

This is for the Sketchup plugin version (if that changes anything).

Thanks in advance

Hi @mach128x

in fact a lot has happened in terms of utilizing hardware ray-tracing. We’ve not only added support for ray-traced sun shadows, but also more recently for ray-traced artificial lights as well.
So depending on what your focus is, I’d indeed recommend to favour a GPU with fast support for hardware ray-tracing (quite frankly this will bring you to Nvidia’s line-up). Usually these GPUs also have quite capable rasterization performance.

If visual quality is not your main concern and you’d rather want to focus on best performance for smooth real-time walkthrough then of course you might be able to get a better “bang for the buck” with other GPUs with less capable ray-tracing capabilities.

You are probably splitting hairs, whatever GPU you get will likely be determined by the client’s desired price point and availability. You can assume that nvidia is preferred over AMD or Intel offerings.

The main thing that will be a factor is VRAM. Many users like creating huge models with tons of assets and not much thought for efficiency. So it becomes a matter of how much you can load in VRAM before Enscape gives up.

“Consumer” GeForce cards give better performance per $ than the “Pro” cards. But the “Pro” series (was Quadro, now just RTX) are supposedly better supported by “Pro” design apps. The consumer side of things is kind of up in the air at the moment, the GeForce 50x0 series was just released but is unobtainable. In the older 40x0 series, you’ll just need to weigh cost vs VRAM. The new 5090 for $2000 with 32GB VRAM would be ideal, but apparently only exists on paper.

Basically if they are OK with a $1500 price tag, get the “Pro” RTX 4000 Ada generation single slot with 20GB VRAM (not the SFF version, though that works well too, just a little slower.) You can of course bump up in the Pro line for much more $$$. Also note the naming conventions between Pro and consumer are seemingly designed to be as obtuse and confusing as possible, so be sure to research thoroughly.