The industry standard 3DMark benchmark has just been updated with a new test aimed specifically at pitting AMD versus Nvidia on the ray tracing battleground.
Now that AMD is only moments away from getting into the ray-traced game with its RX 6000-series RDNA 2-based graphics cards, benchmark provocateur, UL, has released this brand new feature test that's able to effectively test the pure ray tracing guts of both AMD and Nvidia's GPUs head-to-head.
Black Friday 2020 deals: the place to go for the all the best Black Friday bargains.
I can already tell this benchmark is going to make for a great head-to-head once AMD's RX 6800 XT is available from November 18. The Radeon GPU is said to be a true RTX 3080 competitor in rasterised rendering workloads (those mostly used in modern games), but we don't yet know where AMD's RDNA 2 architecture will fall on the performance spectrum for cutting-edge ray-traced effects.
Nvidia is touting a significant improvement to its ray tracing performance with its Ampere architecture, and the RTX 3080 offers pretty healthy improvements over even the RTX 2080 Ti with ray tracing enabled.
Both companies' approaches to how they accelerate ray tracing will vary. We already know Nvidia uses an RT Core and Tensor Core combo to both compute Bounding Volume Hierarchy and denoise a ray-traced scene, but AMD's RDNA 2 approach is still shrouded somewhat in a thick pre-release smog.
In order to find out the lay of the ray-traced land, once both architectures are available to test or buy (well, in theory buy), 3DMark has announced a brand new benchmark built around Microsoft's DirectX Raytracing API both, which GPUs are built to utilise on PC.
The aptly-named 3DMark DirectX Raytracing feature test is "designed to make ray tracing performance the limiting factor", and is drawn in a single ray-traced pass. So we'll see exactly how these cards align with little regard for their traditional performance.
UL already offers one ray-traced benchmark. That's Port Royal. The difference here is that while Port Royal was built to use a game-like mix of traditional DX12 rendering and ray tracing, the new DirectX Raytracing benchmark uses ray tracing exclusively to render the scene.
"In this feature test," a 3DMark blog post says, "camera rays are traced across the field of view with small random offsets to simulate a depth of field effect. The frame rate is determined by the time taken to trace and shade a set number of samples for each pixel, combine the results with previous samples and present the output on the screen."
The benchmark itself is a 1440p run, and you can change the sample count to see how that effects performance.
There's also an interactive mode chucked in for good measure, which lets you play about in the sci-fi 3DMark world created specifically for UL's benchmark suite.
You can access both ray traced benchmarks in 3DMark right now, although we'll have to wait until November 18 to actually get anything interesting to compare against.
For reference when that time comes, a quick and dirty run of an RTX 2080 (with a Threadripper 2970WX) in the benchmark, with a sample count of 12, gave us an average of 20.48fps, while an RTX 3080 (with an Intel 10700K) blazes past at 46.27fps.
Anyone fancy a bet on where the RX 6800 XT will land?