Slow light, testing ray tracing performance with Port Royal

Subject: Graphics Cards | January 14, 2019 - 02:23 PM |
Tagged: 3dmark, port royal, ray tracing

3D Mark recently released an inexpensive update to their benchmarking suite to let you test your ray tracing performance; you can grab Port Royal for a few dollars from Steam.  As there has been limited time to use the benchmark as well as a small sample of GPUs which can properly run it, it has not yet made it into most benchmarking suites.  Bjorn3D took the time to install it on a decent system and tested the performance of the Titan and the five RTX cards available on the market. 

As you can see, it is quite the punishing test, not even NVIDIA's flagship card can maintain 60fps.

View Full Size

"3DMark is finally updated with its newest benchmark designed specifically to test real time ray tracing performance. The benchmark we are looking at today is Port Royal, it is the first really good repeatable benchmark I have seen available that tests new real time ray tracing features."

Here are some more Graphics Card articles from around the web:

Graphics Cards

 

Source: Bjorn3D

Video News


January 14, 2019 | 04:14 PM - Posted by RTXonAndRTXoffFromSomeSoftwareCodePath (not verified)

Is there any news on Microsoft's DXR and ray tracing support on GPUs that do not have hardware support for ray tracing(Software Code Path).

If I rember correctly on MS's DX## versions that MS/GPU maker would have an alternative code path for DX## for any GPUs that did not yet have full hardware feature support for the latest DX/Whatever version's/feature sets. This was limited to any Hardware features that DX## supported that could actually be simulated/emulated with a software code path so not all of the Latest DX/Whatever standard features could be approximated in software.

Nvidia's Pascal and AMD Vega GPUs have shader cores for accelerating Ray Tracing/BFH calculation on the shader cores and really Ray Tracing on the CPU's cores is still there also if one looks at the Blender 3D benchmarks that are popular for stress testing the CPU Cores/Threads using CPU calculated Ray Tracing/BVH(1) calculations.

So how many rays/second can be done on Pascal's shader cores or Vega 10's/20's complement of shader cores? It's not likely to be competative with Nvidia's RT cores in hadrware Ray Tracing output levels but still Pascal/Vega and some light weight gaming titles that do not tax all the shader cores on the higher end Pascal/Vega SKUs could use some limited Rays generated by Shader Core/Rays calculated for lighting/shading effects also.

The hard part about "Real Time" tay tracing for gaming is the limited amount of millisaconds per frame time that is available for any larger numnbers of rays that can be generated in 33.33ms down to 16.67ms/Lower frame times! And RTX Turing makes us of Hybrid Ray Tracing where the RT cores limited Ray Traing output/Per-Frame-Time has to be AI denoised and mixed with the regular raster output.

Ray Tracing and its assoicated BFH(Bounding volume hierarchy) methodology are very compute intenstive operations were mostly done on CPU cores for decades until GPUs showed up with all those Shader cores that could be used for any calculation intensive workloads Via CUDA or OpenCL.

(1)

"Bounding volume hierarchy"

https://en.wikipedia.org/wiki/Bounding_volume_hierarchy

January 14, 2019 | 04:19 PM - Posted by Nobody (not verified)

All DX12 cards need is support for DXR in drivers.

Titan V has them available. It has no raytracing hardware.

It is more than likely that such drivers actually exist for other DX12-compatible cards as well, but there are PR and business reasons to not release them for both AMD and NVIDIA.

January 14, 2019 | 05:12 PM - Posted by imostdefinitelyAMarobot (not verified)

youre not implying that these companies are artificially segmenting their respective product lines for financial and/or other business related reasons.....are you? :)

January 18, 2019 | 06:29 PM - Posted by Scott Michaud

No... they physically cannot make a die that has the features of both Titan V and Titan RTX. 800sq.mm is as big as we can go on a single exposure. NVIDIA seems to be separating their chip into two lines, one compute-focused and one ray-focused.

Of course, they could release a 800 sq.mm chip that has neither FP64 nor RTX features, which could destroy the both of them in conventional 3D games... but then we'd never get raytracing hardware as an industry at all.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.