We are happy to announce the release of a new preview version of OctaneBench with experimental support to leverage ray tracing acceleration hardware using the NVIDIA RTX platform introduced by the new NVIDIA Turing architecture.
This application lets Octane users benchmark the maximum possible speed boost Octane will get from RT Core ray tracing hardware in Turing RTX GPUs. This experimental version of OctaneBench is still a work in progress. It will be tuned and updated in future releases as we integrate RT Core support in Octane 2019.2 and later in the coming months.
RT core hardware provides significant speed gains for heavy scenes like the one in this new benchmark.
The benchmark scene was originally built in 2015 for this Brigade demo:
While not created or optimized for RTX, in our testing to date, it has provided the largest speed boost in OB with 'RTX on' for an actual real world scene, so we have selected it as the basis for this benchmark. Besides being geometrically dense, the scene is composed of hundreds of thousands of scattered instanced meshes. The ORBX in this benchmark can be inspected in Octane standalone for anyone interested in pulling it apart.
This RTX benchmark will provide two OB scores on RTX cards: one for "RTX on" and one or "RTX off" - and an overall speed boost gained from RT Core being enabled. On non-RTX cards, there will be just one OctaneBench result, which should be in line with previous OB results.
In our testing, on a 2080, we are seeing nearly a 3x speed gain for overall OctaneBench scores with "RTX on".
Here are the results on a 2080 Ti and dual 2080 Ti (you can mix and match non-RTX and RTX GPUs in too):
The speed gain is over 5x for info passes - on the very same GPU- with RTX on:
The above results on a 2080 are a remarkable 10x speed improvement over the previous generation 1080:
When considering how RT Core hardware will generally speed up rendering in Octane, it is important to remember that RT core speed gains are currently scene specific, and in many other scenes we have tested, we are getting less than half the speed boost we see here (but much optimization work is still to be done). This benchmark should at least provide a minimum (RTX off) and maximum (RTX on) guideline for the speed Octane can achieve on RTX cards.
Requirements:
- Windows (64-bit) 7 (SP1), 8 or 10 (1803).
- NVIDIA graphics card (CM 3.0 or higher, CM 7.5 is required for RTX support).
- NVIDIA driver 416.94 (GeForce) or 416.78 (Quadro) or newer (except 418 series).
Limitations:
- Benchmark result uploads have been disabled.
- This build will expire in 120 days.
FAQs
Q: How do I know my GPU is using hardware for ray tracing acceleration?
A: At this stage only Turing cards support accelerated ray tracing, if you have the right driver you should see `RTX √` next to your GPU's name.
Q: I have an NVIDIA RTX series GPU but it does is not displayed as having RTX support.
A: Make sure you have one of the drivers stated above or newer.
Q: Does this work with GTX cards?
A: It will work but you won't get any speedup measurements as RTX is not supported.
Q: Does this leverage NV Link?
A: No, the scene in this benchmark will not make use of peer-to-peer memory using NV Link. Future OctaneBench versions will specifically measure NV Link and out of core speeds, as well as denoising speed.
Q: Will you release Linux or OSX builds?
A: We can release them as soon as NVIDIA provides with suitable drivers, no ETA for that at the moment.
Q: When will a version of Octane with full RTX support be released?
A: RTX support is currently planned as an experimental feature in Octane 2019 (with a first integration coming in 2019.2).
Q: What performance boost should I expect?
A: That depends heavily on the scene, we've experienced speedups of up to 5x for best case scenarios, this will be lower with scenes with heavier shading or smaller geometry sizes. For this specific benchmark scene which has 1.7M triangles we have seen figures ranging from 2.5x to 3.x.
Q: Does this new technology affect the quality of the final render?
A: Although it is not directly related to the final quality of your render, using hardware acceleration for ray tracing translates into faster render times so you can render more samples and get a cleaner image in the same amount of time as before.
Q: Do I need to install the Windows 10 October 2018 Update for this to work?
A: No, however the required drivers may require you upgrade to version 1803 (April 2018 Update).
Downloads
RTX OctaneBench 2019 Windows Installer (UPDATED)
RTX OctaneBench 2019 Windows Zip (UPDATED)
Happy rendering,
Your OTOY Team
