TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Raytracing on AMD’s RDNA 2/3, and Nvidia’s Turing and Pascal

100 pointsby treesciencebotabout 2 years ago

6 comments

pixelesqueabout 2 years ago
I suspect the reason the author is seeing very shallow trees for Nvidia might be because the lower levels are done fully behind the scenes:<p><a href="https:&#x2F;&#x2F;forums.developer.nvidia.com&#x2F;t&#x2F;extracting-bvh-from-optix-to-manually-traverse&#x2F;235250" rel="nofollow">https:&#x2F;&#x2F;forums.developer.nvidia.com&#x2F;t&#x2F;extracting-bvh-from-op...</a><p>As someone who deals with BVHs a lot for ray intersection, I find it pretty difficult to believe that leaf nodes with that number of primitives will be anywhere near performant, even with fast dedicated hardware like the RT cores.<p>It&#x27;s true that the Nvidia cards have better intersection performance than ray&#x2F;box tests, but I don&#x27;t believe it&#x27;s in the 100x ratio range which I suspect would be needed if the BVHs were that shallow and leaf nodes that large.
评论 #35266248 未加载
评论 #35266634 未加载
matthewfcarlsonabout 2 years ago
I&#x27;ve often wondered why Nvidia cards are generally so much better at rendering scenes in Blender&#x27;s cycles renderer (a raytracing engine). The benchmarks on Blender&#x27;s website are really telling (<a href="https:&#x2F;&#x2F;opendata.blender.org&#x2F;benchmarks&#x2F;query&#x2F;?group_by=device_name&amp;blender_version=3.3.0" rel="nofollow">https:&#x2F;&#x2F;opendata.blender.org&#x2F;benchmarks&#x2F;query&#x2F;?group_by=devi...</a>) by the fact that the only non Nvidia entry on the first page is the AMD 2X EPYC 9654 96-Core.<p>This really lays out the decisions that Nvidia made compared to AMD and how their approach tends to hide some of the shortcoming of GPUs (latency and utilization).
评论 #35266769 未加载
ladbergabout 2 years ago
Would love to see a more in-depth article on BVH construction itself! I&#x27;m decently familiar with the main concepts but have no clue what the current SOTA looks like (is that even public info?).<p>BVH construction is my favorite question to ask in interviews because there&#x27;s no single best solution and it mostly relies on mathy heuristics to get a decent tree. You can also always devote more time to making a more optimal tree but there&#x27;s a tradeoff where it&#x27;ll eventually take more time than it saves in raytracing.
评论 #35267381 未加载
shmerlabout 2 years ago
Ray tracing on Linux for CP2077 with 7900 XTX is still barely usable, but it&#x27;s getting better.<p>I&#x27;d say RDNA 3 is not really giving useful ray tracing on for example 2560x1440 unless you use upscaling to speed it up. May be in a few GPU generations ray tracing will become usable with native resolutions.
评论 #35268289 未加载
br1about 2 years ago
Interesting that card&#x2F;drivers customize so much of ray tracing, like rasterization in pre vulkan&#x2F;metal&#x2F;d3d12 or even fixed function gpu days.
sylwareabout 2 years ago
I did not get into the real details yet, but mesa radv pulls that horrible glslang due to some shaders related to acceleration structures.<p>Personnaly, I am a dev, then I patch to compile out all that (and all the tracers at the same time) since ray tracing has currently a ridiculous ratio benefits&#x2F;technical costs.<p>This defeats the very purpose of vulkan spirv: getting rid of those horrible high level shader compilers from the driver stack and keep them contained at the application level.<p>It seems beyond clumsy, but as I said, I need to get into the details of &quot;why&quot; those shaders in the first place, and then why they are not written directly in RDNA assembly or SPIR-V assembly (that would require an &quot;assembler&quot; coded in simple and plain C).
评论 #35268163 未加载