New RADV Code Can Deliver 10x Faster Ray-Tracing Pipeline Compilation For Some Games
([Radeon] 14 January 12:42 PM EST
Radeon Vulkan)
- Reference: 0001606136
- News link: https://www.phoronix.com/news/RADV-10x-Fast-RT-Pipeline-Comp
- Source link:
A new merge request opened today for Mesa's Radeon Vulkan driver "RADV" by Valve contractor Natalie Vock provides another significant boost for the Vulkan ray-tracing performance in multiple titles.
Over the past year [1]RADV ray-tracing performance improved a lot as one of [2]the lingering areas where RADV remained behind the now defunct AMDVLK driver and where NVIDIA has long had a performance advantage with ray-traced games. Already this year there have been [3]some nice performance improvements while another exciting merge request is pending for bringing more gains.
The key takeaway for AMD Radeon Linux gamers with today's pull request:
"Compiling RT pipelines in UE4 games with raytracing (e.g. Ghostwire Tokyo, The Callisto Protocol) becomes 10 times faster. Yes, an order of magnitude! In one Ghostwire Tokyo Fossilize capture I gathered, time to replay went from 4 minutes and 20 seconds to just 20 seconds. These UE games also tended to have quite terrible stuttering whenever a new RT pipeline was compiled. That stuttering is gone completely."
[4]The merge request now uses function calls to separate out any-hit/intersection shader compilation. Function calls for ray-tracing are now used for "really cool stuff" in building off [5]this now-merged code that hit Mesa Git earlier today.
Natalie Vock added in today's new merge request:
"On top of that, runtime performance improves by a lot [in affected applications] as well. Who knew that inlining hundreds of shaders into an incredibly hot loop might be bad for performance?! From quick napkin math, I think the pure RT performance in Ghostwire Tokyo improves by over 2x. In any case, FPS goes from ~30 to ~40 on my 7900XTX.
It seems like with the MR, we roughly match Windows performance in the Ghostwire Tokyo scene I tested, as well. Performance improvements on different apps/non-UE4 titles may vary, but I'm pretty sure quite a few apps should benefit. (Cyberpunk 2077 doesn't, though. It only really uses 1 any-hit shader at the maximum and is therefore unaffected by this MR.)"
It seems RADV is off to a really great 2026.
The code is now under review. With Mesa 26.0 code branching / feature freeze imminent, it might not be merged in time for this quarter's Mesa 26.0 release in which case it may not be found in a stable release until Mesa 26.1 in Q2.
[1] https://www.phoronix.com/review/radeon-radv-rt-2025
[2] https://www.phoronix.com/review/radeon-radv-amdvlk-final
[3] https://www.phoronix.com/news/RADV-UE5-Lumen-Merged
[4] https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39314
[5] https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29580
Over the past year [1]RADV ray-tracing performance improved a lot as one of [2]the lingering areas where RADV remained behind the now defunct AMDVLK driver and where NVIDIA has long had a performance advantage with ray-traced games. Already this year there have been [3]some nice performance improvements while another exciting merge request is pending for bringing more gains.
The key takeaway for AMD Radeon Linux gamers with today's pull request:
"Compiling RT pipelines in UE4 games with raytracing (e.g. Ghostwire Tokyo, The Callisto Protocol) becomes 10 times faster. Yes, an order of magnitude! In one Ghostwire Tokyo Fossilize capture I gathered, time to replay went from 4 minutes and 20 seconds to just 20 seconds. These UE games also tended to have quite terrible stuttering whenever a new RT pipeline was compiled. That stuttering is gone completely."
[4]The merge request now uses function calls to separate out any-hit/intersection shader compilation. Function calls for ray-tracing are now used for "really cool stuff" in building off [5]this now-merged code that hit Mesa Git earlier today.
Natalie Vock added in today's new merge request:
"On top of that, runtime performance improves by a lot [in affected applications] as well. Who knew that inlining hundreds of shaders into an incredibly hot loop might be bad for performance?! From quick napkin math, I think the pure RT performance in Ghostwire Tokyo improves by over 2x. In any case, FPS goes from ~30 to ~40 on my 7900XTX.
It seems like with the MR, we roughly match Windows performance in the Ghostwire Tokyo scene I tested, as well. Performance improvements on different apps/non-UE4 titles may vary, but I'm pretty sure quite a few apps should benefit. (Cyberpunk 2077 doesn't, though. It only really uses 1 any-hit shader at the maximum and is therefore unaffected by this MR.)"
It seems RADV is off to a really great 2026.
The code is now under review. With Mesa 26.0 code branching / feature freeze imminent, it might not be merged in time for this quarter's Mesa 26.0 release in which case it may not be found in a stable release until Mesa 26.1 in Q2.
[1] https://www.phoronix.com/review/radeon-radv-rt-2025
[2] https://www.phoronix.com/review/radeon-radv-amdvlk-final
[3] https://www.phoronix.com/news/RADV-UE5-Lumen-Merged
[4] https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39314
[5] https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29580