Portal with RTX has the most atrocious profile I have ever seen. It's performance is not (!) limited by it's RT performance at all, in fact it is not doing much raytracing per time bin on my 6900 XT.
This is one GIGANTIC Ubershader that takes 99ms of frametime by itself. The hardware utilization is horrifically bad, and it's using the absolute maximum amount of VGPRs possible on RDNA2.
To show how bad it is, I have taken a capture from the start area in Portal with RTX. This has 323 VALU instructions per RT instruction, which is a BAD ratio.
Compare that to Control, which is generally regarded as a bad Game in terms of RT performance on RDNA2. Here we have a ratio of 82 VALU per RT instructions in the reflection shader, which is in the normal range (according to @b_nieuwen.)
So, if RT performance can't explain the poor performance, then what can? Well, one explanation would be vector Registers spilling to cache or VRAM. This shader is pegged at 256VGPRs, which is the maximum that's possible to use per wavefront.
I haven't tested it on RADV, but according to @b_nieuwen the situation is the same there, so it's likely not an issue with the Windows driver.
This is an interesting hypothesis on what could have gone wrong:
I have been annoyed with PC Gaming this week, so I think I should continue this thread. I have been annoyed by a visual bug with the RT Reflections in the @Steam release of @ControlRemedy for more than a year now and had basically given up hope:
So when I started a new game last night on the EGS version, I was stunned when I got to the same spot and saw that the issue wasn't there (first screenshot.)
I eventually copied all RT shader object files from the EGS release to the Steam Version, and the issue was gone.(second)
So, what was going on, and why didn't any hardware reviewer complain about visual bugs with Control RT during their testing? Well, the EGS release is using a newer build (0.0.344.1879) than the Steam/GoG release (0.0.344.1873), and all reviewers tested the EGS version.
Regarding the recent @digitalfoundry Scorn video. I believe the "FSR2 is particularly expensive and can take many milliseconds to compute" take is based on the testing in Deathloop earlier this year. Unfortunately those results were ignoring the native-resolution post-processing.
Further benchmarks to prove this point will be coming soon, but here is a preview of FSR2 in Scorn.
1440p native and 4K FSR2 Quality (internally 1440p)
Compute Pass 7 is either TAA or FSR2, Color Pass 10 seems to be post-processsing.
How can I be sure that this is the FSR/TAA pass? -- All render targets before this have a 2560x1440 resolution, while 4K buffers start showing up afterwards:
So it's time for the followup to this thread.
First off, I am severely disappointed by anyone who chose Lithium. Lithium is a trace element on earth, which makes up less than a percent of a percent of earths crust. (0.004% in fact) 1/10
(p.20 of pubs.usgs.gov/pp/0127/reportβ¦)
Secondly, this was a trick question, and the correct answer is either Aluminium, Iron, or Sodium, depending on your perspective.
Aluminium is the most abundant element on earth that can be used as a cation. Silicons chemical properties make it most likely unusable for this. 2/10
And while Oxygen can be oxidized by Flourine, those compounds are not practical for use in batteries.
So, Aluminium-Ion Batteries: This sounds great. Aluminium has a robust supply chain, it's available world wide and it is relatively cheap. What's the catch? 3/10
Let me explain what I think is the reason for the terrible RT performance of AMD cards in the gallery outside the NSC Control Room, the so called "Corridor of Doom." (@Dachsjaeger)
TL;DR: This scene is accidentally hitting every single worst case scenario at once.
A word of warning: This is going to be complicated for anyone who is not familiar with the way Acceleration Structures are handled in DXR and Vulkan RT. I suggest reading up on the relationship between TLAS and BLAS. (Section 8) developer.nvidia.com/rtx/raytracingβ¦
DXR uses 2 levels of Acceleration Structures, Top Level Acceleration Structures (TLAS) and Bottom Level Acceleration Structures (BLAS).
The TLAS contains a number of instances of BLAS, these are identical copies of BLAS that are translated, rotated and scaled in 3D space.
I don't think people know how janky Linux Gaming was a few years ago and how much better it has gotten.
Up until Proton/SteamPlay was introduced, most Linux gamers had 2 or more Steam installation. One for your native games, one under Wine for the unproblematic games, and...1/6
another customized Wine based installation for each of the more problematic games. Lutris and PlayOnLinux were a godsend for managing these installations and quirks. Lutris is even capable of setting up a custom Wine/Proton installation for a lot of GoG or EGS games. 2/6
Optimus hybrid graphics were broken under Linux, mostly due to the nVidia's proprietary driver and it's graphics pipeline, until driver 435.17 from 2019! IIRC the recommended solution required a reboot for each switch from iGPU to dGPU. 3/6