Current Generation Games Analysis Technical Discussion [2023] [XBSX|S, PS5, PC]

Status
Not open for further replies.
I posted this in a dumber thread, but I decided to take a quick look at fortnite to see what the performance is like with nanite and virtual shadow maps. I play fortnite with everything on low (including textures) except view distance epic, lumen off, nanite geometry on, and DLSS Quality at 1440p. I have an RTX 3080 10GB. Landed in a quiet spot and was getting 170 fps at that particular view. Turned on virtual shadow maps (medium is the lowest setting) and my fps dropped to 120. Changed virtual shadow maps to ultra and my fps dropped to 90. Then I switched to TAA with 100% 3D resolution and my fps dropped to 65. And that's with textures, effects, post processing etc all turned to low. That was landing in a quiet area on a small hill overlooking some trees.

The cost of the virtual shadow maps is pretty high. I don't know how the settings in fortnite compare to the ones in Remnant (ie is the medium setting in both equivalent?). Just at a glance, the fortnite performance doesn't actually seem too far off of what I've seen of Remnant 2.
 
I posted this in a dumber thread, but I decided to take a quick look at fortnite to see what the performance is like with nanite and virtual shadow maps. I play fortnite with everything on low (including textures) except view distance epic, lumen off, nanite geometry on, and DLSS Quality at 1440p. I have an RTX 3080 10GB. Landed in a quiet spot and was getting 170 fps at that particular view. Turned on virtual shadow maps (medium is the lowest setting) and my fps dropped to 120. Changed virtual shadow maps to ultra and my fps dropped to 90. Then I switched to TAA with 100% 3D resolution and my fps dropped to 65. And that's with textures, effects, post processing etc all turned to low. That was landing in a quiet area on a small hill overlooking some trees.

The cost of the virtual shadow maps is pretty high. I don't know how the settings in fortnite compare to the ones in Remnant (ie is the medium setting in both equivalent?). Just at a glance, the fortnite performance doesn't actually seem too far off of what I've seen of Remnant 2.
You pay a huge performance tax from DX12 alone. That could be a large part of the prolem in any UE5 game.
 
You pay a huge performance tax from DX12 alone. That could be a large part of the prolem in any UE5 game.

Pretty sure if I turned off Nanite and stayed with my low settings in Fortnite I'd be pushing 300+ fps with dx12. Nanite is probably costing ~2.5ms more. Medium virtual shadow maps about ~2.5ms as well. Epic virtual shadow maps cost ~5.2ms. Those costs will go up at native res. This is all an estimate, but the new tech is costly. If you push AA artwork through that pipeline people may not perceive the benefits because they can't separate the art from the technology.

In the quick test area, I was getting ~5ms added per frame for nanite and medium virtual shadow maps, and ~7.7ms for nanite and epic virtual shadow maps.

3.33ms (+ 0.0 ms) - 300 fps - dx12 all low, dlss quality @1440p
5.88ms (+ 2.5 ms) - 170 fps - added nanite
8.33ms (+ 2.5 ms) - 120 fps - added medium virtual shadow maps
11.1ms (+ 5.2 ms) - 90 fps - added epic virtual shadow maps
 
Pretty sure if I turned off Nanite and stayed with my low settings in Fortnite I'd be pushing 300+ fps with dx12. Nanite is probably costing ~2.5ms more. Medium virtual shadow maps about ~2.5ms as well. Epic virtual shadow maps cost ~5.2ms. Those costs will go up at native res. This is all an estimate, but the new tech is costly. If you push AA artwork through that pipeline people may not perceive the benefits because they can't separate the art from the technology.

3.33ms (+ 0.0 ms) - 300 fps - dx12 all low, dlss quality @1440p
5.88ms (+ 2.5 ms) - 170 fps - added nanite
8.33ms (+ 2.5 ms) - 120 fps - added medium virtual shadow maps
11.1ms (+ 5.2 ms) - 90 fps - added epic virtual shadow maps

Even with identical settings you pay a huge penalty just to enable DX12. It isn't exclusive to AMD GPUS either.
 

At 2:39 the water comparison, looks like they forgot to add caustics on PC

And SSR reflections are definitely bugged on large bodies of water.

Worth noting the RT quality (including shadows) has also been improved on PC this morning via a hotfix. It'll be interesting to see how it effect those issues but I would assume it fixes the over darkening on the grass if it is hopefully now using the correct BVH.

Water caustics is a big issue though, that looks awesome on the PS5, it's a very strange one to miss out. Also, why SSR artifacts with RT reflections? They must be combining effects but it's definitely breaking things.
 
Baldurs Gate 3 benchmarks with DX11 and Vulkan: https://www.techpowerup.com/review/baldur-s-gate-3-benchmark-test-performance-analysis/5.html

DX11 is faster and uses less memory on nVidia GPUs. Shows how good the DX11 driver and how bad a low level API is. Alone 1.5GB or 20% more VRAM can make a difference if a game is playable on a 6GB card.
DF had some details on this -- unsurprisingly, vulkan performs better when (I assume rendering workload) cpu bound: https://www.eurogamer.net/digitalfoundry-2023-baldurs-gate-3-directx-11-vs-vulkan-which-is-best
Disappointing that the gpu performance is a little worse.
 
Status
Not open for further replies.
Back
Top