← sources

Phantom Blade 0 Performance tested on RTX 5070 Ti w/ DLSS 4 & 4.5, FG.

source: reddit (reddit:nvidia)published 2026-10-30 16:27ingested 2026-10-31 02:39

retained text

First thing I want to note is you no longer have to revert back to Mvidia driver branch 617.xx as Nvidia rolled out an OTA update for this game's profile so the latest driver branch will work just fine. I tested this on the latest Hotfix driver FWIW.

System: 

CPU: AMD Ryzen 7 9800X3D

GPU: Gigabyte Windforce 5070 Ti OC

PSU: 1000w NZXT C1000G

Mobo: Asrock B850m PG Riptide w/Wifi

RAM: 32GB DDR5-6000 CL:30

On first boot you will go through a very long shader compilation. It took about 20 minutes or so on my rig. But after that's done the game runs BUTTERY smooth and I couldn't really see much of any further traversal or compilation stutter. I honestly hope more game dev's go this route, I have no problem waiting for a full comp for dozens of hours of smooth gameplay.

Settings: Ran it at 4K and I use DLSS Swapper to set the latest DLSS library file. The game ships with 310.9 I set Super Resolution to Recommended, Ray Reconstruction to Preset F, and FG to Preset B. I prefer the native dll files instead of the Nvidia App API calls as you can normally get the best performance and less artifacting if the files are natively sitting in the game folder. 

(Some games may reset your model Preset like Black Myth Wukong not behave but this one did not... and Multiplayer games with Anti-Cheat need the NV App override)

Secondly you will want to set the output resolution off of "Auto" and use 100% so you can use your native monitor resolution as the output res.

\---

Ray Tracing Off / Software Lumen

4K - DLAA (K) Ultra - 34fps Avg

Changing settings to Very High w/ all geometry & Texture settings to Ultra netted a 17% fps gain with minimal quality loss. All further tests are with my optimized settings.

4K - DLSS-Q(K) - 64fps Avg

4K - DLSS-B(K) - 70fps Avg

4K - DLSS-P(M) - 78fps Avg (No shimmering issues on reflections or foliage)

\---

Note: Similarly to Black Myth Wukong and Alan Wake 2, This game "kidnaps" Reflex behind Framegen so you actually cannot get the snappiest response time without FG activated in this game.

FG: 2x, Preset B - 4K - DLSS-P(M) 122fps Avg - PCL: 42ms (w/o FG: 72ms)

\---

Ray Tracing: On

4K - DLAA(K) - 28fps Avg

4K - DLSS-Q(K) - 49fps Avg

4K - DLSS-B(K) - 56fps Avg

4K - DLSS-P(M) - 61fps Avg

Performance w/2X FG: 102fps (PCL: 49ms)

Performance w/3X FG: 137fps (PCL: 51ms)

Turning on Ray Reconstruction w/ Performance Upscaling and no FG incurred a  11% framerate penalty bringing Avg fps to 55

\---

Path Tracing: On You cannot turn off Ray Reconstruction at this point. All performance tested with Preset F from the 4.5 update.

4K DLAA-RR - 21fps

4K RR-Quality - 37fps

4K RR-Balanced - 44fps

4K RR - Performance - 52fps

w/ FG 2X - 85fps (PCL: 56ms)

w/FG 3X - 113fps (PCL: 60ms)

I will note that Framegen has a very good implementation in the game although it does seem to take a pretty big hit in performance. I couldn't see much of anything in the way of distracting artifacts.





document fetched 2026-10-31 00:38 · canonical https://reddit.com/r/nvidia/comments/1xjk0vq/phantom_blade_0_performance_tested_on_rtx_5070_ti