Dragon’s Dogma 2 – Patch 2.03 Patch
Share the Story:
Stay out of town
Capcom released a new update for Dragons Dogma 2 earlier this week and I was keen to test it out across PS5 and PC with a quick look at Series X|S for completeness. At first glance it appears to have added new modes, but not all is as it seems as the focus here maybe on hardware, not yet available.
Staring with PS5, the High mode of old – the single mode PS5 and Xbox Series X launched with – offers Ray Traced GI and SDF AO using RE engines interlaced 3840×2160 Checkerboard like mode. This benefits from the ID buffer hardware from the Pro and is supported on PS5. Since launch, a mere 6 months ago, the game has been patched often. As I discussed in my IGN performance review would be the case which included a brand new Low mode, sans RT, which offered the biggest boost to frame times. Largely helping the load on the CPU and memory bandwidth, but also dropping resolution and effects such as grass, alpha and LoD. A huge boost about month into launch, but since then only small updates have come to add minor increments on that. The PC version did benefit slightly more at times dues to the inherent architectural differences between the platform and consoles. Fundamentally, the main problem remains as I discussed in that review. State models, dynamic reactions, path finding, and logic all cause big spikes in frame times and the quick fix is to cull NPCs all over the game. This creates a Back to the future scenario of people fading in and out of existence all the time to reduce the amount of data and contention. Ray Tracing is a big CPU burden along with GPU, hence why the Low settings can boost frame rates by some 35% on PS5 and Series X.
What changes in patch 2.03
The latest patch (2.03) that dropped on the 17th October this week came with a new guide on expected rates and resolutions along with a new menu name. Low now becomes Favour Performance and the old High is now Favour Graphics and almost certainly are built for the PS5 Pro update. Dynamic Resolution Scaling may now be a factor here to help edge cases but in reality the game is often fully GPU or CPU bound in that high/graphic mode or both. And then more often CPU bound in the Low Performance mode when in towns or hubs. Testing prior patch 2.020 to latest 2.030 we see almost identical performance levels in those memory/CPU bound segments well within margin of error. Low to High can fluctuate between 20 to 44% faster, section to section, ending up being 35% faster on average over high. And we see exactly those same levels between the new Performance and Graphics mode now, reinforcing the fact this is little more than a rebrand and no big changes have happened here, aside one. It appears a lower resolution base now in performance, the notes state 3072x1728p upscaled to 4K now, a 36% pixel reduction versus the 4K Graphics mode. The interlaced dithered pixels and strong TAA do create heavy ghosting and trails when motion is fast with high contrast which affects both modes. Along with dithering and noise on foliage, which can look like poor Depth of Field, worse in the performance mode. No changes were noted across the modes including the high quality RTGI and AO which can dramatically improve light, shadow and materiel accuracy indoor and out, best noticed with light seeping into doors, windows or canopies emphasising occlusion and reducing light leak and grounded objects more accurately.
What this offers for the PS5 Pro
Using the footage released by PlayStation running the the PS5 PRO, assuming the same patch version and settings which include RTGI, AO and a 4K output with a twist. The PS5 Pro PSSR, rather than RE Engines own TAA and interlacing, resulting in an image that is significantly cleaner, sharper, more temporally stable than the PS5’s Graphics mode. It presents a huge increase from the limited footage released, (Note: I have viewed the ProRes footage but the video uses the public YouTube versions released across many sites). Ghosting is practically eradicated now across those same high contrast, fast moving sections. Texture details, material quality and image quality looks closer to the native 4K and/or DLSS image on PC now. From the limited footage, I cannot be certain but RTGI looks similar to the same settings as before but motion blur looks less aggressive than PS5 with increased shutter speed giving a lighter blur on per object and camera motion, but this is a post effect that can be impacted by TAA and reconstruction depending where in the render pipe it is processed.
How does it compare to PC and PS5
Running the latest patched version on my RX6800 and Zen 3 5600X machine, allegedly a match for PS5 Pro some say, in the same section and settings as Base PS5. 4K interlaced, Ray Tracing on, we are significantly CPU bound with it often only using 70% command load of the GPU and far lower power draw and clock speeds, resulting in higher performance than the base console on the same patch, across the same run. However, this can be surprisingly close with a high delta of 40% to the PC and then a close 8% between, finishing this short run 24% faster on PC that becomes increasingly CPU/Memory bound the more we draw into the town square which can close the gap up between them.
What this also demonstrates is the PS5 Pro pulling ahead over this PC spec, in this game/section with superior IQ and performance than both base PS5 and this PC machine where it appears to be fully CPU and or memory limited. The PS5 Pro remains within the 33ms frame times here with no 50ms spikes, which occur often on PC and are rife on PS5 with RT enabled. Ensuring I cover all bases, the Low mode on PS5 can still run below the Pro here, between 10-21%, which means that is the best-case scenario for potential CPU/data bound games on the Pro and points to possible cache or architecture changes here as I often say, Teraflops, ALU count, cores, frequency are not as vital as many believe. I caution all though, this is early days and if we are looking at exact matches from Graphics mode across PS5 and Pro, then here the professional console is almost 60% faster in the same run and 33% on average than my RX6800/5600x machine, an impressive debut for sure. Although too early to be classified concrete, but once I get hands on, I will be sure to show and tell you all the facts and results.