DF Weekly: Avatar is the latest PS5 Pro patch that looks worse than the base PS5 game
Why is this happening, should Sony step in and can it be fixed?
In this week's DF Direct Weekly, we talk about the mooted PlayStation handheld (spoilers: it won't run native PS5 games) but it's our second news topic of the week I'll be writing about in more depth today. Avatar: Frontiers of Pandora has arrived on PlayStation 5 Pro and its spec points look impressive - you're getting the quality mode visuals of the base console version at 60fps. There's just one problem: the Pro version looks a lot worse than the standard console's quality mode and even its performance mode has some quality advantages over the Pro version. In which case, what's the point?
We should expect to see some outlier software with the arrival of any enhanced console - games that aren't quite delivering what they should. In fact, we saw it with the launch of PlayStation 4 Pro back in 2016. If memory serves, The Last of Us Remastered could run at a lower frame-rate on Pro than it did on the base PS4 - and Watch Dogs 2 also had similar issues. However, these were generally exceptions and were quickly patched. While the quality of Pro versions could vary dramatically, a noticeably worse image wasn't the issue - and yet, this is clearly happening with PS5 Pro. So many titles are affected now that many users are asking for the ability to run PS5 Pro-enabled games without the Pro features enabled. Instead, those games would run with standard model features, with only the extra CPU and GPU horsepower of the new console in use.
My personal feeling on this is that it's a short term "solution" of sorts to a problem that will likely be resolved in the medium term, but to understand the problem is to focus on Sony's machine learning-based upscaling solution: PlayStation Spectral Resolution, or PSSR. This is essentially the PlayStation team's answer to Nvidia DLSS or Intel XeSS, where game engine inputs and a lower resolution are fed into a neural network, delivering an upscaled output fit for a 4K screen.
- 0:00:00 Introduction
- 0:01:19 News 1: Potential portable PlayStation in development
- 0:18:08 News 2: Avatar PS5 Pro patch shows PSSR struggles
- 0:31:25 News 3: Star Wars Outlaws patch brings PC fixes
- 0:37:47 News 4: Intel Arc B580 GPU leaked
- 0:49:42 News 5: Light of Motiram takes heavy Horizon inspiration
- 1:00:47 News 6: Valve celebrates Half-Life 2 20th anniversary with update, new documentary
- 1:19:42 News 7: Dynasty Warriors Origins demo drops
- 1:25:31 News 8: Highlighting Brazil Pixel
- 1:36:33 Supporter Q1: Will CD Projekt Red be able to fix UE5 stuttering?
- 1:44:08 Supporter Q2: How will Microsoft configure their business if Steam can run on Xbox?
- 1:51:21 Supporter Q3: Why does Windows 11 cause so many headaches?
- 2:01:58 Supporter Q4: Why haven't we heard more about a PS5 Pro patch for Black Myth: Wukong?
- 2:07:45 Supporter Q5: How does Nintendo avoid technical issues in their games?
- 2:14:19 Supporter Q6: What would a theoretical PS3 Pro look like?
PSSR has delivered many excellent results: Stellar Blade looks absolutely phenomenal, Final Fantasy 7 Rebirth is a game transformed, while all of Insomniac's games enjoy substantial upgrades thanks to PSSR. There are many others, including Naughty Dog's The Last of Us Part 1 and 2. However, Silent Hill 2 did - and does - have issues despite being patched, Star Wars Jedi: Survivor can look pretty hideous at times, while Star Wars Outlaws also has some profound issues. Avatar: Frontiers of Pandora is especially impacted because PSSR has issues with transparency effects - most notably foliage - and also seems to have quality problem with ray tracing inputs. It's also shown to lose a lot of quality the higher the upscaling factor. Avatar: Frontiers of Pandora seems to operate with a 720p to 1800p dynamic resolution window, which paired with foliage and RT combines to create a 'three horsemen of the apocalypse' effect which ultimately makes the game look worse on Pro than it does on the standard console.
Right now, the sense is that PSSR can be as detrimental to a game's presentation as it can be a boon - and that's a problem when you're selling an expensive enhanced console upgrade to the most dedicated fans in your userbase. The quality bar is very, very straightforward: a Pro-enhanced game cannot and should not ship if it's worse than the output of the standard model. In terms of quantifiable metrics, frame-rate shouldn't be lower and to be fair, we've not seen anything like that from Pro software so far. However, things are trickier when dealing with image quality where it's typically a more subjective appraisal. That said, Avatar is losing so much detail owing to PSSR 'noise' that a cursory A to B comparison should do the trick. Foliage (Jedi: Survivor) or RTAO (Dragon's Dogma 2) shouldn't be flickering noticeably when it does not do so on the original PS5. At some point the question needs to be asked: can't developers see what's going wrong here?
We've often talked about an apparent weakness in QA that allows for 'easy to fix' but impactful problems to make their way into shipping code - and that's still a problem - but what we're looking at here is a basic failure to recognise and to fix superficial issues. It's very, very difficult to understand how the worst of the Pro problems ever made it to users and it suggests that developers are perhaps over-stretched and that QA really is lacking. At this point, there's also an argument that the platform holders themselves should do more as it's not a good reflection on their own hardware.
There has been some tacit recognition that PSSR isn't the complete solution right now and should perhaps be deployed more sparingly. Guerrilla Games didn't use it for its Horizon games and has delivered what we think is the best image quality we've seen in any console game. Both Santa Monica Studio and Polyphony Digital have delivered good PSSR support, but even then, have included the option to drop back to prior upscaling solutions. Meanwhile, with Fortnite on PS5 Pro, Epic has chosen to retain its own TSR upscaler - a technology deeply embedded and integrated with Unreal Engine 5.
The obvious conclusion is that Epic didn't feel PSSR is a good fit for UE5 right now - born out with the UE5-driven Silent Hill 2, where Bloober Team's swap back to TSR for its performance mode has ironed out its image quality issues. And yet, PSSR remains in the 30fps quality offering, which despite the higher resolution arguably looks worse overall than the 60fps performance alternative.
As mentioned earlier, a possible solution is to offer a system-level 'opt out' of Pro upgrades and allow for Game Boost to operate on the standard PS5 version of the game - and such an option is said to exist on development Pro hardware. However, introducing a kind of fallback solution to mitigate failed Pro upgrades sends the wrong message when it's down to developers, publishers and the platform holder to get it right with their software.
So, where do we go from here? More thorough testing of Pro upgrades is a given, based on some of the most egregious examples we've seen. Rather than add a system level feature to disable all Pro features, a more streamlined solution may be to allow for PSSR to be a toggle, as it is in God of War Ragnarok or Gran Turismo 7. We've tested many, many games now and there do seem to be commonalities in where PSSR has issues - maybe there are problems in how it is being implemented, but for so many studios to deliver software with similar issues, it does suggest that PSSR requires some work.
We've heard plausible stories of different versions of PSSR making their way into titles with varying quality levels, along with continual quality improvements - which leads us on to the next point: how best to roll out PSSR improvements. On PC, all games using DLSS from version 2.0 contain a .dll in the game install folder that can be swapped out with later versions, exhibiting clear quality improvements. Nvidia now has four years of DLSS 2.0 development under its belt and while developers can patch their games to introduce later versions of the scaler, users can swap in new DLLs and literally improve anti-aliasing and upscaling in their own games. In theory, Sony could introduce a system level feature to automatically 'patch' PSSR to the latest version for all software. Ultimately, for a console experience, I still feel developers need to validate PSSR updates - but perhaps Sony can play its part in making the patching process easier, or even carry out the integration work itself subject to developer approval. Whichever way you slice it, there's clearly an issue that requires the platform holder to step in.
Ultimately, what we are looking at here is the first console to arrive leveraging machine learning features - and there are going to be teething issues, just as there was with Nvidia DLSS back in the day. It'll take time to iron out some of these issues, but as machine learning is a crucial foundation for the next generation of consoles, I'm confident we'll get there. However, from a short term, consumer-focused perspective, we are dealing with a £700/$700/€800 enhanced console here - and in no way, shape or form should PS5 Pro versions of any game be worse than the same title running on the standard console.