Death Stranding 2 may be the final big single-player Sony game to come to PC, so we were curious to see how well it runs on a mid-range system and scales to more powerful machines. The inclusion of ray tracing, which wasn't present in the PS5 and PS5 Pro versions of the game, also makes the PC release unique. The Nixxes-developed port was released around a month ago, so we're a bit late to the party here, but it's only recently that patches have brought the title into what we'd call a good working window. The initial release included PCIe bandwidth issues, performance problems and other bugs, and felt more like a beta release than a finished game.
We started our testing on a mainstream system built around the venerable Ryzen 5 3600, a mid-range CPU from 2019, and the RTX 4060, the entry-level Nvidia option from 2023. The game runs well using optimised settings, easily maintaining 60fps at a 1440p output resolution using DLSS balanced mode. Hitting higher frame-rates requires more CPU grunt, but it's good to see that the game isn't terrifically CPU-limited. Likewise, GPU performance feels reasonable here for the fidelity on offer, though the 8GB of VRAM on this card (and many other mainstream Nvidia options) means that turning down texture quality settings is a matter of course. There are no disruptive stutters to report either, beyond the momentary blips on cutscene transitions that we've seen before, and using dynamic resolution scaling (DRS) ensures a solid 50-60fps throughout the first three chapters.
Death Stranding 2 on PC also marks the first release of Guerrilla Games' proprietary Pico ("progressive image compositor") upscaling, which is found in the PS5 and PS5 Pro versions. That gives us the novel opportunity to compare Pico to DLSS in the same game, and what we found on our RTX 4070 at 4K resolution surprised us. As the upscaler benefits from being tightly integrated into the Decima engine, it actually outperforms DLSS 4.0 in some areas like foliage, while being slightly more performant than DLSS 4.5 using a matched 1440p internal resolution. It's also platform-agnostic, so Nvidia, AMD and Intel users can all benefit from it, and we recommend giving it a try.
Are you going to play Death Stranding 2 on PC? (172 votes)
- Yes, I've already bought it
- Yes, I plan to get it soon
- Maybe, I will wait for more information or a sale
- No, I'm just reading out of interest
The upscaler of your choice can be used alongside DRS to lock to a given performance target, though you may notice short frame-rate dips, possibly as it adjusts between scaling factors. The effect becomes more pronounced when v-sync is enabled, but less so with VRR enabled as we generally recommend. Still, with the horsepower to increase performance above 60fps on the RTX 4060 PC when using Pico or DLSS, having DRS enabled allows you to benefit from better image quality than you'd get from a fixed scaling factor designed to maintain the same 60fps lock throughout.
Let's move onto the optimised settings proper now. While performance is laudable overall, VRAM remains a key constraint and should be the first thing you consider changing on anything but the highest-end GPUs. The very high mode matches that on PS5, but even high approaches the limit on 8GB systems. Therefore, we recommend the medium setting for 8GB cards, even though it means that the RTX 4060 is unable to match the PS5 in terms of visuals, despite possessing enough compute power to do so. (The same issue also affects the RTX 3070 and 3070 Ti, despite being faster than the 4060 by a good margin - another victim of low VRAM allocations for this generation of cards.)
Reflection quality is another setting with a profound performance impact, especially when you jump from screen-space reflections (SSR) alone to ray-traced and SSR together. The standard high option offers the best balance and stability, equivalent to the PS5 presentation, with high RT introducing some noise and inconsistent performance, while very high RT looks great but saps frame-rates substantially. RT high is therefore recommended if you have GPU headroom, and we recommend pairing it with DRS to maintain more persistent performance.
Likewise, ray-traced ambient occlusion (RTAO) looks good but drops performance by seven percent in our test area, so we have opted for the PS5-matching high option for our optimised settings. If you have GPU headroom, then it's fine to use it.
|
Optimised Low-End (8GB VRAM) |
Optimised High-End (RT/8GB+ GPUs) |
PS5 Perf Mode |
PS5 Quality Mode |
|
|
Upscale Method |
Pico |
Pico |
Pico |
Pico |
|
Dynamic Resolution Scaling |
On |
On |
On |
On |
|
Texture Quality |
Medium |
Very High |
Very High |
Very High |
|
Texture Filtering |
16x |
16x |
2x |
2x |
|
Shadow Quality |
High |
Very High |
Very High or High |
Very High |
|
Shadow Resolution |
High |
High |
High |
High |
|
Screen-Space Shadows |
On |
On |
On |
On |
|
Ambient Occlusion |
High |
Ray Tracing |
High |
High |
|
Reflections |
High |
Ray Tracing (High) [w/ DRS enabled] |
High |
High |
|
Level of Detail |
High |
High |
High |
Very High |
|
Terrain Quality |
High |
High |
High |
Very High |
|
Cloud Quality |
High |
High |
High |
High |
|
Volumetric Lighting Quality |
High |
High |
High |
High |
|
Translucency Quality |
Default |
Default |
Default |
Default |
|
Depth of Field |
High |
High |
High |
High |
The level of detail (LOD) setting controls how far away full-resolution models are rendered, and here the very high preset has a substantial performance impact, so high is recommended. It offers around a 10 percent performance win versus very high with a minimal reduction in distant detail, aligning with PS5's performance mode. Cloud quality also impacts the view of distant vistas, and again it's the PS5-matching high option that we recommend versus the more expensive very high. Translucency quality, which affects the rendering resolution of transparent effects like particles, is best set to default, as it again has a heavy performance cost in some scenes but looks very similar to the higher setting.
With the biggest performance wins out of the way, let's quickly run through less impactful options. Terrain quality should be set to high to maintain complex geometry and surface detail, as going lower noticeably flattens distant areas. The high setting aligns with the PS5 performance mode and offers a six percent gain over very high. (A persistent bug intermittently reduces distant terrain detail and requires a restart to correct - something that has persisted across patches and needs Nixxes attention as it doesn't happen on PS5.) Shadow quality shows negligible performance changes across settings, so very high is generally recommended - unless you're on an 8GB card, where the 250MB VRAM saving is worth taking. Other settings show no significant performance savings, so we'll skip them. Finally, anisotropic filtering, which is limited to 2x on PS5, can be increased to 16x on PC without any performance penalty.
In total then, our final optimised settings look rather similar to the PS5 across the board, save for the VRAM concessions needed to hit mainstream Nvidia graphics cards. To test our optimised settings, we used a demanding dark scene with dynamic lighting and heavy foliage, back on the RTX 4060 at 1440p - with Pico balanced mode enabled and medium texture settings, so we're not limited by VRAM. Here, average frame-rates increased from 47fps on the very high preset to 60fps with optimised settings. Using DLSS balanced mode instead provides similar results, with a 57fps average. It's worth noting that the very high preset doesn't invoke RT, so the performance gulf will be even wider if you compare optimised settings to the game running at max settings with RT enabled.
It will be interesting to see if we do see any more big Sony releases ported to PC, as the strategy looks to have changed as of late and Nixxes no longer mentions PC ports on their website. Kojima Productions is nominally independent, but what about games from Insomniac, Sucker Punch or Naughty Dog? It would be a real shame for PC players to lose out on their future work.





Comments 8
Given that Shohei Yoshida recently said that porting PS games to Steam was basically free money, and given what we know about the rapidly rising gigantic budgets for Sony games this seems like the sort of play shareholders really won't like at all.
If the aim is to sell more Playstations I can't see it moving the needle. Sony first party games are exceedingly rare now, and PC gamers are now used to very high frame rates, ray tracing, ultrawide, mouse and keyboard controls, being able to play on a Steam Deck and many more PC only things. Would any of them really be able to justify the enormous asking price of a PS5 Pro just to play a small handful of first party games at 60fps on a pad?
I can't see it. But who knows, maybe Sony thinks differently.
Don't waste too much time hyping up RT; keep putting it aside for when it's added to the PS5 Pro.
It's insane that, on PC, the quintessential dynamic platform, you only run a locked-in configuration setting.
The Ryzen 3600, released before the PS5, should be retired, otherwise the console-centricity will be too obvious. It's not mid-range, it's the low-end of the last century.
Why does the low-end have to run at 1440p, and why shouldn't it use DLSS performance, which on the Switch 2 seems to work wonders, even at 1080p, and even with the worst version? Why does the PC have to suffer from the damage finder vs. the consoles' damage control, that seems clear to me.
Can the 4070 Ti only use the quality setting? The RT noise is annoying only on PC, while in unusual cases like MHW on the PS5 it's a win-win, even if it sounds like a hornet's nest?
Please, look carefully for cases where the RT isn't visible, and mix the 4070 TI with the 4060 to change the subject, which isn't really noticeable anyway...
We're waiting for the pre-mega patch for PS5 Pro, which will be analyzed before it's available on the PSSR, which was alpha until yesterday, where the significant improvements will be shown in great detail, more or less as has been done before.
You've left Eurogamer to take everything directly, eh...
@MemuAccount Not sure I grasp some of your points, but I'll reply to the ones I do understand.
RTX 4070 Ti - our contributor Rayan tests graphics performance on this because it has the Nvidia feature set and the 16GB of RAM required to see full utilisation of all settings and features. We use DLSS quality mode because DLSS is used by the vast majority of RTX users and often looks better than native 4K. We test at 4K because it's a good fit for the 4070 Ti's capabilities. If we lower the resolution, there's a good chance we'll be CPU limited. We need to be GPU limited ideally to show the GPU cost of each setting.
RTX 4060 - We use this because we have data on how this or RTX 3060 is the most popular GPU, via the Steam Hardware Survey. The RTX 4060 is also very very close to PS5 in terms of compute power, so a useful console comparison on expected performance.
Ryzen 5 3600 - We don't have data on which CPU is the most popular (the Steam Hardware Survey is just core counts basically), but we do know its performance profile is broadly similar to Zen 2 on consoles. It is a useful anchor point, therefore, on whether CPU performance lands according to expectations. This does. Something like the Outer Worlds 2 doesn't. Also, if a game runs =>60fps on a 3600, it sets a good baseline for more modern CPUs in the era of the high refresh rate display monitor and also means that 30fps minimum is usually doable on a handheld (from a CPU perspective at least).
Testing fixed configurations - the point of optimised settings is to offer up our suggestions for a fixed configuration that offers the best balance of visual features and performance. Of course, it is entirely down to the user whether they want to tweak further.
1440p DLSS balanced - 1440p is the sweetspot for gaming monitors and our understanding is that most gaming monitors being sold now are 1440p. DLSS balanced is the sweetspot for image quality at 1440p, though I'll grant you that performance mode can look OK in many games.
I don't understand some of your other arguments but I think I understand the tone which suggests you may not be interested in a good faith discussion any way.
@Rich_Leadbetter
"RTX 4070 Ti - our contributor Rayan tests graphics performance on this because it has the Nvidia feature set and the 16GB of RAM"
4070 Ti has 12GB Vram, are you sure you don't mean the 4070 Ti Super which does have 16GB Vram (which I own)? Or was the reference to 16 GB of Ram a reference to the system Ram?
With the spec difference between the 4070 Ti Super vs 4070 Ti non Super a clarification would be good.
Where are the Radeon cards such as the 9070xt (which I also own) in your PC performance/optimised settings articles?
Many times I have tried DFs optimised settings that are okay ish, on my 4070 Ti Super system, yet are no help for the more performant 9070xt system which can usually run games with much higher graphics settings and resolution than the former.
I have been gaming on the PC since the early 1990s (before that, home computers like Atari St, BBC B and Ti 99/4a) and I really had hoped that features locked to one vendor or another would have been consigned to history 30 years on and it would no longer be necessary to maintain two systems for gaming because of hardware companies and game developers anti consumer practices (which they are all guilty of at one time or another)
@Obsidian76 My mistake, it's a 4070 Ti Super, not the 4070 Ti.
"Many times I have tried DFs optimised settings that are okay ish, on my 4070 Ti Super system, yet are no help for the more performant 9070xt system which can usually run games with much higher graphics settings and resolution than the former."
The performance differentials may be slightly different but to say that optimised settings are "no help" is simply not true. Your game will run faster with optimised settings vs ultra - the point of optimised settings is to give you the best balance of performance vs visual quality.
Have you noticed how our optimised settings are nine times out of ten the same as PS5 settings? It's because the same principle applies even if different resolutions and different GPU architectures are in play. Even different graphics API in the case of PS5.
The general principle doesn't change: running everything at max is ultimately wasteful if you can lower the settings for very similar visual results.
The principle only doesn't apply if your Radeon GPU is producing different visual results to the output of the Nvidia GPU on the same settings which generally doesn't happen.
@Rich_Leadbetter
Maybe I should have been clearer, the optimised settings are of 'no help' to me on my 9070xt system (not my 4070ti super system, (currently with 5800x/x470/16gb Ram, awaiting swap with 5800x3d/x570/32gb Ram) where the DF optimised settings are useful as they can help hit my target frame rates of 72 - 138 hz (4k 144hz VRR display))
On the same display with the same frame rate target, using the 9070xt which is on 9800x3d/X870E/32gb Ram (replaced aforementioned 5800x3d), using PS5 equivalent settings is just leaving potential quality (i.e. not having to use FG, or having higher graphics settings or running at 4k native or Quality FSR4.1 rather than (with or without using optiscaler) dlss performance at my target frame rate) and performance improvements on the table.
Sure some of that can be down to the uplift in cpu performance by moving from Am4 to Am5/Zen 3 to Zen5.
I mean if I was happy with PS5 level quality and performance, my Series X would not be gathering dust as a glorified but rarely used Blu ray player.
What I would like to see in the future, post LLM bubble, from DF or others are PC optimisation guides/reviews that included recommended highest quality settings for targeting a minimum of 60fps (120fps with FGx2) on all three GPU manufacturer's / Both CPU companies (possibly Arm as well, with the N1X et al) at 1080p/1440p/4k for PC rather than just catering to the lowest common denominator.
But as it stands for the 9070xt system, wacking every setting up to ultra or higher works with my frame rate target in 90% of the games I play.
So as I said the optimised settings serve well enough for the 4070ti super system, which I thank DF for.
@Rich_Leadbetter
You're trying to force the science to do your bidding...
You can't match the most used DLSS settings with the 4070 Ti Super, nor can you plug in any monitor you want.
On Steam, 4K monitors/TVs are less than 5%, and cards that perform twice as fast as the PS5 are at least double. It's unclear why you need to go beyond that statistic to choose a monitor, or rather, it's all too clear...
With the 4070 Ti S, you're worried about being CPU-limited (and we don't even know what to do with it), but limiting the 4060 to lock it into the PS5 isn't a problem, it seems.
Aren't you interested in processor purchase data for over five years now? You chose the Ryzen 3600 to level the playing field with the PS5, you've said it 100 times, now you add reasons at will, perhaps to hide the fact that your publication is heavily console-centric?
Why are you talking about a beta right away? Just because it could have been improved at the beginning? How come I don't see the word beta in Starfield for PS5/Pro, even though it looks more like an alpha (the constant crashes on consoles are ridiculous), and why was the analysis released right away? Isn't there a wait? Isn't there a good critique? It doesn't seem bad, after all, there are just a few small issues...
Your unbalanced unit of measurement isn't invisible. Oliver always tries to justify and downplay the low resolution and frame rate drops on PS5 in every way: It's a bit softer, it mostly holds 60fps, etc.
The second excuse is grotesque; even a game that stutters at 200ms twice a second could fall into that category.
Why does DLSS on PC need 4K in quality mode to be better than native, while on the Switch 2, even in performance mode at 1080p—the worst-case scenario—it seems to work miracles?
How come, since DLSS's release, now more than 7 years ago, you've been talking and comparing AI upscalers more directly than in previous years? You clearly think you're invisible, huh?
On RE9, the 3070 needed to use DLAA, otherwise the limited VRAM wouldn't have been saturated enough. Even there, the 4070 TI S was on par with the Pro; the difference in textures, etc., was minimized without even being noticed, and the fear goes away. In the end, you managed to take the Pro to the next level from the 4070, one way or another.
Does the PS5 Pro's popularity matter?
The Steam charts feature stuff from 10 years ago, which only launches a few 2D indie games. At that point, on consoles, you'd have to include the Switch 1 as well, and maybe the older consoles.
Give me the approximate percentage the Pro would get among that crowd, and then explain why it's always in pole position; every sneeze is recorded.
It's clear that you're always trying to force the PC settings toward PS5; just as you can see, anything more is often downplayed, you forgot a piece... As I was saying, anything more will be well-regarded when it arrives on the Pro, or on PS6.
The frame-gen will exist later, as will the importance of 60fps, fast loading, and VRR, something you've talked about five times over the years before it arrived on consoles. After that, the stopwatch even counted tenths, and if it reaches 48Hz in fits and starts, that's good!
The longer version of damage control vs. damage finder.
@Obsidian76 I think the point I'm trying to make is that optimised settings are the best balance of computational cost vs the visual output. What it isn't is a personalised recommendation for every GPU.
In the case of your 9070 XT, what you will get will be much higher performance vs maxed. Now, it's up to you to to decide how to adjust from there. Do you want higher resolution? Faster performance? A lower upscaling factor? Even when making these decisions, the optimised setting procedure - breaking down each setting and analysing its cost - will better inform you on how you may wish to change your settings.
Show Comments
Leave A Comment
Hold on there, you need to login to post a comment...