Shadow settings and performance

[Billcimus]

Yes, I hardly have the optimal PC to run this, being an ancient i5 3570k and 980 ti.

I still think it's odd that changing only the shadow setting from medium down to low doubles my performance in certain areas, say in the caves with the light rays. In general it seems to give me around 15fps.

I can't work out why there is such a dramatic disparity in performance when making this one step-down change. It doesn't even seem to make all that much visual difference, with the exception being it does seem to render the light rays much more simply-- which is odd given this is supposed to be a shadow setting.

I wonder if this setting stealthily alters other settings.

[Ordisoftware]

Shadows and textures are the critical point, not only because of the memory, but also because of the calculations to be performed.

Some modern techniques have become so as well.

You need to upgrade your GPU as well as your CPU for such a game.

https://store.steampowered.com/news/app/2561580/view/4547039255696769239

[normanpcn]

Shadows are normally an expensive operation. Shadow quality setting typically change the render resolution for shadows and may also add/remove items which get shadow maps rendered.
You may think the game renders your frame once. Yes there is the final render but for things like shadows they are a completely different render pass to create a shadow map which can be applied to the final render. These specialty renders may be at lower resolution for performance and they do not need things like textures and lighting and such. Each unique shadow casting light source needs its own shadow map render, from its point of view, to generate the shadow data. Outdoors the sun/moon is an obvious one but there can be point lights, like torches/lamps, that may be shadow casting lights. This is one way where in certain scenes the shadow computation can get very heavy.

[Billcimus]

Originally posted by Ordisoftware:

Shadows and textures are the critical point, not only because of the memory, but also the calculations that have to be performed. Some modern techniques have become so too.

It does seem there's something a little odd going on. I'm running 1080p with FSR balanced mode.

I wasn't expecting it to run well given my rig, but surely going from medium to low with just one setting shouldn't literally double FPS in some areas, and add 50% in general.

What's even more odd is if I disable shadows entirely it runs exactly the same as on low.

Perhaps there's some feature or effect that is enabled only at medium, that my caveman rig just doesn't have the tech to run. I have noticed that I'm oddly CPU bottlenecked-- it's running at 100% full whack on medium whereas my GPU is at something like 70% or less.

As I mentioned before, the look of lightshafts when going from medium to low shadows changes quite dramatically.

[Ordisoftware]

Indeed, the GTX 980ti is below the minimum recommended configuration which is 1650 or RX 5500XT, as i3-8100 or Ryzen 3 1300X. It's time to upgrade if you can. For my part I am very satisfied with the Ryzen 5800X (2021) and the RX6600 (2023) that I got at a good price a few years ago already, for example. Choose the motherboard and a NVMe carefully.

[Billcimus]

Originally posted by Ordisoftware:

Indeed, the GTX 980ti is below the minimum recommended configuration which is 1650 or RX 5500XT, as i3-8100 or Ryzen 3 1300X. It's time to upgrade if you can. For my part I am very satisfied with the Ryzen 5800X (2021) and the RX6600 (2023) that I got at a good price a few years ago already, for example. Choose the motherboard and a NVMe carefully.

Actually, my GPU has some petrol left in the tank. It's my CPU that running flat out at 100%

[Ordisoftware]

You can't play top AAA games like the Horizon saga in its current state anymore.

[GrimmsChoice]

One thing that might affect the effect of the shadow setting, is the other lighting settings like SSAO, anything that affect and/or alters shadows would probably be affected by the shadow setting.

[𝔗𝔥𝔦𝔫𝔤𝔲𝔰 𝔄𝔠𝔱𝔲𝔞𝔩𝔦𝔰]

3570k is dead at this point if you want to play fresh game. I'd say you need at least a modern 4c/8t cpu i.e. 12100f or faster. 3570k is a bottleneck even for 980ti

[Billcimus]

Honestly there's little point in recommending upgrades. It's pretty obvious isn't it, come on. I am fully aware that my PC is very outdated-- and I was never expecting to run this game at even 60fps 1080p.

I was expecting to get something like 30fps 1080p, on medium settings.

To me, it is a bit odd that even shadows - which yes are generally CPU intensive - would make such a colossal difference with one step difference. It is odd that shadows completely disabled runs no better than shadows on low, and that going up to just medium literally halves my frames in underground tunnels and reduces it by about 15 everywhere else.

When you tweak settings on hundreds of games, you start to notice aberrations when certain titles don't conform.

In the end though, it doesn't actually looks all that bad with everything else on a combination of medium to high, and just shadows on low.

[Ordisoftware]

In a way you can't get anything on modern big games with this configuration that is more than ten years old.

There is nothing to do except upgrade considering the current specifications of minimum, recommended and advised as for the Horizon saga, because this hardware is obsolete and if it is not yet the case, it will soon not even be able to launch the games that will crash or say "incompatible hardware" if well coded.

The next games are going to be worse because with raytracing and various things like AI and new mathematical developments to process pixels. We are at a new remarkable tipping point of 2020-2025 for video games. This is one of the fundamental reasons for this remaster to align with West and have two homogeneous episodes.

A lifespan of 10 years for this computer equipment remains exceptional compared to 5 or 2 years for gamers.

[Billcimus]

Originally posted by Ordisoftware:

In a way you can't get anything on modern big games with this configuration that is more than ten years old. There is nothing to do except upgrade considering the current specifications of minimum, recommended and advised as for the Horizon saga, because this hardware is obsolete and if it is not yet the case, it will soon not even be able to launch the games that will crash or say "incompatible hardware" if well coded. The next games are going to be worse because with raytracing and various things like AI and new mathematical developments. We are at a new remarkable technological tipping point of 2020-2025 for video games. A lifespan of 10 years for this computer equipment remains exceptional compared to 5 or 2 years for gamers.

Thing is I can run this game on average well above 40fps, with some settings even set to high. I'm intentionally focusing on just the shadow setting and the difference in performance between very low and low - I made a mistake earlier in saying low to medium, I meant very low to low.

Really can we forget everything else, as it's not relevant, and just talk about this shadow setting and specifically the difference in performance between very low and low. I've tried to research this by searching for it, but couldn't find anyone else in videos and such who have brought up this. Given this I am curious to know if there's some kind of issue with 900 series cards or my general setup-- and is anyone else observing this performance difference between very low and low shadow setting.

Oh I know that once the next generation appears I won't be able to run new games at all. As I said, for new Ps5 games, I don't expect much, and I know I can only expect 30 maybe 40fps at 1080p and medium settings.

Honestly I just can't afford decent gaming PCs these days. I'm thinking of getting a Playstation, and I've already bought an Xbox Series S.

I only got this as it was a £10 upgrade.

May some else try setting shadows to very low? and see if the performance jumps up as much as mine does. Comparing very low to low-- particularly in the intro area in the underground cave with the technology.

[Ordisoftware]

Nowadays in this kind of game the difference between no shadows and the first lowest level of shadows is almost zero in terms of performance, this is normal. It is the same between an old antialias technique and no antialias. Just like you will probably see almost no major difference between the first two levels of textures.

But on your hardware, especially shadows as well as with antialias and tesselation more than textures, it is normal to have such a difference between each quality level, as I told you.

I read that your question does not concern the performance of your card and an upgrade, but answering you is telling you these things because your hardware is obsolete and is no more made to run current games as indicated in the configuration tables.

Think of it as a spring that is stretched more or less depending on the age of your equipment, or a staircase with more or less high steps, because of the speed of processing calculations by the CPU and the GPU, not to mention the transmission of data between these elements as well as with the memories, etc. Thus the more recent it is, the smaller the spacing, and the older it is, the wider it is.

Here are some links in case you are interested in this topic:

• Wikipedia's Shadow mapping
https://en.wikipedia.org/wiki/Shadow_mapping

• Using light and color in game development: a beginner’s guide
https://medium.com/my-games-company/using-light-and-color-in-game-development-a-beginners-guide-400edf4a7ae0

• Lighting theory for 3D games, part 5: the rise and fall of the cult of hard shadows
https://www.blog.radiator.debacle.us/2017/03/lighting-theory-for-3d-games-part-5.html

• Fundamentals of Computer Graphics
https://www.amazon.fr/dp/1032122862/