Engine upgrade and DX12 coming to ETS2 and ATS!

[mmorselli]

It's quite hard to generate meaningful benchmark results in an open world game that doesn't have a dedicated benchmark function baked in.
Maybe it could work if you deactivate traffic, then define a specific starting point (e.g. city border of LA) and route at a certain time and weather (at night with rain).

OK, but an exact comparison may not be needed. In VR, using the fpsVR software, you always have CPU Frame Time and GPU Frame Time data as an overlay. It would be enough to see if in a fairly long journey the CPU Frame Time exceeds too often an arbitrary value (e.g., 11ms corresponding to 90 fps )

No GPU will ever be able to surpass the CPU Frame Time, but a sufficiently powerful GPU can guarantee to match it

[harishw8r]

I speculate more of Texas would come with the Oklahoma update. That should also give them enough time to squeeze in the Peterbilt plant.

I’m more curious about the game features and technical upgrades. Apart from the fact that they are working on it, everything is in the dark. Pretty confident something big is in store.

[Some newbie driver]

Do you want data on gaming rigs? Go check any Gamer's Nexus reviews and comparisons in they blog or YT channels; pretty serious stuff that put on shame some "serious" journalists. Also they use to include at least three gens of CPUs when doing reviews of new ones; so you could see that in gaming os single-threaded software (not synthetic benchmarks) there's not so much difference as you think.

And stop comparing OTHER games. Games can't be compared ones to others; every game has it's constrains and caveats to deal with. At most, you can compare games that share a base equal platform/engine and not even then you will narrow down enough all the variables to do a comparison that allow you to extrapolate the results to other games. Not to mention that almost no game you could check out there in almost any review of modern CPUs is going to have the same constrains that SCS games has (because those games use not to be popular enough to find room on reviews of top tier hardware).

So, stop going around the same thing. You are NOT going to gain a performance that will worth the expense needed for a 13000K computer. You are not going to improve linearly the performance on this game regarding your current CPU at the same ratio that any simplified % figure you could find in any review out there. And that's not going to change until SCS could not be able to absolutely change the current architecture of their engine because that's where it lies the bottleneck that makes it unable to scale the leverage of better hardware. POINT.

Understand that I'm not saying the stupidity that will be "you will not improve". Of course you will; but if to improve a 30% you have to expense a 300% (hypothetical figures); there's an atrocious diminishing return there that makes the purchase absolutely worthless if aiming to play SCS games NOW. And those are not assumptions, my work is to know and understand those things; and I've played this games in several different rigs all those years to have seen how it behaves in different scenarios and what they have improved and what they could not.

You don't believe me? Then waste your money and cry later. But if so, then cry somewhere else, please. Excessive salt in the room isn't good for trucks maintenance.

BTW: They had been working on a new render code, hinted here, for near 1 year (probably a lot more). IF, and only IF it has the kind of changes we hope; then it will make sense to pay big chunks of money to play those games at high performance. But probably for when that change will be released, even if at the end of this year (better not to have too hopes on it), the 13000K will be nothing more than an outdated and overpriced piece of silicon already. So, that discussion would be pointless anyway.

Regards

Edit: Typo

[sikejsudjek3]

Personally I hope that they go for vulkan as I'm still playing in dx9 to have compatibility with 3d nvidia vision (yes I know nvidia gave up supporting it - but there are so many fixes from the community it works better now than it ever did before when nvidia were supporting it!).
Vulkan has a 3d stereoscopic wrapper written for it, dx12 doesn't. I tried to get ets2 dx11 working with nvidia vision, but there were too many linked shaders to simply turn them on or off with the shader hacking tool. So unless that's changed I'm stuck with dx9 unless Vulkan comes out. Oh and by the way - with 3 stereoscopic screens it looks absolutely amazing!

[Appusle]

Do you want data on gaming rigs? Go check any Gamer's Nexus reviews and comparisons in they blog or YT channels; pretty serious stuff that put on shame some "serious" journalists. Also they use to include at least three gens of CPUs when doing reviews of new ones; so you could see that in gaming os single-threaded software (not synthetic benchmarks) there's not so much difference as you think.

And stop comparing OTHER games. Games can't be compared ones to others; every game has it's constrains and caveats to deal with. At most, you can compare games that share a base equal platform/engine and not even then you will narrow down enough all the variables to do a comparison that allow you to extrapolate the results to other games. Not to mention that almost no game you could check out there in almost any review of modern CPUs is going to have the same constrains that SCS games has (because those games use not to be popular enough to find room on reviews of top tier hardware).

So, stop going around the same thing. You are NOT going to gain a performance that will worth the expense needed for a 13000K computer. You are not going to improve linearly the performance on this game regarding your current CPU at the same ratio that any simplified % figure you could find in any review out there. And that's not going to change until SCS could not be able to absolutely change the current architecture of their engine because that's where it lies the bottleneck that makes it unable to scale the leverage of better hardware. POINT.

Understand that I'm not saying the stupidity that will be "you will not improve". Of course you will; but if to improve a 30% you have to expense a 300% (hypothetical figures); there's an atrocious diminishing return there that makes the purchase absolutely worthless if aiming to play SCS games NOW. And those are not assumptions, my work is to know and understand those things; and I've played this games in several different rigs all those years to have seen how it behaves in different scenarios and what they have improved and what they could not.

You don't believe me? Then waste your money and cry later. But if so, then cry somewhere else, please. Excessive salt in the room isn't good for trucks maintenance.

BTW: They had been working on a new render code, hinted here, for near 1 year (probably a lot more). IF, and only IF it has the kind of changes we hope; then it will make sense to pay big chunks of money to play those games at high performance. But probably for when that change will be released, even if at the end of this year (better not to have too hopes on it), the 13000K will be nothing more than an outdated and overpriced piece of silicon already. So, that discussion would be pointless anyway.

Regards

Edit: Typo

I do agree with this statement. Though not about the 13th series. While it might be overpriced at that point, it won't start to lose performance all of the sudden.. so it will run as fast as it does now, 5 or 10 years from now. Unless ETS2 suddenly gets a completely new engine, the performance remains the same. But.. I will list my specs below. I do however, agree that you should NOT upgrade to a new CPU solely for ETS2 reasons. the engine is single core only and you will not be getting that much more performance. Even with the rig listed below I needed to use the FSR OpenVR mod at 50% render scale to get it to run smooth. This effectively reduces the resolution by 50% then upscales it back up with a smarter way of upscaling so you don't really notice it.

My current rig consists of an i7 13700K, RTX 3080 10G and 32GB 3600 CL16 ram.
my VR HMD is a HP Reverb G2, this has a resolution of 2160 x 2160 per eye. (Although SteamVR some how doesn't recognize this for some reason and pins the 100% resolution above 3000.)
I run SteamVR at a custom resolution of 150% and a per app resolution of 300% (If I don't forget I will edit this post with exact resolution values given I forgot what they were since the seemed like arbitrary values given the 2160x2160 base resolution of the HMD) Then I use the OpenVR FSR mod at 50% render scale to further smoothen it out. I just can't figure out fi they use FSR 1.0 or FSR 2.1... I tried to run his newer project that does the same VR Performance Toolkit but that one doesn't seem to work with ETS2 at all.

When I was optimizing my VR experience I noticed that my GPU, regardless of resolution in VR my GPU never exceeded 70 to 80ish% only when it was at 500% resolution scale did it go to maybe 94 in some cases but the FPS and Frame delivery suffered greatly. I am hoping that when they update the engine to support multithreaded workloads it will smoothen things out. When I'm just driving on a highway or locations where nothing is really going on it runs fine, the graph is mostly green and usually hovering around 5 tot 8MS with few orange or red lines. Orange means a frame was delivered late, Red means a reprojection was needed or the frame was missing, this leads to stuttering. In VR what that results to is either a feeling of your game suddenly being fast forwarded messing up your turns for instance or you feel something is off, the motion isn't smooth and for some this can be sickening even. (I got used to it really fast though, and I was the kid that got motion sick in vehicles)

While I know, as a VR user I am in the minority, and it's a godsend it works as well as it does given one developer of the team works on it,(Unless that is changed now..) but I do hope.. with DX12/Vulkan and an engine that uses the hardware at its disposal it will smoothen out. I'm also not asking for the game to run at double the frame rate it does now, the game is from 2012, and the engine is also fairly old. there is only so much you can alter or add.

I would also hope that at some point they are able to enhance the games fidelity a bit more too, in VR having sprite based foliage is very odd, when you move your head from left to right, so do the sprites, so you notice that they are just flat textures. it's a very minor thing and sometimes fun to see. Another thing I have is that I hope they can add FSR 2.1 to the game to further help with performance on older hardware. Given FSR is platform agnostic it will also work on every machine the game is released on. This, hopefully, also addresses the lack of AA in VR. the Jagged edges are not just distracting but after a bit of play makes you feel a weird pressure on your eyes. They also make it hard to see objects in the distance, or park using your mirrors because the outline used is barely visible in your right mirror. So hopefully.. this will also be addressed at some point. Though I'm not a VR Dev and I have no idea how hard it is to make AA work in VR. or how that would even pertain to their engine.

Though.. Currently.. The frame delivery and lack of AA are my 2 biggest gripes with the game.
and the frame delivery might be solved, or mostly solved with the move to DX12 or Vulkan.

I went on a bit of a ramble here.. mentioning what came to mind as I went too so I hope it isn't too all over the place, thanks for reading though if you did, you are a trooper!

EDIT 13-FEB-2023:

My SteamVR Settings:

Advanced Settings: Show
Render Resolution: Custom
Resolution Per Eye: 150%
Advanced Supersample Filter: On
Overlay Render Quality: High

Pre-Aplication Video Settings (ETS2):
Resolution Per Eye: 300%
Resolution: 6692 x 6548

Currently testing with 500%
Resolution: 8640 x 8456.

HMD Resolution Per Eye:
Resolution: 2160 x 2160

Unsure if the above values even apply.. I see no performance difference at all whether I go higher or lower.
I am using SteamVR to handle the OpenVR/XR API rather than Windows Mixed Reality.

[Tomsun]

@Appusle How is your experience in other more modern (e.g. multithreaded) titles, is there no problem with frame delivery?
i assume that neither the "simple" switch to DX12/Vulkan nor the "simple" switch to a multithreaded Gameengine including restructuring the game will bring the "good" result, but only the combination of both. But if i see the current situation in my personal rig (5800x3D, 1080ti, 32 GB CL18, all NVME distributed drives in 1440p-UW 100 FPS) even in situation i have Framedrops in average only 20% of the situations i would see the GPU bottleneck, in 80% of the cases its complex situations, heavy traffic, Game save, where the single thread is at its limits, GPU Load as a rule of thumb somewhere between 40% (old regions) and 60% (new regions), CPU Load somewhere between 8% and 13% (where 12,5% means 1 core FULLY loaded, as its 8 core CPU, for other core counts respectively)

[Berniyh]

Are there actually any comparable titles available with DX12/Vulkan and multithreading support as well as well optimized VR?
Off the top of my head, I can't think of any. At least not in the sim world.
Maybe Half-Life Alyx? But I haven't played that one, so can't say for sure. It does have a Vulkan port at least.

In the future there'll be more of course, e.g. DCS will most likely drop multithreading and Vulkan support this year. XP12 is already using Vulkan, but still single-threaded, it'll move to multithreading at some point (possibly not this year).
Also a new Assetto Corsa will come out this or next year and I suspect they might be using DX12/Vulkan as well with their newly-built engine (but we don't know yet).
Especially for DCS and XP12, I'd expect significant improvements in VR support when they go Vulkan+multithreading. Right now both of them are ok, but seriously limited in VR by the single graphics thread. Just like it's the case with ATS/ETS2.
Not that you can't do VR with singlethreaded graphics (iRacing and the original AC do well here), but both of these are limited in the overall graphics quality that you can achieve. And Assetto Corsa specifically has a serious, basic limitation in its graphics engine meaning that it can only use a single light source. (Everything else is shaders or post-processing)

[angrybirdseller]

I heard for years about the virtues of multi-threaded games, but I think even in 2023 it comes single core thread efficiency for game performance at 1080p or lower resolution. Even if DX12 arrives in next couple months think it will take years to make ATS and ETS2 multi threaded to see performance improvements. There are some here understand that scs has to rewrite the code all to make workable for DX12 and multi thread performance.

[Berniyh]

I heard for years about the virtues of multi-threaded games, but I think even in 2023 it comes single core thread efficiency for game performance at 1080p or lower resolution. Even if DX12 arrives in next couple months think it will take years to make ATS and ETS2 multi threaded to see performance improvements. There are some here understand that scs has to rewrite the code all to make workable for DX12 and multi thread performance.

Well yes, but I hope that progress is already ongoing for years as well. When they introduced DX11, they must have had multithreading on the roadmap already.
Which does not mean they started with that right away, but I'd also expect that they didn't just start on that now.

See DCS: they will provide a first step towards multithreading this year, splitting the graphics thread from some other stuff (most likely Physics calculations).
Then the next step will likely be Vulkan and then – possibly much later – there will be multithreaded graphics.
While the last step will surely improve the performance further, even the first two steps should already improve the performance, especially on VR.

[flight50]

Isn't SCS already using multi thread though? Its multi core we don't have. I'd rather have multi core support so that it can take advantage of multi threading.