Thoughts On Ue 5 And Optimization From The Thames Valley Creator Interview

(original discussion)
https://www.rail-sim-community.com/...view-with-train-sim-society-the-optimization/
--
I post it not just because of the route(s) which looks kind of quaint and cute, but also for the brief discussion of optimization and UE 5.

The developer says that there is no real advantage to going to UE 5 since they have the "same limitations" and that "optimization" was up to each developer by taking "shortcuts."

This isn't the first developer that has said this, although folks keep tossing around UE 5 as some sort of "magic bullet" that DTG NEEDS for TSW to expand. From these comments it seems that it's not necessary and UE 4 is fine, it's just the actual details of the route that need to be ironed out better in the design phase with these "shortcuts"... not the UI being used.

Now here's my take as a project manager in the past (not in video games however.)

Bigger teams can create more, but communication becomes crucial. The more hands you have in the pot, the more chance for potential conflict simply because you aren't always watching what everyone else is doing on the team. You are too busy getting your own work done. That's just the nature of joint projects.

I wonder how much of the bigger projects being less optimized is because there's multiple developers working on it, while smaller routes have 1 or 2 developers who are each handling more and KNOW what they're optimizing.

If you're a single person doing everything, you know what "shortcuts" you took or how you programmed something. If you are part of a larger "team" then you have to really communicate a lot or things might get lost along the way. Then the QC later is ANOTHER set of eyes on the same project that don't know how it was made, just how it looks from the outside.

It works here with "smaller" projects because 1 or 2 people can easily work alone or side by side and work almost as one mind on it. However, that does limit the scope a bit to these smaller projects (and/or smaller scope like this route that's maybe 20 miles and just one locomotive, no layers, etc) That's not bad. It's just how much a person can reasonably handle.

When you expand projects you take on some risk with more people, more labor management and more crucial communication.


What is your take?

Video interview here

 

The size of development team can have pros and cons. It's all depending on the knowledge in the first place I guess. A single developer with a lot of knowlegde can make a good product, but has to do it all alone. With a team, you have most likely more knowlegde and you can help each other, however with big teams, you often see communication becomes a problem as everyone does just his/her own part.

UE5 does offer some new techniques, however this comes with a performance cost. Switching to UE5 will not solve the current issues and most likely you will get more issues in the end. They're now on a old UE4 version and there is no support for FSR and DLSS for example, I believe those techniques where not supported right out of the box on this UE4 version.

If you look at UE5 games, the overall experience is not great, bad performance and low resolution. Consoles are not happy with UE5 games, unless you not use techniques like Lumen and Nanite. Leaving out those techniques are even more reason not to switch to UE5.

 

In my opinion, TSW has poor graphics, especially when not played in environments with predominantly sunny weather.

If switching to UE5 can improve the situation, then it should be done.

If better work (?) by DTG on this current graphics engine can improve the situation, then it should be done.

Doing neither will leave TSW mediocre.

Before any misunderstandings arise (the usual ones), it's important to clarify that switching to UE5 means developing a new rail simulator from scratch.

 

What is your standard for "good?"
Why do you think UE 5 would be better?
I do think there's a newer version of UE 4 available with more options, not sure if that'd be difficult to port over or not.

 

A software developer gives a reasonable and well-informed opinion.

At the risk of stating the obvious, every extra mile of TSW route is an extra mile's worth of developer design, implementation, QA, all of which are measurable in units of time. Managing larger software development teams is a professional art for which huge companies pay obscene amounts of money to consultants to try and figure out, and they still manage to screw it up. It's hard, and "but this version or that version is better for the user" is almost never a decision without consequence, and the best make those choices thoughtfully and with as much information as they can reasonably gather.

 

I just explained it above.
I challenge anyone to say that TSW, with, say, medium cloud cover, can be considered a graphically acceptable product in 2026.
Are Lumen and Nanite good or bad?
During MR's development, DTG raved about these features.
Now they're obsolete?
There are also has other flaws that completely ruin the immersion, not directly related to the graphics engine.
Like flying cars...
But that's another story.

 

You didn't compare TSW to anything above. You just said it's vaguely "poor."

That's why I asked what you were comparing it to.

 

Everyone has their own opinions.
But when I hear that a graphics engine like UE5.8 is no different from UE4 (which is over 10 years old and no longer supported by Epic), I feel a little disappointed...

 

there is and that unreal 4.27 and it support dlss 3.0 and i think fsr plugin. matt said something like it take few months to upgrade from 4.26 with nothing else being worked on. i think he said there little benefits. i tried to fine the post matt posted but it gone.

 

Reading this as an old duffer and someone who has merely been a spectator of the growth of gaming since the early 80s.
Is there a bigger problem here? Have we reached the outer limits of games development or at least games that can be played on consoles and reasonably priced gaming PCs?
Is the next step-up going to require major developments in hardware and a possibly very expensive jump in the cost of games machines for the end user?
Is the next stage going to need super-fast broadband as a game's world environment is streamed?

Is the dream of an simulated world looking 100% like real life never going to be possible?

 

The dream of a world simulated 100% like reality will remain an illusion, but a simulated world crafted with a little more care would be desirable...

 

Should just do similar to MSFS. Stream the environment and overlay with handcrafted assets. Use the freed processing power on the train and signalling simulation.

If they could get access to AIVR and routeview data in the UK like I use at work for signalling design you'd have the latest cab view videos and overhead photos of routes to use.

 

I think most people did not really understand what Johannes said about UE 5.

First, you need to keep in mind what the questions was. He was asked what was the biggest technical challenge.

He said UE5 makes some things easier, but it is not a magic button that fixes it all. Yes, UE5 removes the need for rebasing, handles loading the scenery better and nanite pretty much removes the sensation that an object is being replaced by another as it gets closer. Those are all improvements.

BUT it does not remove memory requirements, doesn't remove the need for the developer to still think about how many things can be placed in the world, how the frame rate will be for most users, etc ... UE5 runs great in good hardware, but it does not automatically remove all (or most of) the work related to managing performance from the developer. This was the question that was asked to him and what he was answering to.

Keep in mind that all these improvements UE5 brings have hardware requirements, and if your hardware does not meet them your performance is going to tank hard.

Also, as others have mentioned, you need to remake all the content for UE5 to make use of the new features. They can't just load the UE4 project in UE5, press "compile" and ship it. Not even close.

 

Good point Inkar.
Windows 11 has significantly higher hardware requirements than Windows 10 so older machines struggle with the newer system. It's approximately twice the memory requirement of the older program.
Plus there are software minimums you must have too. (Direct X12 is required vs DirectX 9 I believe)
I don't know what UE5 requires, but there is SOME increase that would make older systems and consoles obsolete even quicker.
That's before the added work requirement to rebuild the whole game in the new system.
I still haven't been sold on just how much advantage would be gained for all that cost.

 

Something's not quite right.
Why did DTG attempt to develop MR with UE5, with such a grand presentation (a year ago)?
Also, I don't think all games developed with UE5 (if well optimized) crash consoles.
Just think of the optimization work done for FH6 (which doesn't use UE5 anyway) on Xbox.
A perfect job done by those who know what they're doing.

 

It's a different game.
They can try it on whatever system they want.
They stopped developing Metro Rivals, so does that tell you it was a good or bad experiment?

 

I believe MR was abolished not because of the power required by UE5, but because of the absurdity of the project itself.
Then again, if the problem with TSW is that it can't handle a certain amount of add-ons, why continue to "pump out" DLCs like crazy?
Isn't 10 years of addons a bit too much?

 

I think it is not a problem of what is required to use UE5, but what system would be required to make it perform at acceptable levels.

To make an example with current technology, TSW6 only requires 8 Gb of RAM and 2 Gb of VRAM to run it but I guess we all know your experience with that is going to be terrible.

Same happens with UE5 and consoles. A TSW Next with UE5 on a console would run, but not as you would want. Probably to a point where you would prefer to stay playing the older UE4 version.

 

This is my concern. With MS extending the Win 10 lifecycle by another year and the current price of hardware, I'm hanging on to my current PC for at least another 18 months, though I am about to (finally) get a slightly better video card than my aging GTX 1650 and, if I can figure out how to do it, trying to swap in a SSD for my C Drive. With the current spike in hardware prices it's hard to justify spending a substantial four figure sum just to play a few games. If the hardware prices stabilise then there's always the option later of doing a motherboard and CPU etc. update.

 

He did, and I've linked to that thread a couple of times in the past when DLSS came up, but it's giving an error now. Must have been in the Off-Topic subforum.

CMs, can the thread linked below be moved into TSW General Discussion, please? It fits, and there's valuable info inside, especially a DLSS discussion between Matt and JetWash.

https://forums.dovetailgames.com/threads/train-simulator-unreal-engine-5.76804/

 

Adding to this part, I think the what was said was that nothing added to 4.27 had a real bearing on TSW, so it wasn't worth going to the trouble to upgrade.

It's also worth tossing out a general reminder that TSW is using a very heavily modified version of UE4 that also includes useful features backported from UE5, so the full upgrade would not be a straight UE4 --> UE5 conversion. I imagine it would involve not only a complete rewrite of TSW, but also a substantial rewrite of UE5 to fit.

 

Is ten years of add-ons enough —— no! Still good add-ons coming out for TSC. I think I have most all of the DLC for TSW and I don’t have any problems with blurriness, slow loadings or crashes. Of course I gave a pretty hefty PC since I use it for flight simulation as well. That aside, I don’t see were the conversion to UE 5 would be of any benefit for many of the arguments made above. It seems the most common argument to move to 5 is “it is newer so it must be better. Not a philosophy that I hold.

 

People need to give up this weird dream of UE5, for now. It wont be what they are thinking it will be.
I'm not trying to lose all of this content that has been built up, I don't want to wait 10 more years for my route to be picked and then built in a new engine again. No thanks, the trade-offs are not worth it, not even close. It's not as if some UE5 version wouldn't also have problems.

 

Maybe because they realized full MCP integration would be implemented in UE6 in the 2027 beta release?

 

UE5 is a better engine. There is no doubt about it, and if you read this thread specific technologies like nanite and lumen were mentioned.
Reducing the arguments to make a UE5 version of TSW to "its newer so it must be better" is not fair.

I think it is the other way around. It is you who do not understand how the release of a UE5 version of TSW would be.

First is that you would not be giving up on anything the current UE4 version has. If a UE5 version of TSW was released it would be as a new product. It would probably start slow at first, releasing few DLCs for some time while the UE4 version keeps releasing DLCs at the normal pace. Remember when TSW was released just with Heavy Haul while TSC kept releasing DLCs? And guess what ... there are new TSC DLCs still being released today. Forget the idea of TSW UE4 or TSW UE5. It will be TSW UE4 and TSW UE5.

The UE5 version of TSW would be technically better, look better and have less stuttering. People with the appropriate hardware for it would buy it but a considerable part of the user base would still stay only with the UE4 version for some time (I'm thinking years). Adoption of the UE5 version would not happen immediately. It takes time as it always does, and people would jump over time from TSW EU4 to the new UE5 version as the hardware prices allow. For quite some time, many people would buy both TSW UE5 and UE4 DLCs, as it happened with TSW and TSC.

We are at a point now where if you just look at TSW graphics you can tell they are old compared to what the newest games are showing. Not terribly old, but noticeably old. That is why it is time to start thinking about a version of TSW with UE5. It will not be a fast journey, and it needs time to be done right.

 

here the thing maybe it not bad idea to start new tsw on unreal 5 soon. here me out. every time dtg release yearly tsw they add feature that are only current to that verson of tsw. for example station announcements....everyone was requesting this feature for years and once that feature was announced it be feature in tsw alot of people were happy but once news broke that older routes couldn't support it with out full rework of route alot of people were disappointed and to this day alot of people are asking when will station announcement be on this route and dtg doesn't have answer yet. same with suspension feature. heck they got license from Boeing and used it what once or 2 times. one other feature people are requesting is AI train horns. that would be cool and i think it be in future tsw game but it will be to that verson of game and might not be support on older routes with out rework the route.
if dtg start new game they can add these features in base game and it be adding going forward with other features TOD would current and they wouldn't have to go back to route and add it.
metro rival pretty much had i think core basic like sim graph and support for high speeds trains and stuff plus pcg in place before it got cancelled.

right now dtg is fixing blurry issue on consoles. witch i donno how long it be before it fully fixed maybe 2 or 3 years. by then the new xbox and ps consoles will be out and i think the blurry issue would go away on those consoles.

 

UE6, PS6, next-gen Xbox, RTX 6090...
Let's wait until 2035 and everything will be ready, unless someone starts throwing bombs at our heads...

 

It's unclear what DTG is doing (honestly I don't really care anymore) as a developer (UE5? UE6? UE7?, nothing...).
It's just that we've reached the point where third-party developers are eyeing projects from individual modders who are deemed worthy of further development for a future release (see the "GAS" tanks).
A subcontract that gets subcontracted...
They are getting closer and closer to the Trainz philosophy...

 

I know very little on this subject really, however it seems the core game is not solid enough for it to be farmed out to 3P developers to create additional content. Is every developer creating their own shortcuts or workarounds to make their DLCs work at the same time DTG are working on core game optimisation? This seems silly to me unless like operator#7940 says, there is excellent communication and collaboration across DTG and all 3Ps. But I'm not sure there is.

 

Its an interesting thought. I was playing TSC at the time TSW was first release, and I entirely understand what you're saying. I always thought of TSC -> TSW being entirely different to UE4 -> UE5, and UE5 just instantly becoming a complete replacement. Simply because TSC and TSW are such different games, it's not just TSW twice with different engine.

But actually, what you say is likely correct. Even ignoring hardware limitations with UE5, I honestly don't know enough about the subject. If a new game was to release, 3rd parties probably wouldn't even have the ability to create content of it (If it was to be anything like current TSW) and so they would only continue as normal create content for this game as usual - nothing would change with this game.
I just never thought about it in this way.

 

I'm sorry, better hardware does not equal a fix, certainly not when hardware pricing has been so screwed since COVID. Heck. somehow the current situation with pricing actually makes me beg to go back to 2020 pricing, at least RAM sticks were still dirt cheap then. Times like this make me glad I choose to over spec my PC now, I probably wouldn't have been able to get a second SSD or more RAM if I didn't.

Realistically any new consoles that come out are gonna be easily some of the most expensive in history. And they're going to be out of mass adoption range for quite a while, heck, I just saw this article talking about how even the former Sony CEO is scared of the pricing.

https://www.gamesradar.com/platform...w-much-more-ray-tracing-can-you-put-in-there/

If a guy with a more than likely pretty healthy bank account with personal bias to the company is wonder what the hell Sony can offer to make it worth $1000 in his eyes. Imagine how many general consumers are just going to nope out of this next generation of systems?

By extension anyone insane enough to try and make games functionally needing a PS6 to be enjoyable might as well just beg for bankruptcy. DTG's gamble of hoping next-gen machines bail them out of their tech dept is exceptionally stupid when I think it's at least gonna be 5 years before those outside the enthusiast can start thinking about upgrading. Even if DTG do a new train sim that sim still likely needs to run on current gen machines if they want any healthy sales, and currently i have no confidence a UE5 TS would run at a solid 30FPS with no blur issues.

 

Indeed. The RTX3050 I just ordered was £170 yet this time last year were selling for around £150 and by rights a year on should have dropped to around the £100 mark.

 

Conceptually it was absolute nonsense. Even just on a matter of foundational gameplay mechanics.

 

The crucial thing there being that it's a proprietary in-house engine the respective FM and FH teams have been iterating on for decades. Not an off the shelf "jack of all trades, master of none" engine like UE.

 

I think this whole argument over "engine" misses the bigger point of hardware.
UE5 MIGHT be able to run higher speed stuff. (Note we're not comparing to UE 4 to UE 5, we're comparing the UE4.X that TSW is running vs base UE5)
The point is that more "stuff" means better hardware in your machine.
There's no "something for nothing."
The bottleneck is not the game engine.
It's the hardware requirements.
You can already create a "fire hose" of content in UE 4 that is way beyond what most peoples computers/consoles can handle.
In railroading terms, it's a bit like arguing over the benefits of a "high speed train" to go 100 or 125 mph... when most tracks on the line are only rated for 75mph. It's an irrelevant argument because most people don't have the computers/consoles to even use the current capability of either UE 4 or UE 5. To be honest, we're running into the situation where many "branch lines" (customers) are still only maintaining 55mph rails and the latest TSW version is trying to run at 75mph for the "majority" of the mainline.
Hence the buffering/stuttering.

If you think of yourself as the track owner, TSW as the "train" and DTG as the train operator, then you can see the dilemma. Yes some local industry and branch lines are private and don't want to upgrade, but want the higher speed trains on their lines. Meanwhile some people are begging for even HIGHER speed, not asking what the "cool new train" would actual run on.

 

People keep treating UE5 as some kind of holy solution, but that’s simply not true. The Bus originally ran on UE4 and later switched to UE5, and in terms of performance you could barely notice any difference. What was clearly noticeable is that it looked much better thanks to Lumen and other graphical improvements UE5 brings.

Is it a real game changer in terms of FPS? No!, it isn’t. As a game developer you still need to make optimisations yourself, and the same limitations on what you can build still apply.

 

No need to get hysterical, the point is the productivity increases with agentic AI would be phenomenal. So, yes, waiting for UE6 is the right move. Had they built a new foundation on UE5.8 (with essentially alpha MCP builds) - you would not be able to reap the benefits of having Fable spin up 10 agents to do all your work. UE5 still relies on blueprints, UE6 is moving away from that model. LLM's cannot work with blueprints, UE6 won't have that problem. A new game on UE6, will allow them to essentially run agentic AI and if they do it right,(which history and prior track records should make this a HUGE if) - it could be spectacular. I personally, don't have much faith.

https://www.unrealengine.com/news/the-road-to-ue-6

 

Yes, anything is possible.
In the meantime, keep giving them your money for the current TSW so they can fund something innovative.
I'll wait for a simulator with UE6, if I'm still alive by then...

 

You may not have to wait much longer

 

Let's hope so...

 

No thanks on the AI slop

 

I kind of think that if they do want a rail sim in UE 6, it should be a fresh slate on a new game from the ground up using the full capability of the new program. TSW can be maintained and added to (but stabilized over time at the core) like TSC was. Even today TSC has a core that is kept stable and updated and there is lots of third party content being released.
TSW would just become something like TSW Classic or TSW Gold or something with no new update and new features every year.
It would create a nice "tier" of capability for different players for different level systems.
None of them would be "going away" but they'd be fully playable, and players can choose what level they want to game at...much like many players on older computers choose to play the less resource intensive TSC and their large existing library.

 

Is it really slop? I'm assuming from your comment, that you must have months if not years of experience using an Frontier LLM on an api or enterprise account with full access to all of the developer tools available? You then must have experienced the massive productivity gains from being able to write an overview and instructions markdown, spin up 10 to 20 fable agents and have it code your entire complex physics solvers and economy logic ( a process that would take months if done by hand) and have it done within two hours? Certainly, you've downloaded the June 5.8 UE build that ships with MCP plugins and you've connected your Claude Max 20 account to it and gave it a command to auto generate an detailed urban city, and watched with awe as it accomplished that in less than an hour? Or are you just making a blanket statement based on the manufactured outrage about AI propagating on social media?

 

Blanket statement. Screw AI

 

Regardless of the wholesale utility or cost effectiveness of AI as a concept, I think that like past technologies many in business will embrace it for the promise. Whether that ends well or not remains to be seen in each individual instance.
We may be having an entirely different discussion in 10 years if AI starts doing the actual coding. I haven't brought it up so far because that changes the whole current paradigm and on an unknown timeline.
Currently it's individual creators doing the job "by hand" (I get it.. on computers with a UI... point being it's much more direct implementation)
If we do end up going to just telling AI what we want to see and they do everything behind the scenes to the point there is no one looking at the actual code, then we're talking about a whole new animal.
I'm not sure how far down that road we want to look because in the mid-term future I can see people making their own content entirely through AI. For example if you tell your AI agent "Build me a realistic train simulation game" it would just do it.
There'd be no large scale distribution needed either because everyone would have their own, tailored to exactly what they want.
If you like what your AI makes your neighbor just says "make me the same thing...with X,Y,Z added"
But no one would be actually making or selling it.
This will be down the road, but that's the trend we're seeing.
You can already with a few prompts have short videos of something like "Batman fighting Superman at a mall in California" and the AI makes it up with you defining it with more prompts.
If we go too far into the "AI" track then eventually we reach a point where it's too soon to predict how different the world will soon be.

Now is that "good" or "bad?"
I think there's some of both and people will diverge more and more along those opinions as time goes on.

 

I appreciate your opinion, but in my experience (as someone who uses Claude Code everyday in my work) the people that scream loudest against AI, often have the most basic understanding of it. Your post points out that dichotomy. AI, including MCP agentic integration in game engines is here, and its being used and I guarantee you in the last year you've played a game or interfaced with an application that was coded by Claude or Codex. You can look at the recent Netflix app release that contained claude.md files in it. I'll just address your point of "telling my ai agent to just build me a realistic train sim". Ironically that's exactly what I did. And here's what Fable told me: I can help you with about 80 percent of the tasks you would need to do in UE5.8, but the rest of the visual and artistic design decisions, such as creating models, using Blender, and blueprint work (especially in UE5) would need to be done by hands on UE dev with indepth systems knowledge. What Fable DID do for me however, is spent approximately 20 5 hour sessions over the last month writing a detailed physics solver and schema's for the underlying game logic that lays out all the fundamental architecture for a train sim that can then be implemented by competent UE and Blender devs/artists. How many would I need? I don't know - but I sit on a considerable amount of investment resources which could be reallocated to hiring 2 devs for a year, at market rate to do all the work that is NOT possible through AI integration in UE. So, i guess the big question, if you want to continue this discussion is - what background or knowledge do you have, or what experience do you have running a Frontier LLM, managing multiple AI agents running complex coding tasks? Because if you have none, you just have opinions - and there's lots of those out there, which as some of the previous posts in this thread reveal - are based upon emotional responses, rather than experienced background in the systems being discussed.

 

I think this is probably the third thread I’ve seen you throw ‘to participate in a discussion with me you need XYZ experience’. That’s not how public forums work. I’m sure many people here could attempt to exclude you from conversations regarding trains & rail infrastructure on the same basis.

In any case, I seriously doubt AI isn’t being used in TSW development already, we’ve already had several art assets leaked into the release build which were later replaced once reported, so it’s something they’ve certainly (at the very least) trialed.

As for UE5, I don’t think it would achieve anything other than degradation. The problem is the development & publishing philosophy DTG has. As a console user, I haven’t really enjoyed anything new such as rail singing, because every route that’s featured it has been a crashing, blurring, buggy mess.

It’s not like moving TSW to a PC only environment is going to do much either, the money consoles brought in was clearly attractive enough to bring in several third parties, some of which were adamant about TSW being a DOA product or that TSC was where the money is.

 

Twobuss, while I respect you may have some experience (going on your word for that), I am speaking of future technology which you yourself pointed to above is not here yet. So saying that it's limited to just doing 80% of the work TODAY is not negating what it can do in 10-20 years as I said.
Where was AI 20 years ago?
There was none.
Where was it 10 years ago?
5 years ago?

Saying I'm wrong in 20 years because I'm not working in a program from today that won't even be used in 20 years is missing the point.

I think you're even missing the original focus of the thread, which was that at scale if you are working with other people there is a limit on what is possible without a lot more coordination. Now AI on that scale won't really change much because any time you work with people it's not about the tool so much as the interpersonal communication. Hence the root of my comment on AI...which was that one person will have a unique vision and introducing more people will diversify that view. The more people...the more diluted that single vision.

Now using AI, sure the developer on Thames Valley COULD have done it in a FRACTION of the time.
Yes.
That's true.
And 10 guys could in theory do it faster.
Just as possible now.
However, any time you have 10 people working on a project (be it manually with 10 hand tools, on 10 computers, or through 10 AI agents) you aren't going to get that 10 times as efficient without a MASSIVE amount of communication, and in total efficiency, it MIGHT be easier for smaller or single developers to do some projects in the long run because they have a better focus with less chance of miscommunication.

Because human beings are human beings.

 

Maybe but you should see how Denshattack is doing.

The concept seems pretty popular and could be a contender for game of the year.

 

There's no way this could be viable. DTG is already maintaining seven builds for TSW. Even if, for the sake of argument, we eliminate the two gen 8 builds that seem to be on their way out, that leaves five, which you're proposing doubling to ten, but still only consisting of duplicate versions of each of the five system builds based on two different versions of the same engine. And the customer base is going to fund this by cheerfully buying both UE4 and UE5 versions of each DLC during the proposed years-long transition? Not on your life.

TSC and TSW worked in parallel because they are two completely different games. Also because TSW is additionally available on consoles and TSC is not.