RRT3 crashing

Tips & suggestions for a good RT3 playing environment.
User avatar
undertoad
Watchman
Posts: 85
Joined: Sun Oct 04, 2009 8:21 pm
Location: Glasgow

RRT3 crashing Unread post

Hi

I hope I'm in the right forum here for "RT3 is behaving strangely" problems.

My current RRT3 game is becoming a bit unstable. Here are the symptoms:

- Bulldoze becomes extremely slow and unresponsive. Bulldozing just one piece of track (or maybe a service tower) is a matter of "point, click, wait". Sometimes the "wait" part is up to 10 seconds.
- The Train List in the Ledger is starting to cause crashes. I went through the pain of testing this with multiple restarts. What causes the crash is trying to order trains in the list by Age, Type or Profit (this year or last year). Funnily enough, ordering by Speed, Oil or Water doesn't cause a problem. When RRT3 crashes, sometimes I get the "RRT3.exe has encountered a problem..." Windows message. Sometimes the message is sort of there on screen (the mouse pointer becomes the standard Windows pointer in a message-box-sized area of the screen), but the message box itself isn't visible!

My system is RRT3 1.05, Windows XP SP3, dual 2GHz processors, 4Gb of RAM. The scenario is Germantown - by about 1892 (playing beyond Gold, which is what I love doing, if only RRT3 will behave itself as the economy and rail system becomes more complicated...)

I went into this game and added a train, just to check.... no, the number of trains hadn't reached 255 (that would be a dead giveaway!), but 218.

This is odd, as this post: viewtopic.php?f=82&t=3376#p32538 in one of the Sticky threads in this forum mentions running 866 trains!

Are there any known fixes or recommendations to get RRT3 to be more stable?

thanks for any suggestions!
User avatar
RulerofRails
CEO
Posts: 2063
Joined: Sun Dec 08, 2013 1:26 am

Re: RRT3 crashing Unread post

The bull-dozing delay is pretty normal in my experience with a large network.

I know there is a problem with sorting large numbers of trains and industries with patch 1.06 which I believe was inadvertently caused by some of the patch changes. I haven't heard of this problem in version 1.05.

I have two questions:
Are there any AI on the map? Do they have any trains?

Have you found this problem on other maps as well, or just this one? If you want I could dig up some big saved games if you want to quickly try some different maps.

As far as the 866 trains, you need a super system for that. Take a look at thietavu's system specs in his signature (below his post). With my laptop (2.66Ghz, Win 7 32-bit), 300 trains is a safe limit before overload and things slow to a crawl. The game shouldn't crash when this is exceeded, it will just get painfully slow for time to pass and the graphics will stutter. Using some type of gamebooster program might help to raise the limit on a slower system.
User avatar
undertoad
Watchman
Posts: 85
Joined: Sun Oct 04, 2009 8:21 pm
Location: Glasgow

Re: RRT3 crashing Unread post

Hi RulerOfRail!

Thanks for your reply.
RulerofRails wrote: I have two questions:
Are there any AI on the map? Do they have any trains?

Have you found this problem on other maps as well, or just this one? If you want I could dig up some big saved games if you want to quickly try some different maps.
At this point there are no AIs on the map. I do remember this problem also happening on Unification of Germany, though I don't have the save anymore. As on this map, I was enjoying playing well past Gold, trying to keep a larger and larger network working efficiently, just as thietavu does. Actually I don't remember this specific "Train list in ledger crash" problem - just that (as you say below) the game became painfully slow.

If you do have some large save-games handy, that would be great to try them.
As far as the 866 trains, you need a super system for that. Take a look at thietavu's system specs in his signature (below his post). With my laptop (2.66Ghz, Win 7 32-bit), 300 trains is a safe limit before overload and things slow to a crawl. The game shouldn't crash when this is exceeded, it will just get painfully slow for time to pass and the graphics will stutter. Using some type of gamebooster program might help to raise the limit on a slower system.
Thanks, that's a useful benchmark showing up how relatively outdated and slow my system is: 2x2.0GHz processors, 256Mb graphics memory. I'm guessing that those - especially processor speed, because RRT3 almost certainly won't use multiple processors effectively - are the limiting factors.

You mentioned gamebooster programs - do you know of any particular one that works well with RRT3, or should I just trawl the web for some and try them out?

thanks!

undertoad
User avatar
Hawk
The Big Dawg
Posts: 6504
Joined: Fri Nov 10, 2006 10:28 am
Location: North Georgia - USA

Re: RRT3 crashing Unread post

undertoad wrote:because RRT3 almost certainly won't use multiple processors effectively - are the limiting factors.
RT3 doesn't even now what a multi-processor is. :mrgreen:
Hawk
User avatar
RulerofRails
CEO
Posts: 2063
Joined: Sun Dec 08, 2013 1:26 am

Re: RRT3 crashing Unread post

256MB of graphics memory is on the low side. I have 1GB.

Right now I have JetBoost, free version installed. (Please be aware that this and other freeware options need to be installed carefully as they are often sponsored and will try to set your homepage to bing and other crazy stuff. Always choose custom install and read all the info carefully before and during installation.)

I don't use the booster all the time. I will run the game with other programs running. When performance suffers I close the other programs. If I play on for some time and performance starts to suffer again, I will then turn the booster on. I don't use it very much at all.

A better every-day use option if you are running Windows 7 (probably true for most operating systems after XP?) is to search windows (in Win 7 from the start menu) for View Advanced System Settings. Then un-tick most of the Visual Effects options. These are just visual, so it's a question of function over form. This made my computer a lot faster at general tasks. The other thing I did that helped was to install a solid state hard drive a couple months back. They are constantly becoming more affordable.

A fair portion of my large saved games are from paying Arop's maps, so they are 1.06. I found 3 for 1.05. The maps are Michigan 1830, North-South Corridor, and Four Roads to Berlin. Michigan 1830 is the most insane. It takes maybe a minute for me to load and has 358 trains. Let me know how you go.
Attachments
Mich1830 114.zip
(7.6 MiB) Downloaded 269 times
NS 004.zip
(6.83 MiB) Downloaded 259 times
Four Roads 143.zip
(7.46 MiB) Downloaded 268 times
User avatar
undertoad
Watchman
Posts: 85
Joined: Sun Oct 04, 2009 8:21 pm
Location: Glasgow

Re: RRT3 crashing Unread post

Hello again!

Thanks for the reply and those save games.

Here are the results:

1. Michigan 1830
This took a whole 5 minutes to load on my machine! But loaded with no problems. But funnily enough, once in, it ran beautifully, at least in my terms. I generally play at Slow or Very Slow speed at this stage, due to the sheer number of things to do, so the game's ability to zip through time at Fast/Very Fast is not that important.

Interestingly, the Train List in the ledger re-ordered trains in the list with no problems (unlike in my game with only 218 trains). But... when I tried to Page through the list, it got more and more sluggish, until on about the 4th page-turn the whole game hanged (I also observed this in my own game). I wonder, does this also happen with this game or similar large games, on your machine? Perhaps you don't use the Ledger Train List much and thus haven't come across the problem?

One thing I noticed in your Michigan game is that though there are a lot of trains, the map itself is very sparse. There are far fewer buildings and industries than on my Germantown c.1895 map. (Continuing to play in this crowded, complicated environment is exactly what I'd love to be able to do in RRT3 - I'd love the cost of demolishing to find a road to become a problem, as it is in modern-day Europe). I want to load both games up and compare the Editor's object counters, but I haven't had time yet.

I suspect that RRT3 runs into problems not just because of the number of trains, but because of the number of buildings or even trees. Looking at the save-game files: my Germantown one that's causing problems is 36.5Mb on disk, about as large as your NS004, and getting close to your Michigan game at 45Mb.

Off-topic: how are all your trains at 99% oil and water? Excellent management, or a cheat? :lol: And that 4-way junction SW of Chicago - how on earth did you manage to build that using RRT3's "interesting" track-laying tool?

2. NS004
I couldn't load this one, as I got an error "Pack element < P-2T_Body.3dp > not found". I suspect this is a car or loco that I haven't got installed.

3. Four Roads
This one got to "Loading Car Database", like NS004, but then hanged. I suspect that there was actually a "Pack element" error, but it was invisible. This is a perennial problem with RRT3 on my machine: when an error occurs, the display doesn't react properly, even to Alt+Tab or Ctrl+Alt+Delete. The mouse pointer changes (I can even get it to change to the "resize window" arrows if I move it carefully to the right place) - but the window(s) itself isn't shown, as the screen (apart from the taskbar, sometimes) remains stuck on the hanged RRT3 display. Have you ever come across this, or is it perhaps something that happens because of my low (256Mb) graphics memory?

Could you tell me your own machine spec, and what problems (if any) you run into running large RRT3 games like these? I'm thinking of buying a new machine, with Win7 64 and and SSD drive at the very least, and so it would be good to know how much a better machine solves these problems. On my own machine, I notice that RRT3 exe seldom takes more than about 200-250Mb of memory - but consistently takes 50% of CPU power, for as long as it's running. (Oddly, this 50% load seems to be distributed between the two processors I have: something in the hardware must be distributing the load, without RRT3 "knowing" that there are two processors).

My own spec:
Laptop
2x2.0GHz processors
4Mb memory (XP only uses 3.5 max)
Win XP
Very aged hard drive: solid and reliable, but slow and full.
256Mb Nvidia Go 7600 graphics card (a superb laptop card back in the day, but very outdated now!).

I'll give Jetboost a go as well. Thanks for the warning! I hate those little "default Yes" checkboxes: "YES, I'd like you to install loads of random stuff, change all my settings and take a small, random amount out of my bank account every month, no need to provide card details as you have them already!".

thanks!


undertoad
User avatar
Wolverine@MSU
CEO
Posts: 1166
Joined: Fri Nov 10, 2006 2:14 pm
Location: East Lansing, MI

Re: RRT3 crashing Unread post

undertoad wrote:1. Michigan 1830
One thing I noticed in your Michigan game is that though there are a lot of trains, the map itself is very sparse. There are far fewer buildings and industries than on my Germantown c.1895 map.
Off-topic: how are all your trains at 99% oil and water? Excellent management, or a cheat? :lol: And that 4-way junction SW of Chicago - how on earth did you manage to build that using RRT3's "interesting" track-laying tool?
As the author of Michigan 1830 I can answer these two questions. The map is sparse because in 1830 Michigan was still a territory and was pretty much a "wilderness". I've set the growth rate somewhat on the high side for the first 20 years so while it starts out slow and you have to be judicious in placing your initial track and trains, things pick up rapidly. As for the 99% oil/water, there is an option to have trains refilled at stations for an increase in station cost. If the player chooses this option the event reduces oil/water (and I think sand too) to the point where they use virtually none over the course of the game.
User avatar
RulerofRails
CEO
Posts: 2063
Joined: Sun Dec 08, 2013 1:26 am

Re: RRT3 crashing Unread post

I tested this some more in the Michigan map on two machines. I can sort all the categories of the engine list fine from the first page, but after sorting for any of the profit categories I get a crash when going some pages in. I also see the slow down you are describing as I go further into the list, but it never freezes completely. So I can only assume this a bug in the game when some internal tables are overflowing. If theitavu or others that run large games can test or give info on whether they are able to do these actions that would be great cause it looks like my game is doing the same thing yours is.


NS004 and Four Roads:
Sorry. :oops: That's the problem with loading other people's saved games especially those containing the "Allow EVERY ... engine" switches. I had assumed that this map didn't use the "Allow EVERY North American Engine" switch. But, it does. This means that any custom engines installed will be available to play with. (I don't understand the philisophy behind using these switches because allowing new engines from the future by default changes the game play.) So I used quite a variety of engines, and thought that they had been included as extras to install with the map. Anyway, the scenario uses the G5s and P-2 engines, as well as Ned's custom skins for the F3. I forgot about the F3 skin and assumed that the G5s and P-3 were extras that downloaded with the map. If you want, try installing those things and seeing if it works.

Four Roads uses the "Allow EVERY European Engine" switch. This allowed the AI to buy some P-2s so you will need that. In this case I didn't check what the AI had. What I could do is replace all those engines and re-upload the files. Why don't you try and see if adding the engines and skin I mentioned form the New Engines section will work. Either way let me know.


Mich 1830:
I was testing a newer laptop which came with Windows 8.1 and Nvidia graphics when getting the fast load time. It uses the Vista Fix of disabling Hardware Transformation and Lighting, so due to wash-out on the price maps it's no good for playing the game, unless I get this alternative method running without using too many extra resources (yet to try it, will report). The newer machine gets 1 minute 15 seconds load time without booster. On my main laptop it takes 2mins 45secs with the booster enabled.

Saved game size is dependent on the size of the map itself (the gmp file). I haven't checked it out too thoroughly but most if not all of that info is the basis for the saved game file (gms file), with obviously more info added. This must be taken into account when saying that a certain game is very big. I have a saved game from Hexum that is 75Mb but only has 180 or so trains. In 1.06, I have had a saved game at almost 90Mb. I think that the amount of trees and the size of the map may cause some slow down, but hitting these overflows in the code is what really causes the freeze type slowness.

That interchange near Chicago is a bit cheaty, because there is a crossing of track without overpass or connection right in the middle. Basically the idea is to use short-radius turns because after all, RT3 doesn't slow trains down very much on turns. I probably used the bulldozer to make a track stub heading in roughly the right direction and then do the connect x2 connection starting at that stub, I just don't remember exactly. I hope my track laying skills have improved. Now I use a simpler system that takes a little more room but is still quite effective. It's basically two T's (connections in both directions) as close as possible. I can mock up an example if you like.

If you want a better computer to play RT3 I would recommend something without Nvidia. I am not sure how much benefit 64-bit is either. I don't think RT3 uses enough memory that it matters that 32-bit operating systems are limited to 4Gb of total RAM for the system as long as you aren't running it with lots of other programs at the same time. Clock speed matters especially when coming up against these barriers of overflow. I need to use laptops because they are portable, but their clock speed is in general lower than desktops.

My main laptop (over 5 years old) is a 2.66 GHz dual core, ATI Radeon 4650 graphics. This setup needs good cooling. I have had to disassemble, clean the dust, and re-do the thermal paste a couple of times. It's hp, the cooling was never it's strong point. The SSd is only 240Gb. My original drive was failing. I could have had a 1Tb normal drive for the same price, but I already had back-up drives for storage so I figured I would go for it.

Obviously as shown above, this machine gets the same problem you describe here. Only oddity is when starting the game. I need to wait for the game to start before clicking on anything on the desktop etc., otherwise the game will be running in Task Manager but never appear. As long as I remember that, the game starts well every time.

As I said, this computer isn't ideal for these really large maps. It will go into extremely slow "freeze" mode on some of these large games at the 300+ trains level. If the game is running on very slow, it's possible to go further before this becomes noticeable. (My gameplay preference is Normal or Fast and then pause when I want to perform an action.) The freezes are less after I changed the Windows Performance settings and got the SSD. In fact I haven't played a super large game since I don't believe.


Interesting that you like to do some bulldozing. Some maps have cities where you need to bulldoze to get a route through (Caucasus in Asia section comes to mind, it's not a perfect map but worth a look), but I haven't seen a whole map built that way. The demand map for a resource can get quite complex if there are many industries that demand it. My methods of play involve some control over this, but it's only a guiding hand to the game's natural systems. With a map like that, I would have to abandon any hope of influence resource travel and just accept more randomness. As with any new idea, there's good potential to make an interesting and unique scenario/s.
User avatar
undertoad
Watchman
Posts: 85
Joined: Sun Oct 04, 2009 8:21 pm
Location: Glasgow

Re: RRT3 crashing Unread post

Hello again, and thanks for the reply!

That's interesting that you too are getting the crash in the Ledger Train List. Reassuring, in a way, in that it's not just my ancient machine that has this problem. But bad news in another, in that your better machine doesn't solve this problem. It would be great if thietavu stopped by, interesting to see if he ever runs into this problem, on his warp-speed machine.

No problem about the extra engines. I think testing on Michigan has shown that this ledger bug isn't just a weird one-off.

I'm pretty sure that my save game file is so huge because of the densely-populated map. All those houses, industries, and "cars" of goods knocking around have to be recorded somewhere. The goods map on my Germantown c.1890 game makes the goods map for your Michigan game look like a desert! (As Wolverine explained, this is deliberate).

I'm trying to look for a better machine that will also be good for other things (particularly my work, which involves Visual Studio and SQL Server), rather than just to solve the RRT3 problem: that's why I'm looking at Win7 64, and lots of memory (at least 8Gb) which RRT3 will never come close to using.

That makes for an immediate dilemma: processors these days tend to be multiprocessors, with each individual processor not being that much faster than the 2.0HGz I have. That's great for modern software that can use it - but as far as I can tell from the expertise on these forums, what RRT3 really needs is just one, really fast processor. (The fact that my hardware assigns the RRT3 load across the 2 processors probably - I guess - just saves on cooling and prevents one processor from maxing out. It doesn't make RRT3 run any faster, probably because of some optimisation for parallel processing that RRT3 doesn't have. Does that sound right?).

(Thinking about it, the amount of number-crunching RRT3 does in the background, as a fascinating economy-simulator, would lend itself really well to parallel processing. Oh for the source code!)

So while I can go for a modern 6-core (e.g. the AMD FX-6300, clocked at 3.6GHz), I don't know how much difference it'll make to RRT3.

You also mentioned nVidia - is that the general impression you get from the forums, that RRT3 doesn't play nice with nVidia graphics cards?
Interesting that you like to do some bulldozing. Some maps have cities where you need to bulldoze to get a route through (Caucasus in Asia section comes to mind, it's not a perfect map but worth a look), but I haven't seen a whole map built that way. The demand map for a resource can get quite complex if there are many industries that demand it. My methods of play involve some control over this, but it's only a guiding hand to the game's natural systems. With a map like that, I would have to abandon any hope of influence resource travel and just accept more randomness. As with any new idea, there's good potential to make an interesting and unique scenario/s.
That, in a way, is what I enjoy about playing beyond Gold on the campaign maps. It becomes a completely different scenario by itself: one of managing a busy, complex network, trying to capture as much traffic as I can, and keeping my industries running at max efficiency. All with plenty of money to spend, and the prospect of more and more interesting trains to buy! It raises interesting challenges:

1. Managing complex industries like steel mills, and preventing them from stealing one resource while being starved of another. My steel mills in Wisconsin have iron to spare, but little coal. Meanwhile the ones in Kentucky and Michigan have plenty of coal, but need more iron.
2. More generally, running long-distance trains to distribute goods across "low-price sinks". For example, I have profitable lumber mills in the Ozarks, pumping out lumber which would sell well in Ohio and Michigan. But it won't move by itself without a long-distance run, because there's a low-price "sink" around St Louis and Indianopolis which they can't get across by themselves.
3. Managing and maximising passenger traffic. Especially difficult because the Express Goods map is confusing, showing arrivals as well as potential departures unless you click on a particular station to get the breakdown. And you can't get a clear idea of how good your existing schedule is (e.g. "5 trains take passengers from this station in this direction, here's their ETAs).
4. Keeping profits rising to keep the Board happy. Though I don't see why I should listen to them, I'm the 76% majority shareholder... :lol:

I'm thinking more and more that this a different kind of game, that RRT wasn't really designed to handle (and certainly wasn't tested for!). But it's the kind of game I like.
User avatar
undertoad
Watchman
Posts: 85
Joined: Sun Oct 04, 2009 8:21 pm
Location: Glasgow

Re: RRT3 crashing Unread post

Wolverine@MSU wrote:As the author of Michigan 1830 I can answer these two questions. The map is sparse because in 1830 Michigan was still a territory and was pretty much a "wilderness". I've set the growth rate somewhat on the high side for the first 20 years so while it starts out slow and you have to be judicious in placing your initial track and trains, things pick up rapidly. As for the 99% oil/water, there is an option to have trains refilled at stations for an increase in station cost. If the player chooses this option the event reduces oil/water (and I think sand too) to the point where they use virtually none over the course of the game.
Thanks! That's an interesting option for the oil and water. I spend so much of my time worrying about off-route maintenance.

I get what you say about the sparseness of the map. 1830 is really early for a railroad in that part of the world. Everywhere I play a map (Michigan is obviously part of my Germantown game), it gets me interested in the actual history of the place.
User avatar
Hawk
The Big Dawg
Posts: 6504
Joined: Fri Nov 10, 2006 10:28 am
Location: North Georgia - USA

Re: RRT3 crashing Unread post

undertoad wrote:(The fact that my hardware assigns the RRT3 load across the 2 processors probably - I guess - just saves on cooling and prevents one processor from maxing out. It doesn't make RRT3 run any faster, probably because of some optimisation for parallel processing that RRT3 doesn't have. Does that sound right?).
Where did you get information that RT3 is actually using both processors?
undertoad wrote:So while I can go for a modern 6-core (e.g. the AMD FX-6300, clocked at 3.6GHz), I don't know how much difference it'll make to RRT3.
Nope!
undertoad wrote:You also mentioned nVidia - is that the general impression you get from the forums, that RRT3 doesn't play nice with nVidia graphics cards?
I run NVidea with no problems. There is a problem with RT3 on newer cards though. Cards with over, I think, 768 MB RAM
Hawk
User avatar
undertoad
Watchman
Posts: 85
Joined: Sun Oct 04, 2009 8:21 pm
Location: Glasgow

Re: RRT3 crashing Unread post

Hi Hawk

Thanks for the reply.
Hawk wrote:Where did you get information that RT3 is actually using both processors?
From Task Manager. Overall, RRT3.exe is using about 50-55% CPU. This is the only heavy CPU usage going on (apart from 1-2% processes that blip in and out of existence on the Processes list). When I look at the Performance CPU graphs, both processors are running continously at about about 60% usage. If RT3 was using only one processor, I should be seeing one loaded processor (the one RT3 is using, loaded to 100%, which is 50% of 2 processors), and one very idle one (perhaps handling some of the "blippy" things). So it seems that RT3's work is being distributed across the two processors.

I'm no expert on multi-processor programming - but it seems possible to me that while the hardware is managing to distribute some work from RT3 (in a way that might be "invisible" to RT3 itself), RT3 isn't really designed for it, which is why everyone knows that RT3 doesn't use multiple processors well.
undertoad wrote:So while I can go for a modern 6-core (e.g. the AMD FX-6300, clocked at 3.6GHz), I don't know how much difference it'll make to RRT3.
Hawk wrote:Nope!
Do you mean, nope, it won't make much difference? I'm hoping that it'll at least make some difference, even if it's only the difference between one 2.0GHz processor and one faster 3.6GHz one.
undertoad wrote:You also mentioned nVidia - is that the general impression you get from the forums, that RRT3 doesn't play nice with nVidia graphics cards?
Hawk wrote:I run NVidea with no problems. There is a problem with RT3 on newer cards though. Cards with over, I think, 768 MB RAM
That's good news! I haven't followed modern hardware closely enough to say sweeping things like "nVidia is better than AMD" or vice versa - but it would be a pity if RT3 just didn't like one or the other of them, forcing me to go for the one it does like. Over here (viewtopic.php?f=82&t=3376), thietavu is finding that the problems are in the 512-768Mb range, and that over 768Mb everything works fine. Of course, his whole setup is pretty hardcore. But I hope he's right, as it's pretty difficult to get a video card <1Gb these days.
User avatar
RulerofRails
CEO
Posts: 2063
Joined: Sun Dec 08, 2013 1:26 am

Re: RRT3 crashing Unread post

thietavu stops by from time to time. But, it can be months. Who knows if asking question on one of his threads would help. Depending on whether he has email notification setup.

Did you try Oilcan's trick of setting up trains between the Steel Mills so that they make a round trip hauling Iron in one direction and Coal in the other? It's a nifty little trick. Also, I realized that a big stack of Coal at a Steel Mill at low price is a great source of short-term profit. Depending on where Coal Mines seed, the Italy map can have a nice big stack of Coal at a Steel Mill just over the border in Yugoslavia. This can be worth at least a million if it's exploited right. In this situation the mountains are helping to keep prices low.

In my opinion the clock speed should help. I don't understand it very well, but multiple cores seem to be for parallel processing where a program is sending out simultaneous request for different things that will end up working together to achieve the same goal. Older programs are designed generally to send out something to do and then wait on that to finish before performing the next step. (Don't know if that's true, but that's my simplistic understanding.)

I believe thietavu is talking about the blurry textures issue in that post. Needing the Vista Fix is a separate issue. I can say that my main laptop ATI Radeon 4650 (1GB, integrated) has the blurry textures issue. But, there are two ways to patch this that work perfectly.

Out of the cases where the Vista Fix is needed at least with Windows 7 (my recollection from reading the various posts and my own experience), the majority of cases for needing the Vista Fix seem to be for the Nvidia cards. But, since Windows 7 doesn't come with the typical computer anymore that may be less relevant. There's no way of knowing if Windows 8 or 10 behave the same way until we get more facts reported.

Thankfully, you can patch both issues, so you can get the game running nicely (hopefully on any setup), it will just take a bit more effort. See my experience with Windows 8 here. Others have had success with Windows 10 as well.
User avatar
undertoad
Watchman
Posts: 85
Joined: Sun Oct 04, 2009 8:21 pm
Location: Glasgow

Re: RRT3 crashing Unread post

Hello again!
RulerofRails wrote:Did you try Oilcan's trick of setting up trains between the Steel Mills so that they make a round trip hauling Iron in one direction and Coal in the other? It's a nifty little trick. Also, I realized that a big stack of Coal at a Steel Mill at low price is a great source of short-term profit. Depending on where Coal Mines seed, the Italy map can have a nice big stack of Coal at a Steel Mill just over the border in Yugoslavia. This can be worth at least a million if it's exploited right. In this situation the mountains are helping to keep prices low.
Yes - funnily enough, I came up with exactly the same trick, without realising that Oilcan had documented it! Steel Mills (and all 2-input industries) can be a bit stupid like that, stockpiling an enormous stack of input goods even when the other good isn't available. So I have my Steel Mills in Wisconsin and Minnesota hogging all the Iron in the region, but a bit short of Coal - while the ones in Kentucky are hogging Coal, but short of Iron. A couple of dedicated trains evens it out.

There's something slightly weird about this - an example of a breakdown in RRT3's otherwise excellent economics engine. 2-input industries create such a price-mountain around them that I wonder whether the (relatively-tiny) demand for e.g. Coal in the Houses around them ever get satisfied. It's an exploitable bug, as you say, because these industries can act as free gatherers for a good which you can then haul somewhere else up a steep price-hill the game hasn't evened out (e.g. because of mountains, or sheer distance).

I had this happen with a Lumber Mill I built somewhere around St Paul. Pumping out lumber by the millions of yards, and all of it being taken by a completely inactive Weapons Factory in Minneapolis. I could make even more money by holding a train at the Lumber Mill station, loading up lumber to take to the people in Houses in Iowa, who otherwise wouldn't have got any. But then if I wasn't careful, an auto-consist train from e.g. Des Moines would haul it straight back up to Minneapolis to sit in the yard of the Weapons Factory, before the Iowans had had a chance to get their hands on it! Market failure... :lol:
RulerofRails wrote:In my opinion the clock speed should help. I don't understand it very well, but multiple cores seem to be for parallel processing where a program is sending out simultaneous request for different things that will end up working together to achieve the same goal. Older programs are designed generally to send out something to do and then wait on that to finish before performing the next step. (Don't know if that's true, but that's my simplistic understanding.)
That's my simplistic understanding as well. I don't know about chip hardware, but the SQL Server database engine (which I do know about) can figure out whether a particular operation allows parallel processing using multiple processors, and makes that happen if it does. This works very well (with a few gotchas). I suspect that something in the hardware does something similar, and tries to turn RRT3 operations into parallel operations as much as it can. Which is probably not much: SQL operations are completely neutral about the messy details of how they get their results, because SQL is a very high-level language. RRT3's programming, much more low-level, probably has the necessity for one-after-the-other single-processor operation much more "baked in".

So while I'm seeing RRT3 putting a load on 2 processors, it doesn't result in much benefit, from the point of view of RRT3 itself or the tycoon playing it. SQL Server used to suffer from something called "wait states" - where an operation has been turned parallel, but the different parallel threads are constantly waiting for each other to finish something (because the operations really are better suited to "one-after-the-other" processing, and the optimiser has made a bit of a mistake making them parallel). That could be what's happening with RRT3: code that was designed for a single processor is being turned into a multi-processor code on the fly, but not very effectively.

I'm glad that modern hardware, and at least Windows 7 (results not in yet on Windows 8) seem to have few problems with RRT3, and there are fixes around (I'll be sure to look at your Windows 8 experience if that's the OS I go for). I should be able to afford a new machine in a month or two - look forward to smoother play!
User avatar
RulerofRails
CEO
Posts: 2063
Joined: Sun Dec 08, 2013 1:26 am

Re: RRT3 crashing Unread post

I know what you are talking about with those examples with Coal and Lumber. It's a product of the low demand amounts at the houses and how easily they are overwhelmed. This is exaggerated by the fact that deliveries by train are dumped at the station and must spend some time spreading out on the map towards the various houses before any meaningful consumption takes place. In that time price is dropping and likely the cargo can find a new place to go before much cargo reaches a house at all. This process can repeat many times. Good example of auto-consist re-hauling the same cargo.

The game's system of calculating the demand amounts is very robust from my experience and I know of no way to fool it. (I think most people that have played Trainmaster understand this pretty well.) An example of this is the size of Green Triangles for demand and Pink for supply. I haven't seen an example where the game calculates their size wrong. The triangles for each house are barely visible, but they are faintly there.

The house demand in RT3 is more useful for its demand instead of actual consumption of cargo. This isn't realistic really. A tedious way to help this issue is to manually seed (or place by event) many of the houses in a city on the same cell (or 3-4 nearby cells). I have never tried it as a map-wide thing. Visually cities wont look as good, but maybe who cares. (I didn't try the reserve cells trick. It's said that the game doesn't always remember manual changes, but I did notice that if you delete a city, it's reserve cells stay permanently on a map. Possibly placing and then deleting bogus cities could be used to careful increase the reserve cells of a certain city size. If this would work, then we could use the auto seeder to place the houses and then event the placement of industries.)

At one time I experimented with increasing the house demands. The problem was that the change in at-station demand between cities becomes even more dynamic which is a bad consequence. I saw the same cargo jumping between nearby cities even more readily. I thought to decrease the base prices for cargoes, but I have never invested the time into doing this. Maybe a combination of this and placing houses on same cells nearby the stations would improve this aspect. One of those things to test out that I might get to one day. . . .

The evented cargo production modifiers are robust. In long-term games I would encourage decreasing production of resources somewhat as time goes by to help prevent cargo gluts especially on medium to high density resource seeds. As a simple preventative, that's probably best and will bring some challenge to the longer-term game.
Post Reply