TANE & TRS19 won't run with a RTX 3090

Another longshot, if you have Nvidia Experience running, remove it, that caused problems here, also the Win10 Game bar doesn't like Trainz much.

Got any other games you can try? may narrow it down to if it is a Trainz problem or driver / GPU one. If no suitable games perhaps the free trial of Passmarks performance test? It carries on working after 30 days with some of the more advanced stuff disabled that's aimed mostly at the benchmarking obsessed. https://www.passmark.com/products/performancetest/

Another thought the 3900s need two separate leads from the PSU, not one with multiple connectors, the builder may have slipped up there.

Maybe run GPUz at the same time and see what readings it's showing, also performance monitor and see if anything else is running, in particular on the TRS19 drive.

I have not waded through this thread, but I do open the Nvidia control panel and there is a tab to choose pre-configured settings for different games. I use that for TRS19. The other day, Black Ops was hanging after installation. And this technique fixed it immediately.

clam1952 said:

Another longshot, if you have Nvidia Experience running, remove it, that caused problems here, also the Win10 Game bar doesn't like Trainz much.

Got any other games you can try? may narrow it down to if it is a Trainz problem or driver / GPU one. If no suitable games perhaps the free trial of Passmarks performance test? It carries on working after 30 days with some of the more advanced stuff disabled that's aimed mostly at the benchmarking obsessed. https://www.passmark.com/products/performancetest/

Another thought the 3900s need two separate leads from the PSU, not one with multiple connectors, the builder may have slipped up there.

Maybe run GPUz at the same time and see what readings it's showing, also performance monitor and see if anything else is running, in particular on the TRS19 drive.

Click to expand...

Malc,

The xbox Games Bar is disabled and I've uninstalled Geoforce Experience. Neither make any difference to Trainz freezing.

However, the Passmark performance testing software indicates something wrong with the 3D performance of the 3090. It whizzes through part of the various tests with a ridiculously high frame rate showing but stalls with an "Unexpected Error". The driver is the most up-to-date for the GPU from Nvidia so ..... I've asked the PC provider (Chillblast) to have a look. They do have a very good after sales service so I'm hopeful they can spot a GPU driver problem and put it right.

The RTX 3090 is fairly new design and, as you mentioned in a previous post, the usual teething problems with early Nvidia drivers seem to be showing themselves.

Lataxe

Not sure if it helps, I'm running an RTX 2060 with the 461.92 Driver. Trainz is fine with that set up (Swap you?).

My PC was also built by Chillblast and has been pretty much rock solid.

I have my Virus Checker excluded from the Trains directories (and the Trainz directory in the user directory (only the Trainz directory - not the whole thing!)

One thing I have changed in the nVidia Control Panel is to limit the GPU to 80fps maximum. This has reduced the load on the GPU without noticeable effect.

Have you given TrainzUtil.exe sufficient rights?

Last thought - I've found that if I have to usw task manager to end Trainz, then I have to wait a little bit, then do a database rebuild - same as for a program crash as it appears to leave the database in a messy state -

Are you using "maximise compatibility" in settings?

Colin

Thanks all for the helpful suggestions so far. Unfortunately neither they nor the suggestions from the N3V Helpdesk have changed anything. Both TANE and TRS19 continue to freeze sooner or later (5 minutes at best, usually 20 seconds) after "Start Trainz" or starting a session/route from Content Manager.

I have managed to update the Nvidia drivers one increment, which should be automatic but isn't quite as it involves some sort of choice between a direct Nvidia download of the driver vs a Microsoft Store (Win10) download. That's now been sorted out and the GPU shows the performance it should when tested with the Passmark software that Malc recommended. No change to the Trainz problem though - which doesn't seem to be a GPU problem anyway as Task Manager shows little GPU activity when Trainz is started and none when it freezes, whereas there's significant amounts of CPU and RAM activity - doing what, I don't know. The logfile freezes too, with a final entry of:

; <NULL> GSRouterne(), 4 outstanding node(s)
; <NULL> GSRouterne(), 4 outstanding node(s)
; <NULL> GSRouterne(), 4 outstanding node(s)
; <NULL> GSRouterne(), 4 outstanding node(s)

This morning I managed to get a clean install of TRS19 to run a Kickstarter-2 session for nearly 5 minutes but it too froze and nothing would restart it. Same syndrome of zero GPU activity but an ongoing churn in the CPU and RAM.

The (non) progress with the Help desk has been somewhat glacial as I get one message every 24 hours (weekends excluded) with simple things like "try an extended database repair" or "move your Trainz installation to a different location". As though I wouldn't have already tried every get-it-moving-again fix already found a hundred times in the forum!

On the other hand, both TANE and TRS19 ran on my old Win7 PC with just the usual common-or-garden variety glitches such as CTDs, BSODs, periods of stutter and red bugs. The very same stuff transported wholesale on their dedicated SSDs from one PC to another won't run on the new PC. Neither will a newly installed "clean" version. This suggests that there's something in Win10 Pro or in the new hardware driver stuff that Trainz doesn't like. But, as ever, every other program I have transferred without issue to the new PC and they all now run like greased lightning.

SO - I have nothing left to try other than to wipe Trainz from the new PC and take up another creative PC-based hobby (probably 4K video). I'll miss the creative aspects of route building but, to be blunt, the last 6 months when Trainz was "working" have seen so many glitches and associated frustrations that I was about to give up on it anyway. Being unable to get it running at all on a modern PC with an RTX3090 GPU, a Threadripper CPU, and tons of Firecuda M.2 storage as well as 64 Gb of RAM ...... well .... goodbye Trainz and thanks for all the mish(mash).

Lataxe

Are you using "maximise compatibility" in settings?

This is an extremely important tip, even if it does not solve the OP's problem.

I have seen the program freeze, graphics wise, but the CPU and disk continue to operate. This lack of GPU activity is due to nothing moving around on the video card and the program is stuck because something is stomping on it. Have you checked your system components? I see you updated the graphics drivers, but how about checking your physical disks for errors, or perhaps your RAM? How about your Windows 10 install freshly installed or not?

Freezing after 5 minutes or so seems to be hardware related and not software.

For starters, run SFC.

Open a command prompt with admin-permissions.

At the prompt, type SFC /scannow and press enter.

This will run and clean up any Windows anomalies that may have occurred if you've reset your PC via the reset button or powered off while things are running.

After that I would check the RAM. Use MEMTST86, or the built-in memory tester.

I would check the disks for errors, and also check the RAM, power and heat.

It could be that one of these components, regardless if they are new or not, is having problems. It could be as simple as adjusting the timing on your RAM. I had an odd issue with this, which unfortunately lead me to replacing my RAM, due to the components having a timing problem. I would freeze randomly, crash to the desktop more often than not, and eventually I had only 32GB instead of 64GB installed. I removed the upper 32GB and the system ran fine, but if I installed the other 32 GB again, I could no longer boot! I then checked the RAM spec while it was what it was supposed to be, I should have used one on the recommended list given by the motherboard manufacturer. When I switched from Crucial to G-Skill, the problem went away.

RAM too could be sensitive to heat. A cold solder joint on one of the surface-mounted RAM chips on one of DIMMs can wreak havoc and create all kinds of weird and nearly untraceable problems due to thermal expansion. As a board technician, these things are pure hell to find and we ended up twisting the motherboard to find them because it caused the ever-so-tiny legs to lift off the surface just enough to interrupt and fail. Heating up, the components will move just slightly and also fail in a similar fashion, and this effect happens over time and not immediately. A RAM chip its self could be trashed and losing a bit here and there just at the critical time. A normal application may not hit that location, but if you're so happen to be lucky, or unlucky at this point, the programs will fail. I've seen this a few times as well. This is where MEMTST86 comes in handy.

Power supplies can do this. Trainz, being a graphics and system-intensive program, will push components hard. This doesn't matter what version, or what operating system and hardware drivers. The particular power supply that gave me problem was brand new. I got it along with other components, not for the system I mentioned before, but another build I did many years ago. Everything would run fine for about a half-hour, but as soon as I put any kind of load on the system, things would freeze. It turned out that the power supply couldn't put out what it was supposed to under load and one of the power-rails would drop suddenly. Thinking about it now, it was probably a bad voltage regulator or capacitor. I used to fix power supplies in the past and this was usually the problem. (Given that this was a switching power supply and I didn't have schematics, I would not attempt a repair on this unit. It's a good way of getting killed!)

Heat... Components don't like to get warm. Ensure your vents are adequately clear and that the components are adequately cooled. As things heat up, they become flaky, and over time sadly they become more sensitive to heat leading to early death. Been here and done that as well as a tech. We used burn-in ovens to find the early failures, but PCs operating can act as burn-in ovens since the heat inside the case is additive. This means the heat from your CPU is adding to the heat from your video card, RAM, disks, etc., and not being vented out fast enough. The solutions for this is add more fans, or changing the case for a more spacious one to allow for adequate airflow.

And finally, have you run an EDR? This may very well be required now that the program has crashed more than once. There's a good chance that you're driving yourself into a circle of failures due to the crashes on the crashes as things get more muddied and worse. It's like digging a hole and getting in deeper while trying to get out of the mud. So after checking the other things mentioned, run an EDR before trying again and see how that works out.

I truly understand your frustration.

adjusting the timing on your RAM.

if your bios has a default setting for RAM, or "use SPD" then try that. Pushing RAM timing will often result in not even being able to boot.

Re that GS Router problem, see this thread post 5 onwards seems to be related to a faulty locomotive in this case. https://forums.auran.com/trainz/showthread.php?148688-Environment-loading-please-wait

Lataxe - Don't give up hope - you know that TRS19 is worth fighting for!
There's a new Driver ver. 465.89 from Nvidia that is worth installing (and you can elect to omit the GeForce Experience package from installation).
This driver brings resizable BAR to the RTX-3090 and all other 30-series Ampere cards.
I had some start-up issues with an update to one of my Affinity desktop publishing programs recently which was resolved by removing the old drivers and DLLs for my unused Intel GPU integrated with my i7 CPU.
Perhaps you are experiencing something similar from the support drivers for a legacy hardware item or device?
Or perhaps your new Threadripper CPU is introducing some new spanners in the works when run in association with an nVidia GPU when run with high-intensity 3D loads?
My view is that your freezing issues with Trainz is likely less to do with your GPU than the CPU, and some sort of blocking protection is occurring whilst under certain loads.

After many weeks and a great deal of hardware swapping in and out, the problem described in the OP has been solved (sort of) and trainz is now running once more for me. Well, TANE SP4 is running whilst TRS19 is running in the fashion it did on the old PC - falling over regularly with a crash to the desktop (in Build 105100 - I daren't try those later builds yet)!

The problem has turned out to be a secret spanner in the works of the TRS40-Aorus motherboard with a Threadripper CPU and an RTX3090 mounted. A swap of GPU, CPU, power supply, RAM et al made no difference to the problem; but the substitution of a TRS40-Pro motherboard for the TRS40-Aorus version did the trick.

These component swap-outs were all done by Chillblast, the PC supplier. They eventually discovered that it was the TRS40-Aorus motherboard causing the problem but the MB was (naturally) the last component they tried swapping out. And they only did so when a (very) few other similar-build PCs were returned with similar issues. For the most part, an RTX GPU & Threadripper CPU do their jobs on the TRS40-Aorus motherboard; but various high-demand programs (such as 3D graphics building and rendering programs - and Trainz) caused glitches, from freezes to switch-off of the GPU use in program computation functions to crashes, both to the desktop or a BSOD.

There is still no absolute knowledge of what the MB problem is but this TRX40-Aorus issue was passed from Chillblast to Asus for investigation.

**********
So, if any out there have managed to acquire a new PC with Nvidia 30-series and Thredripper CPU on a TRX40-Aorus MB, and you're getting the Trainz-freeze issue ..... an MB swap is the answer. Such a PC with an Aorus TRX40-Aorus MB may also be the root of any problem with programs such as Blender; or any other computation-intensive 3D building/rendering program.

*********

Although Chillblast (the PC supplier) might be judged guilty of sending out a defective PC in the first place, I have nothing but praise for their after sales service. They spent much time giving me advice (things to try) then performing remote access investigations. When none of that worked, they sent me a new boxed Nvidia RTX3090 to swap out; and subsequently (when that didn't work either) took back the PC for extensive testing and hardware swap-outs of very expensive components at no cost to me other than a long wait for the fix. I've got an upgraded motherboard and power supply at no cost to me, not to mention new RAM and CPU (as a precaution). They've also restarted the 3 year parts, labour and return-to-base guarantee from when the PC ws finally fixed rather than from when I bought it nearly 3 months before.

Praise where its due.

Lataxe

Lataxe - So pleased that you finally got it sorted out - and grateful for you sharing the resolution here, for what ended up being a truly perplexing issue.
Glad to hear that the problem was indeed with hardware - specifically a mobo - and not TRS19-Platinum or the RTX-3090.
I was planning a Ryzen 9 5900X upgrade and pairing that with a RTX-3080 or 3080Ti and held off due to lingering doubts raised by your experience with a RTX-3090 and AMD's Threadripper.
So thanks! PC.

I too am glad to hear this, Lataxe.

I agree it's not the supplier's fault and it's the fault lies with the motherboard manufacturer. Reading about this, makes me think it's a timing issue with the motherboard that causes problems with the PCI-e buss. This is all tied in with the same circuitry that drives the NVME M.2, and even the RAM. All it takes is something that's out-of-spec in a critical timing circuits and everything else falls apart rapidly.