8 core Any Value Over 4 in TS19?

boleyd

Well-known member
I currently have a very old 2500K 4 core over clocked to 4ghz and only 8gb of 799mhz of DDR3 memory. It does amazingly well with TRS19. All settings max ain't so good but slight reductions gives me good performance. So, the question is, setting aside clock speed, does TRS19 take advantage of the 8 threads (Hyper-Threading) in a late i7 CPU? Or, should I stick with a fast ,and cheaper i5?

I could probably continue this way, but I also sometimes have an additional window up to watch the news while a train is executing a 1 hour run to check trackage and signals on a modified route. The TV window runs well and I do not care about how well the TS19 window looks. BUT, the temperatures hit the CPU's thermal limit if I am not careful. Have to shrink windows.
 
Last edited:
It's your gtx 1060 than does all the work. The CPU matters but not by that much, so an i7 will give you around 1% better frame rates on average than an i5. The thing it does have is better caches than the i5.

Cheerio John
 
I have an 8 core computer and I don't think I'd ever want to so back to using a 4 core computer. My old monster of a Xeon workstation is not exactly new and the cpu speed might seem on the lowish side, but it's the old story of the wide slow moving river shifting more water than a narrow swift flowing stream. I run TS2019 at around 50% across all cores even on large and detailed routes and core temperatures stay inside normal ranges. I've got my eye out for the later model dual quad core cpu version of my HP workstation with improved specs over what I have and as far as I'm concerned nothing else will do.
 
This made me chuckle.


You are aware that when TS12 and TANE ran on the same machine the CPU temp was lower and the GPU temp was higher for TANE / TS19?

As long as the CPU in sufficient to feed the GPU that's all that really matters. Scripting is done on the CPU true and if you cherry pick the assets you might see more of a difference. TS12 yes you could overheat the CPU on many machines but TANE and TS19 do a lot more processing on the GPU.

Cheerio John
 
Agree with John.
My old pc hit CPU overheat problems with the older Trainz programs, and a modest CPU was fine.
New pc has water-cooling CPU (usually less than 40C), but the CPU fans spool up like a jet on takeoff (70C+).

It's well worth cleaning out the flurb from the fans, filters and case. The stock CPU coolers are rarely able to cope with prolonged heavy use.
Maybe worth removing the heatsink, cleaning and using new thermal paste. I changed cases on the old pc (more cooling fans + vent over the CPU). That reduced the CPU temperature significantly.

Colin
 
My most recent tests with 105096 on a heavily treed route, with a 1 hour run time, showed 100% many times on the CPU but only 40% on the GPU. Yes, the CPU feeds the computed sets of digital info to the GPU. And some of that (probably floating point) is done on the GPU. So, in my case, the conventional profile seems reversed. Or, the philosophy of system workload allocation does not hold in the case of TS2019.

My case is a Thermaltake with four 6-inch fans. CPU cooled with a large fan fed copper finned arrangement. Recently vacuum cleaned the the system - no significant dirt accumulation. More testing is needed before dusting off the magic $$$card.
 
An old CPU can bottleneck a new high end GPU. 100% CPU and 40% GPU is a good indication that it's getting bottlenecked.
 
It's well worth cleaning out the flurb from the fans, filters and case. The stock CPU coolers are rarely able to cope with prolonged heavy use.

I very much agree. My old Xeon workstation gets regular clean outs with both CPU coolers being removed and cleaned before reinstalling with new thermal paste. Being a professional workstation it has fans on everything including the RAM so its cooling is very good.
 
I am with boleyd on this. I am running an AMD Ryzen 7 2700 with an nVidia RTX 2070. I have 16G of dual channel memory. All Trainz settings are maxed out except post-processing. On an extensive tree laden layout, I get 30 -110 FPS. I see all 16 cores maxed out quite often at 100%. I have not seen all of the GPU memory used and system RAM used is generally around 7G.
 
I am with boleyd on this. I am running an AMD Ryzen 7 2700 with an nVidia RTX 2070. I have 16G of dual channel memory. All Trainz settings are maxed out except post-processing. On an extensive tree laden layout, I get 30 -110 FPS. I see all 16 cores maxed out quite often at 100%. I have not seen all of the GPU memory used and system RAM used is generally around 7G.

I wonder if the trees are making a major impact. I noticed when building Middleton for laptops adding trees put the frames per second through the floor. I think it also depends on the number of different trees used.

Having said that an i5-2500 first came out in 2011 so is a bit dated.

On my system TANE gives better frame rates than TS12 running the sort of layouts I normally run but that's a six core e5 1650 v2 with an RTX 2070. Running with a GTX 1050 ti which I was for a while until I picked up the 2070 gave lower frame rates and a lower load on the CPU.

The benchmark CPU score is more than twice that of an i5-2500k.

https://www.cpubenchmark.net/cpu_list.php

Cheerio John
 
Depends on which Trees are being used. With SpeedTrees some have little effect some are quite bad. On my Routes buildings seem to have more of a negative effect probably due to many having no lod.

The Trees? A bit tedious as I've done it but load each one into preview asset and test performance, if it drops dramatically towards zero fps don't use it, good ones maybe drop 2 to 5 fps.
The 9th Generation i5s actually have 6 cores, i7 8 cores neither hyper-threaded and the i9 8 cores 16 hyper-threaded. I would suggest unless obsessed with benchmarking the i5 9th generation CPU's would be more than adequate for TRS19.
 
Trees are mostly ULTRA series.

I started up my Lockheed Martin flight simulator. I had a similar spread of CPU higher percentage usage than GPU. Summer, some clouds, no rain low wind and set to show cloud shadows on the ground. I forgot about the 6 cores in latest i5's. I had priced a set of new components and, at that time, believed that the 9th gen units were more than adequate. Newegg has some bundles/kits) of motherboards, CPU, and memory. My case is old but very strong all steel . It weighs 47 pounds plus a 750 power supply. Its only drawback is that it does not have a big glass window so all of my guests can admire my COMPUTER and be very jealous. Cool baby!
 
Last edited:
Trees and what they are do make a difference. I changed out all the Ultratrees on a layout for Malc's billboard trees and the frame rate doubled.
 
I am probably using 80% Ultra Trees. Speed trees do not look as real and they also look brighter at their LOD level distance. Should be the opposite.
 
I am with boleyd on this. I am running an AMD Ryzen 7 2700 with an nVidia RTX 2070. I have 16G of dual channel memory. All Trainz settings are maxed out except post-processing. On an extensive tree laden layout, I get 30 -110 FPS. I see all 16 cores maxed out quite often at 100%. I have not seen all of the GPU memory used and system RAM used is generally around 7G.

Likewise here, running a Ryzen 3800X with a GeForce 2080 Ti and both were purchased with TRS2019 in mind especially after I saw the advanced post-processing that would be getting added to the game last year. Memory usage on the heaviest routes (really that has to be Central Europe at the moment though C&O can get close) does come close to 7G and when the load is heavy the GPU can show close to 100% utilization with the CPU not so much although the game is fully using 4 physical cores and a bit of the hyperthread pseudocores although the rest are also going to the OS doing file loading and video drivers. So as one poster here said, lots of objects with no proper lods seems to kill performance as that will get even this combo down to ~30 fps though as long as it maintains 30 I'll still be happy with that.


I currently have a very old 2500K 4 core over clocked to 4ghz and only 8gb of 799mhz of DDR3 memory. It does amazingly well with TRS19. All settings max ain't so good but slight reductions gives me good performance. So, the question is, setting aside clock speed, does TRS19 take advantage of the 8 threads (Hyper-Threading) in a late i7 CPU? Or, should I stick with a fast ,and cheaper i5?

I could probably continue this way, but I also sometimes have an additional window up to watch the news while a train is executing a 1 hour run to check trackage and signals on a modified route. The TV window runs well and I do not care about how well the TS19 window looks. BUT, the temperatures hit the CPU's thermal limit if I am not careful. Have to shrink windows.

Would you consider picking up a Ryzen 5 3600 (that is a 6 core 12 thread part for $200) and a suitable motherboard? It is not a major investment and the performance uplift would be noticeable for object-heavy routes, especially going with dual-channel DDR4 to bring the memory up to 16 GB - I think this combo would comfortably run TRS2019 (nearly) maxed for a good long time. Also as Malc said, current i5s are also 6 core parts so again, rather decent bump there.
 
I read this thread awhile ago, and since I was getting ready to make a new layout, I thought I would test a few things.

First of all my specs: Ryzen 7 2700, 16G dual channel memory, RTX 2070 GPU. The layout is 250 bases. I started by laying a little over 100 mile of sisde by side track for a scenic route. There is no AI involved.

After laying the track and having it reach the top of a rather high mountain (which uses about 1/2 of all baseboards), I ran one locomotive. All settings on high. FPS stayed constant at 200. I then laid 21 other tracks totaling about 200 miles. Ran loco: 200 FPS. This is still without any scenery whatsoever: no ground textures, buildings, etc.

Next I added consists on every track. Some were quite long, some (4) were just street cars. Ran loco again. FPS now at 35 -50. Got all of the other consists moving: 27 in all. FPS: 35-50. What changed? The CPU! Massive change. Before adding the consists, my CPU and GPU were pretty much doing nothing. RAM usage was about 6G. But, with the consists added, whether moving or stationary, the CPU started working quite hard. This CPU shows as a 16 core and about 1/2 of those cores were hitting 100% most of the time. The GPU was starting to work harder, but still pretty much below 50%.

I added ground textures, roads with moving cars, buildings, a few thousand trees, etc. FPS still 35-50. GPU doing a little more work, CPU now working its butt off: at 100% on most cores. RAM usage at 8G.

While I realize that you are not going to run Trz2019 while using an onboard GPU, don't underestimate the CPU's involvement. Had I been using any AI, the CPU stress would have undoubtedly increased.
 
Back
Top