Optimization...

Can we have some more optimization features and improvements? I understand that we can't have everything we want. I'm sure there are other people like me, A terrible computer and not getting a better one anytime soon. Only able to run one map cleanly...Is it possible we could have a sort of "deletion of shadows" type thing or something to make it a teensy bit better for us sucky-PC users? Maybe a setting underneath "Lowest" called "Stupidly Low"?

I absolutely love this game but wish it ran better with my PC. Thanks for reading!

Rage

 

"lowest" setting is really the lowest I think.. I mean everything is disabled (shadows, reflection, HDR etc...) and everything is at the lowest quality (textures, lights, meshes)

 

I had thought so, but there could be things to take off to improve the performance even more. I dunno.

 

if you could make the brush so that everything is forward facing billboards no matter how close or far away from them you are, there would be a lot less lag. also lighting quality in the "lowest" mode is still somewhat high.

 

There were talks about people wanting a distance of render option. I would agree. My PC is an old (4 years old) mid-end by the time. And until i get 2k dollars to get a new PC i would like to see BeamNG better optimized and not just "cant do anything more lol, sorry. Get a better PC".

 

well said, mr death

 

$515.04. http://pcpartpicker.com/list/KVxyxY
It's a common misconception that you need a computer that costs in excess of $1500 to play PC games. That computer will run AAA titles are high (not ultra) settings at 1080p 60fps. Another $85 for a Core i5-6500. Both the i3 and i5 will run BeamNG quite well.

 

I live in southamerica where most parts are hard to get, are double the pricetag and mostly have a 50% tax for imports.

That would already make the same setup around 1400, lets sum up that fuck i3 (i have an i5 2500k and i REALLY want an i7), 8gb its almost not enough, but it doesn't has to be DDR4, the motherboard of mine i still think it works fine, and...


You know what? all i would need its a new CPU, a new GPU, and a new SSD and a new case. Maybe watercooling, but still, everything would cost me a little bit too much.

 

8GB is enough. It isn't future proof, but it's cheap enough, and doesn't normally have to be replaced. And with that build, yes, it does have to be DDR4. The motherboard doesn't support DDR3L. I'd also like to have confirmation that parts cost 3 times as much where you are.

 

I dont know if they are exactly 3 times as much where i live, but they do cost quite a bit more. And when i said "dont have to be DDR4" i meant because i think i could still use some juice out of this old mobo.

 

At least what I have observed, is that with amount of objects increasing, CPU load of single core increases and that is only thing really limiting performance on my system, by setting lower details I can get easily GPU load to be such that it runs on any map, but only by putting lighting on lowest and shader quality on low or lower this CPU load on single core can be made lower.

For that easy cure would be to set drawing distance to less than whole island, sadly that changes also terrain to be rendered at lower distance.

On official maps there are less objects than on many mod maps and on those GPU is only limiting factor with single vehicle, even with having only GTX 1050Ti I can still max out settings (not dynamic reflections though) on many official maps and still have above 60fps at minimum.

On many mod maps I can set everything to max and it will still run well over 100fps, but when looking inland where all those objects are, single core gets maxed out and performance decreases because of that, I'm not exactly sure how many percent faster single core performance there is available than on i7-6700, but I doubt that even 50% increase on single core performance will be enough.

Setting drawing distance less will help with that, but it can look funny as high mountains are not drawn to screen.

So if you could set land/terrain to be drawn for 10km but objects to 500 meters or even less, you would not get so much visual impact, however all those objects would not choke single core.

That would allow a lot bigger islands, even the official ones, as you could have a lot more objects and only those that are drawn close will need computing, bridge or lamp post that is 2km might not need to be computed.

Terrain itself is not making much of performance hit, physics alone are not too much, especially with just one vehicle, so I can see fairly well how optimization could give a lot of performance boost.

Shadows use quite bit of CPU time too, again same core/thread as it is with other graphics related CPU usage hit, so keep those off with lower end system, I think that even light quality set to lowest, there was still performance to gain by disabling all shadows, but it can also be my memory does tricks to me.

Difference between i7 and i3 is not much unless you like to run more than 1 vehicle or by night lights and shadows enabled. Difference between iGPU and GTX1050 is a lot more, with faster GPU difference between i7 and i3 could be more, but even 1050 is enough to run normal to high, on 1080p or less as long as you are willing to keep dynamic reflections off on most maps.

There are few (or at least one) mod map(s) where 16GB of ram is a must, I kept constantly running out of 8gb on other stuff I do on this system, so 16GB is really nice to have, but official content of BeamNG works fine with i3 and 8GB of ram, using only iGPU, it is not pretty, but it does play quite smooth as long as one keeps settings enough low and those shadows off.

@Nadeox1, this is short sample of my upcoming performance testing, not sure what we can do with it, but I try to cover all settings and their effect to my current system, will try to do 1050 vs iGPU comparison too.
Hopefully it can help to choose settings some day.

These are with i7-6700 GTX 1050 Ti:
Lowest settings, windowed mode 720p:
At East Coast Usa Highway , with D15, default time of day, hood view with UI off I get
-light quality lowest, shadows off 305-310fps, single core CPU load 100%, GPU load 70%
-light quality lowest shadows on 280fps decrease of 25-30fps, single core CPU load 100%, GPU load 68%
-light quality low shadows off 298 fps, single core CPU load 100%, GPU load 72%
-light quality low shadows on 180fps, single core CPU load 100%, GPU load 66%
-light quality medium shadows off 298 fps, single core CPU load 100%, GPU load 72%
-light quality medium shadows on 180fps, single core CPU load 100%, GPU load 68%
-light quality high shadows off 298 fps, single core CPU load 100%, GPU load 72%
-light quality high shadows on 180 fps, single core CPU load 100%, GPU load 80%

 

It took a long time to write up all the performance data, I still need to do iGPU version, but so far it looks like that my system is CPU limited more than GPU limited if I keep reasonable settings, despite having only 1050Ti .

Also it looks like that shadows are causing big load on CPU, also mesh quality seems to tax CPU somewhat, having that at lowest saves some CPU cycles.

There are barely any GPU load increase with shadows and only slightly with mesh quality.

SSAO and dynamic reflections (if set to very high setting) tend to load GPU most.

As all objects cast shadows, it is natural that maps with lot of objects cause CPU to choke.

 

Nice. But what did you expect from a 1050? you have a killer locked CPU and the worst new GPU you can get from Nvidia?

 

What do you mean? My testing shows that GPU is not limiting performance, CPU is limiting performance. Only if I set ridiculous graphics settings GPU becomes a limiting aspect, but only barely.

I run everything on high, only dynamic reflections is off, everything else maxed out and GPU is not reaching 100% load it is always CPU limiting fps.

60fps is what I limit framerate as monitor won't show more anyway and usually I use 720p resolution in a window, so I would not gain anything from faster GPU, only if BeamNG is optimized I can see need for faster GPU.

Bigger resolution is then another point where faster GPU would be needed.

Problem with getting faster CPU is that there is not really lot faster CPUs when it comes to single core performance, K models and Kaby lake are faster, then some really expensive HEDT platform CPU's are faster, so I really don't understand claims that 1050 would be a bottleneck, when it is mostly waiting CPU.

 

Was the CPU bottle neck after a certain point? i am OCing my CPU mostly to play BeamNG in a pretty good FPS with up to 4 cars in the screen at all times.

 

BeamNG is still in the process of being optimized. Only a year ago, I could only run ECUSA with one vehicle at 45 fps. Now with the same specs and graphics settings, I can run two vehicles on that map with 60 fps. So just sit tight.

 

Yeah, only with really really high graphics settings CPU was no longer bottle neck, using SSAO and dynamic reflections at very high settings are only things to bring GPU on it's knees at 720p.

On normal overall setting I can tick every eyecandy box and get around 80fps, CPU load on single core is 100% while GPU load is 100%, so that is about where it starts to be GPU limited, but if toning down dynamic reflections a bit, I get to be CPU limited. If I turn dynamic reflections off, I can set overall profile to high and have 90fps with 100% GPU load and single core of GPU something around 80%
All of those settings are still running solid 60fps when I set limiter on.

High graphics setting with dynamic reflections on, but turned down a little still gives me 60fps, albeit it starts to drop below on some places and that is with around 75% single core CPU load and 93% GPU load as you can see from screenshot below:




If you then play with 4k or fullHD, then you might need faster GPU sooner as resolution makes performance to be GPU limited much sooner.

I can run 6 cars until I start to notice slowing down (actually that + effect of resolution is something I want to test and put to my spreadsheet too), 11 cars still works somewhat, but is really laggy, one would need HEDT platform to run that many cars. I need to keep dynamic reflections off when running several cars and then it is pretty much CPU that limits again. Turning off SSAO will help with weaker GPU also.

For this low cost GPU what 1050 Ti is, I think one can't except much better than normal graphics settings, high would be for higher end GPU's, so I'm quite amazed how well this GPU performs in BeamNG.

Also I'm looking forward of what tricks devs have on their sleeve for future as they certainly know that they need to get as many CPU cycles freed as possible in order to improve performance in BeamNG.

For example shadows seem to use CPU quite a lot, I would love to test faster GPU and see what effect that would have to CPU usage of shadows for example. Also Bob wrote how he get lot of slowdown in FPS if camera is facing out to sea on his SoCal map, while I get slowdown when looking inland, not out to sea, he has AMD GPU, so it can be that different GPU has different effect on CPU load.

I intend to test iGPU as that is only other modern GPU I can test and maybe that shows some difference in CPU load, or maybe not, we will see that in a future.

Interesting curiosity, on normal overall setting every eye candy box ticked, I'm taking 47fps hit from checking those eye candy boxes, while shadows alone make 84fps hit on every light quality level.

Also not much point going lower than normal light quality on normal overall profile, there is no fps gains from some odd reason, well more testing, writing up and analyzing is needed to understand why what is happening: