Your performance testing results

Holy crap. I can't wait to get my hands on the i7-7700K. After 2 years with my toaster (AMD FX-9370) which can't handle a single T75 without heating my room (I'm serious, after a gaming session my bedroom gets like 2ºC hotter than any place of the house), I will be able to do things like smashing 7 cars at once.
--- Post updated ---
After installing the i7, I will beamchmark (please make this a thing) for you guys.

 

I would love to see your results with FX too, would make nice baseline for you too to compare to and who knows might help some guys looking to build computer

 

Later today I'll try to do it .

Ah, I can help them. Do not buy AMD unless you're a professional overclocker
--- Post updated ---
Here we go...
CPU: AMD FX-9370 @3,80GHz
GPU: GIGABYTE GTX 1060 3GB G1 GAMING
RAM: HyperX Fury 8GB DDR3 1866 MHz
Motherboard: AsRock 990FX Extreme 3

BeamNG, High preset (no dynamic reflections), no FPS limiter, no v-sync

Gridmap:
Ibishu 200BX, Going to the ramp after the loop
Avg: 59/Min: 58/Max: 62
T75, Going to the ramp after the loop
Avg: 59/Min: 56/Max: 62

Jungle Rock Island:
Ibishu 200BX, Roads around the volcano
Avg: 53/Min: 49/Max: 57

Hirochi Raceway
Ibishu 200BX
Avg: 54/Min: 52/Max: 57
T75
Avg: 52/Min: 50/Max: 55

Not bad, but the results change a LOT when using more than one vehicle and after 5 minutes playing at 3,80MHz (enough for decent gameplay) the CPU thermalthrottles.
--- Post updated ---
I forgot the GPU things... crap

 

It would also be even more helpful if you could do exact same settings as in OP, because in BeamNG different settings change things a lot, they change CPU loading and GPU loading, each setting differently, so that is why my idea is to have everything with same settings to really see what difference hardware makes

 

Effect of overclocking.

I did overclock CPU from 3.7Ghz to 4.4Ghz, also ram went up from 2666Mhz to 2763Mhz and in addition to that Bios is older F5 instead of new F21 (I did check and difference between Bios versions to FPS was tiny to non existent).

Glass view mountains map was CPU limited 37 FPS at minimum, with overclock that increased to 44 FPS

Cinebench result was up from 163 to 195, so there was slightly smaller increase in FPS than Cinebench result, which might be within margin of error.

Average framerate increased 2 FPS, mostly because map was GPU limited most of the time, so faster CPU is not helping there, but faster GPU would increase average and max FPS on that map, faster GPU would do nothing to Min FPS though.

 

Ahh, sorry, when I have time I will do it for you :/

 

Cool, thx!

I discovered logging function of HWiNFO64 program, all hope is lost from this point on, be warned, only those who get their kicks from statistical analysis may survive

Still reading? Quick version, any dip of blue line means CPU is not enough fast, any dip to under 60 of orange line while blue line at 99-100 means GPU is not enough fast to run this map at 60fps and this applies only to this map, using settings and hardware described in first post.



Longer version:

Spoiler

Each map is different really, only vanilla maps seem to follow GPU rule under this test scenario (even some of those do get CPU intensive under certain situations), with mod maps it is more or less GPU or CPU, also each system is different, different GPU, different CPU cope with tasks differently, hardware sites show benchmarks and often give overly simplified claims that X is better than Y, while especially with BeamNG it is not possible to make such simplifications.

There are three major components to consider, CPU load (GFX), CPU load (physics), GPU load.

For me with Bios version F21 it is very easy to pinpoint which core runs GFX and which runs physics, however after downgrading F5 bios, it is lot harder as where F21 bios kept certain task running on certain thread as long as application did run, older F5 swaps threads around constantly. Seems not to have effect on FPS though, but figuring out why FPS gets low and GPU usage gets low is a lot harder, especially with task manager not being able to keep up with rapid changes.
I believe most will have this kind of tasks rapidly swapping cores situation.

Even with data logged with 500ms intervals it is bit hard to see why and what, but we can see how heavily cores are bouncing off at 100% limit at times and every time any core hits 100% something gets delayed, that something tends to be data GPU needs to proceed, which lowers FPS and GPU usage.





With F20 and F21 bios only one core would of been at or near 100% which would make easier to read graph, but I'm too lazy to change bios and also it might not be what most are seeing.

Another thing we can see from graphs is that GTX 1050 Ti GPU is not able to meet 60fps target even when CPU is not overwhelmed, GPU load remains at 100%, but FPS drop below 60, faster GPU would help there, also it would allow to run higher graphics level and effects, like enabling SSAO.

Now, overclocking CPU or getting faster CPU is option only so long, I don't believe there is CPU that would run this map at 60fps. That means to get map run at 60fps some other means must be examined. One would be editing visual distance in world editor yet another option would be turning shadows off which often helps CPU in such situations, but for example at Hirochi raceway shadows don't seem to load CPU much at all, so again, each map is different, like completely new game to be dealt with.

With Glass View Mountains map, lowering resolution is not going to increase Min FPS much if at all, Tennessee Min FPS is same, lowering light quality and/or shader detail level can in some cases help CPU, it is not completely clear why, but I presume it has something to do with number of objects casting shadows and needing shader effects, deleting objects works too but spoils the map.

Another strategy could be setting vsync on and setting FPS limiter to 35 or so, which should give constant 30fps and somewhat smooth gameplay, it is far more important to have solid set framerate than to have high framerate. Having constant framerate CPU and GPU has always little bit of extra computing power available and that is what makes gameplay smooth.

So by having framerate capped to minimum one can achieve on map is what is more important than 60fps, but 60fps does mean bit faster response when turning etc.

Learning about framerates different hardware can achieve by using same settings, one can learn which brand and model of hardware is best value for money for one's preferred maps.

When bananabench works again, I would think it should give quite good indication of which hardware works best for those needing to run several vehicles.

Update: I love Tennessee map, but TBH, my system can't run it very well. However it is quite simple fix to get map running just fine at least on my system, setting visual distance from over 30 000 to 2000 (details in first post, look for "How to double FPS" under Tennessee results) causes this much of difference:

Middle area you can see when I enter world editor, change value and hit play button, after that FPS almost doubles, turning vsync on would leave me some computing power to spare and map will run nice and smooth.

 

How do you make the FPS charts?

 

Best way is to get this and Fraps (link in 1st post), have Fraps running and hwinfo recognizes it as a sensor automatically.
https://www.hwinfo.com/download.php

Then in Hwinfo go to options, there is poll interval or something, default is 2000, have it 1000 or if you have fast system (including fast SDD), maybe 500 works too, but experiment with that.

I start logging after map is loaded and I'm ready to drive, alt-tab or alt-enter to desktop, then back, making sure your settings stay when swapping between full screen and windowed (resolution might be different among other settings). After driving I switch to desktop and stop logging.

Log is CSV file and just recently MS Office got new import tool which is awesome (you have to have insider fast version though), after importing start creating new chart to new sheet, select values for FPS, GPU load, etc. what ever you need, there is a ton of stuff to choose from.

I think that it could be possible to make some kind of script that does graphs from CSV logs, but I resort to manual labor, but those that are Excel gurus might get that part done a lot easier.

 

I did today Desert Highway, but it seems iGPU had been disabled from bios as I messed up with Bioses and overclocking, so I happily drove around thinking it will be recording, but it did not, however OBS still robbed some FPS on background.

Anyway, here is chart that might give some idea about it, it runs really well, despite there being something that is quite heavy on CPU, FPS remains high and with Vsync limited to 60fps there won't be any performance issues, however with weaker CPU single core performance, it is excepted that FPS drop will be greater.

 

I will try to record a B-25 Landing right now, can you upload a photo from Desert Highway ?

 

I think your in the wrong thread..

 

There is no need to complain . The record is made to show my laptops performances and I asked him about the photo to see how it looks like with his PC powers. Before you complain and put some hate , read the massage 100 times.

 

I don't understand.

 

Nothing man , was a a little bit annoyed about your massage. And sry fufsgfen for making this off-topic a little

 

Took awhile to get 3.2GB uploaded successfully, 32bit browser likes to be unstable from some reason.

If you watch full screen at 1080p you should see GPU load and it is near screenshot quality, I think.

 

The audio is cool, I can even hear the river sound at 1:30. 110 Fps is really good man , do you have any tips to get my pc/laptop faster? Because I really need that x), as I said before my game runs with 30-40 fps only.
Btw If I had a PC like yours I would beamng 24/7
(Just a fast off-topic question , when I ever try to record using the Xbox live thing on my laptop the bar says : you are currently not playing" please help)

 

Look "How to double FPS" from first post, there is information in a spoiler.

You could also set FPS limiter to 35 or so, have vsync on, it would not have more FPS, but it might be smoother. Then disabling shadows on some maps helps, but not always, it really depends what is limiting FPS, it can be CPU, it can be GPU.

If lowering resolution is not helping, most likely CPU is limiting factor. Run MSI afterburner and add GPU usage from options, then when you play you can see it same way as I have on video, whenever GPU usage drops below ~99% it is very likely CPU is too weak to keep up with GPU

If GPU usage stays at 99-100% whole time, then adjusting graphics details might help, also lowering resolution will help, but you have to experiment with it.

Reading longer version also might help to understand how to get better performance and at least it should help to figure out what is limiting performance:
https://www.beamng.com/threads/your-performance-testing-results.37866/page-2#post-568949

Xbox things, I have no idea.

 

I did test new map, called Beam Mountains, drove around almost 12 minutes and had no idea if I had passed starting location and how many times, lots of roads, inclines are quite ridiculous, but probably good for crashing things.

What was interesting for me was how map is GPU limited same way as Hirochi raceway is, there is two hiccups at beginning where you can see FPS dips sharply down, in reality map runs over 60fps whole time, so if I ignore those hiccups, MIN 63, MAX 117, AVG 86 and Median 85 FPS, while at maximum any CPU core does not go much above 20%, while GPU remains at 99-100%. With faster GPU CPU load will be higher, but what is interesting is that some maps have lot higher CPU use while FPS are in same range.

 

I have been thinking multiple vehicle testing, but haven't really come up with anything useful.

But I just noticed when making video for @Nadeox1 that difference with 1 cement mixer vs 2 cement mixers is not much more than 4-5FPS to slowest framerate during big impacts, would need kind of rail where to load many vehicles and let em loose to Cliff, terrain placeable tv-cameras etc. But there is something satisfying even with only two of them rolling down: