Beamng On Server Grade hardware

Right now my main rig is a 4690k at 4.5 ghz and 12 gb ram with a 1060 6gb
I got a hp z600 workstation with a E5620 right now I'm planning on putting in 2 xeon x5670 chips in it to have 12 cores 24 threads with 16 gb ram
Would this out perform be better then my i5 4690k? In terms of score the i5 4690k is 26% faster then a single x5670. But I didn't really find any solid information in terms of dual cpu performance..
Any help would be appreciated

 

Well, it surely will be able to handle waaay more vehicles at once

 

Of couse a dual socket system will be quicker than 1 CPU alone, but it´s like with SLI.
Don´t expect double the performance

 

Not really, well, actually that depends.

Graphics in BeamNG get very heavy on single core performance, for example i7-6700 paired with gtx1080 is quite often single core limited in fps, 8086k is not much better really, overclocked 8086k is of course.

When you keep SSAO off, reflection off, even shadows off and use normal graphics quality, you will get better use of several cores, but if using SSAO, reflections (even slightly), that is when your single core starts becoming limiting factor, especially when you have view / map with lot of materials, or several different cars that are using a lot of materials each.

In 0.13 BeamNG you need really fast single core to enjoy highest graphics quality in WCUSA or ECA, dual Xenon is bad news, but then again in a future this might change.

Physics wise, Xenon will be really good, but it is graphics that are problem currently.

 

ive noticed a single e5620 (what it has right now) is slower then my i5 4690k in terms of performance points (400 vs 800 cinebench points) but somehow on west coast usa it performs fairly simelair to my 4.5 ghz clocked i5. if not less micro stuttering. so this promises some good results

 

West coast usa is quite heavy on GPU, with 1060 you might be maxing out your GPU and difference in CPU might not show up directly in FPS, but with that kind of difference in Cinebench, I would think your i5 will do more cars at west coast than your Xenon, even with two Xenon chips you will probably not be able to meet performance level of overclocked i5.

Of course if GPU is enough slow, there will be not much difference, but put some faster GPU in and you will see limitations even with single car.

Bigger cache helps of course, but if you aim to keep up 60fps, then it might not be much good.

 

right now i am running a e5620 with 400 point in cinebench. my overclocked i5 700 the dual xeon x5670 gets 1400 points in cinebench. so 1 cpu is about 700 cinebench points. ive noticed for 10 euros more i can get dual x5675's and they get 1550 points so its something i gotta figure out. but sofar the 1060 6 gb is holding up nicely.. i see it isnt being used heavily since the fans cycle between idle and low rpm modes
--- Post updated ---
i know the single core performance is alot worse on these xeons but alot of games dont use single cores anymore. maybe minecraft but thats about it also beamng is the only game i play to be honest.. maybe some gta or far cry every so often but i'd rather waste my life on beamng

 

BeamNG is game that is most heavily using single core, it also uses all cores, but single core is very much what limits performance in BeamNG as way graphics are done, it uses very heavily on single core.

With I7-6700 I have been more often single core limited than GPU limited when FPS has dropped below 60. Of course there are maps where single core performance is not so important, but for example in WCUSA there are places where 3 cars make game CPU limited, because of single core limit.

Most people don't really realize this, but you can't rely on task manager or Afterburner to evaluate if single core is maxing out, not in BeamNG, you are hitting 100% in quick spikes that don't show up in any graphs, GPU usage drops and fps is not keeping up at 60fps.

Banana bench, 1 car tells about single core performance, my 6700 gets this and it is barely enough, put some headlights on and it is not enough, with much slower single car result you are going to struggle:


Then again, it might be that in future BeamNG is relying less to single core performance, but nobody knows when and if that will happen.

You know AMD Threadripper? Not much better than my CPU, even when paired with 1080Ti, GPU does not make much difference above 1070 in BeamNG.
https://www.beamng.com/threads/threadripper-bottleneck.56810/

In fact, between 1050Ti and 1080 at East Coast USA there was not a lot difference in slowest FPS, single core limiting so much there.

So I doubt very much that giving up single core performance in BeamNG will give you performance increase that multi core performance would make you believe.

 

Why use a server for gaming tho?

 

I think @atv_123 could give some more insides on how BeamNG runs on server-grade hardware

 

Indeed, also this thread might be interesting:
https://www.beamng.com/threads/interesting-low-performance-3770k-1080ti.56906/

 

I have uploaded other performance data and what not other places on this forum, but that was so long ago there just seems to be no way I can find that post. I'll upload a banana bench later when I am actually at my computer.

Mine is a:
Mac Pro 5.1
2X Intel 6 Core Xeon X5670 Processors at 2.93GHz (I think... I will have to look later if its these or the slightly slower ones)
32Gbs of DDR3 RAM at 1333MHz
and an NVidia GTX 1070

So mine will reflect fairly well what kind of performance you can expect... no I just have to get you some hard data...

Edit: To clarify... for the amount of hardware that is sitting there it seems like it should monster something like BeamNG drive, and compared to a normal computer, I guess you could say that it does... however the numbers you get back from a setup like this are what I would call "OK" at best. Yes, you can practically run 20 something cars before really noticing any lag, but your not going to be running it at 500FPS with 20 cars or even 1 car... the single thread performance is what BeamNG really needs and these processors don't really have a lot of it.

Edit2: Also... before anyone starts the "thats because Macs sucks" argument... I got the computer for next to nothing and I mostly use it for heavy FEA calculations that my work computer is incapable of doing in a timely manner with a poultry 4 core i7. When you have 24 simulated cores, FEA all of a sudden becomes a bit of a joke...

Edit3: Here you go... this is basically how this type of system will handle BeamNG.

 

Most people who want to build with server hardware are in it for the cpu prices:
https://www.ebay.com/p/Intel-Xeon-X...0614X5670-Processor/82011864?iid=332843159086

 

this looks better then my i5 4690k. also knowing bng wil be optimized in the future i think this is better

i'm going to go for 2x x5675 cpus. they're a bit better but not a whole lot. but as far as i can see the real time at 10 cars is higher then my i5 had. but yeah its fine for me knowing that beamng wil be optimized at some point(give it 2 years) and wil perform maybe better on high core count machines. but i geuss thats a risk i'm willing to take. and if it doesnt work out ill switch back to my i5. but so far i'm seeing simelair fps on a e5620 (even though its slower) then my i5 after 2 cars it starts to lag but thats due it being a rather low end xeon imo.

 

Well the biggest thing left to do is really optimize the graphics engine... the physics engine has been MASSIVELY optimized since the games inception. Granted... I always think they are done doing optimizations, and then they go an pull some magic out of their hat and optimize it even more... so perhaps it can even get better somehow...

But yeah, my computer usually handles just about anything you can throw at it, so anything you do with the faster processors should be fine.

 

CPU is being hammered by graphics, that is where they have ton of optimization room left.

For example ETS2 that is DX9 game can pull off 7500 drawcalls without a sweat, while BeamNG is usually maxing out at 4000 and that is with DX11, which at least in theory should have some sort of threading in too.

Not the physics, but LUA, UI, something, then something, which then limits available single core headroom I guess.

Also where other games don't render much behind the camera, BeamNG renders it all, whole map and every object, except few small ones. Not with all shaders though, but that kinda adds up to it why BeamNG is so incredibly heavy to run.

It just means that faster GPUs are not really worked out fully and that single core has to be pretty darn quick.

That part however is possible to optimize a lot and I don't see reason why not as it will help game to run at more cores. Single core performance is not going to rise much from where it is now in 2 or so years, they need to do something to make Italy to run 3 times faster and there is pretty much something that can be done.