More FPS

Spawn 5+ cars, and we'll see this smoothness

 

Quad core i5 750 @2.67 Ghz, Nvidea GT 320 and 4GB ram. everything on normal runs good. I get some lagg spikes in maps with alot of trees.

 

That is wrong. An AMD processor like an FX - 8 core will blow any Intel CPU in the same price range out of the water in BeamNG because it is heavily threaded, meaning it takes advantage of all the cores instead of just one or 2 (like RoR).

 

ok, sure. 5+ cars are insane for any system in high quality, HT or not.

 

AMD isn't well optimised in BeamNG and there are many threads complaining about it. It's not that the CPUs are bad it's that the single threaded performance really lets it down. There are Intel quad cores that run BeamNG better than the AMD Eight cores.

 

Yeah but play the full game. It lags a lot more.

 

Ah, that's my my game is inexplicably laggy. Guess all us AMD users will just have to wait for an update.

 

False. If you turn down the visual settings and look at peak FPS (an unimportant metric) then the single thread performance appears to be doing something. Of course, no one plays like that and all that really matters is how low your FPS goes, upward spikes do nothing to improve the experience. When it comes to handling as many operations as possible in parallel (many cars could and would be an excellent example), physical cores are going to provide a superior result, unless there are counter-productive 'optimizations' in place.

So far, I've see two people on the forums with quad core intels, one of them an ivy bridge, complaining about severe performance issues. This thread is about an elder, low-lock AMD dual core. Granted, I don't spend a whole lot of time looking around.

I suspect any AMD specific performance 'problems' would result from using both cores of one module at the same time when it's not necessary. I can safely assume they wouldn't run a different instruction set based on brand, that would be highly illogical, even compilers (usually) do make that mistake. Due to the modular nature of the modern designs, you only get ~80% scaling when compared to a single thread utilizing the module. Also, any 'optimizations' made to benefit HT are entirely counter-productive to physical core performance, including the physical cores on intel's non-hyperthreaded CPUs. Cinebench is an excellent example of the counter-productive attempts you see at utilizing a complex pipeline as much as possible by implementing what is a rather dated technology with little regard for overall performance. When Intel's pipeline is being properly utilized by the primary thread, HT provides 0- 20% scaling at a *maximum*, a metric that continues to decrease with every subsequent Intel CPU release as they improve their single-thread pipeline utilization, leaving little to no extra 'resources' available for solving whatever the secondary thread requires. Of course, if there is an expect result that must come before another, HT can result in negative scaling as a 'halt' (or whatever you care to call it) is necessary for certain kinds of sequential data. VMware would also be a good example, as thou shalt *not* assign a tile to a virtual 'core'.

There are many ways to utilize a pipeline properly, and HT isn't really one of them, especially when latency is a factor. Of course, the torque engine may have deeper 'optimizations' in place, the industry has been rather slow to move from the majority of all tasks single worker thread you find in dx9-based engines and the ones that simply add dx11 features on top of a dated engine without a deep revision will still exhibit the behavior of old code (because it is).


*edit* ^To the guy above, what's your configuration?

 

Reply to myself! I turned of HDR, changed my pc setting to 1 screen and updated the Intel drivers. It worked for me because I now have a steady 80-90 FPS. Some 2 car crashes will drop it to 15 FPS. How are the devs going to fix this if they come up with 8 car banger races??

 

yes, true. This is a pre-alpha, i'm sure that developers will optimize this game for a mid-human pc

 

No... He is right, the AMD cpu's no matter what are not optimized for BeamNG.drive yet and will lag... My friend has one of FX series "8 core" and it is total crap in this game compared to my 3770K!

 

AMD A8-4500M APU with Radeon graphics
1.90 GHz
6GB of RAM
64bit, x64 processor
Quad core

I know it isn't the best by far. I usually get around 7-8fps, but the first day I had the game, it ran at an acceptable (for me) 14fps, then the next morning ran at 7, something that makes absolutely no sense.

 

I've got
AMD Phenom II X4 965 at 3.4GHz,
AMD Radeon HD 6870,
8GB 1333MHz RAM

No overclocks or such whatsoever, getting a quite stable 45 frames per second.

 

3770K is over $100 dollars more expensive so apples to oranges. Also I have a Phenom II x4 and I get around 30 - 50 FPS so your friend is probably running SSAO + 4X AA on a some crappy low end GPU or full of bull.

 

For the price of the FX-8350 + 30 dollars you can get the intel 4670k which is the same for gaming than the 3770k because hyperthreading doesn't affect gaming performance. And remember power consumption. AMD CPUs usually have a TDP of 125W while intel's are around 77 and 84W.

 

So you're saying that BeamNG - Drive is more like a CPU powered game than a GPU powered one?

 

It is heavily CPU-dependent.

 

Hyperthreading can affect gaming performance, usually not very much, but if the game is coded to have good multithreading it well. Hyperthreading helps a lot in BeamNG.

 

Is it normal that BeamNG uses mainly only one of my four CPU cores (Intel Core 2 Quad Q6600)? Overall CPU usage is usually between 30-45%, while one single CPU core is at around 95% usage...

 

Yes, the physics runs one car per core.