Multicore optimization

[Bearcat]

I came across this post on Steam:

 

 

 

UPD: found some interesting information on Russian part of the forum. Game uses 3 cores (1 for interface, and 2 for math and graphics, measured on quad core CPU). It is confirmed that min FPS in most cases is limited by CPU and guys found 3.8 GHz acceptable (about 17 fps min) and 4.45 GHz enjoyable (34 fps min), measured on GTX680. I have i5 3.4 and GTX770 (it's on pair with 680) with 23 fps min so this knowledge seems to be true. It is also confirmed that the main CPU consumer is aircraft flight and damage model and that's the maximum devs were able to optimize without compromising the aerodynamics too much. Devs had a choice to simplify physics for bots only to find they become overpowered or too simple. Well, at least that's what I was able to find on forums. Also, some guys suspect the FMOD sound engine to be a problem in some setups. Try to disable all sounds in the game and restart it. If you get a good fps boost that means you probably can dig on FMOD vs Soundcard topics on the Internet.

 

Can anyone suggest what to use if you have a multicore CPU and you want to try to optimize performance for a multicore CPU? Can one of the devs perhaps drop in and just offer up a tip or two? I am just trying to squeeze out the most with what I have...
 

[-TBC-AeroAce]

Darn I have 8 cores, I wish it could utilise more as my game play is so blatantly cpu bottlenecked. Can get like standard 60 fps ultra on a sp mission with moderate amounts of ai but that will half or even more on a full mp server

Edited April 15, 2015 by AeroAce

[6./ZG26_Gielow]

I would like to know when we will have SLI/Crossfire available??

[Jupp]

~S~ Fellows,

 

You know,

 

Setting Affinity for something like your Anti-Virus Firewall to run on one core (4th?), or anything else run at the same time as BoS...

 

Might be something worth bringing up. I read through a few articles about this, and that was my big ah-ha thought of the moment.

 

http://www.tech-recipes.com/rx/37272/set-a-programs-affinity-in-windows-7-for-better-performance/

[Dakpilot]

SLi has been included in Nvidia driver as a profile for a while as I understand

 

Cheers Dakpilot

[AndyHill]

I haven't tested with BoS but I don't know why it would be limited to 3 cores. I think I once got RoF to go 98% utilization on 4 cores (a quick mission with lots of planes), I've never seen a game do that and I was quite impressed. You can specifically ask the OS to run threads on certain cores, but I don't know why you'd want to do that, usually you just spawn threads and let the OS run them. Of course you get diminishing returns with higher numbers of cores depending on the nature of the calculations you're doing, but things like flight modeling might actually scale quite well.

 

Does anyone have a mission with lots of planes in it handy? I would be interested in testing BoS scalability as well, but the QMB doesn't seem to be as flexible as it is in RoF so I can't just pump more planes into it until I run out of CPU.

[coconut]

@bearcat What you want to do is make sure you get as much sequential performance out of your system as possible. Making the game better use multi-cores is something only the devs can have a significant impact on.

 

At the level of the OS, disable background programs you don't need while playing. Make also sure that your power profile is set to "High Performance".

 

At the level of your hardware, make sure your RAM is well configured (running at the proper frequency, with proper delays). If your CPU supports hyper-threading, you can try disabling it (in the BIOS). In theory that might give improved performance due to avoiding having two threads share the L1 cache on the same physical core, but in practice I've not noticed any improvement when I disabled HT.

[402FOX]

I would like to know when we will have SLI/Crossfire available??

Crossfire is working for me, has been for months!

[Zettman]

Crossfire is working for me, has been for months!

 

Well there is a difference between 'working' and working. YSeah the game doesn't crash while using more than one GPU but you do not really see any improvements by doing so. Here are some test I did with no SLI vs 3-way SLI.

 

 

 

 

S!

 

Ok here are the test I made.

 

First round, climbing from 6000m to 7000m and going down from there at about 400km/h until ground level. All tests done in Quick Missions.

 

First screen with 3-way SLI (all 3 GPU show the same usage)

 

 

 

 

 

 

 Second screen with SLI turned off (only one GPU has usage bigger 0%)

 

 

 

The pictures show that what I was exspecting. GPU usage is high at the beginning and goes down a bit at groundlevel. Also what ever thread is running on CPU3, seems to be the rendering thread, cause it has the highest overall usage and goes up when I go closer to the ground.

Honestly I was exspecting something similar with your system, but your screens don't show it. So I have no idea what it could be.

 

Another thing I noticed is, that while using only one GPU, I got better overall FPS. About 80 FPS with one GPU at groundlevel vs 60 FPS with 3-way SLI. Will post this somewhere as a bugreport, cause it seems to me we got negative GPU scaling all over again.

I did a second test where I dived at a higher speed and directly from 6000m down to groundlevel.

 

First screen with 3-way SLI

 

 

 

Second screen without SLI

 

 

 

They both show the same results. The GPU usage going down at lower altitudes is a bit easier to see in them. Still getting higher FPS with only one GPU. So it seems that negative GPU scaling is really there.

 

 

Zettman

[LLv44_Mprhead]

I asked about multicore optimization quite a long time ago. Then the answer was that game uses 2 cores, Operating system one, so you would get quite a performance boost going up from dual core to quad core, but no benefit from more cores than that. I have read that dx12 is supposed to bring big boost in performance for 6 and 8 core processors, at least in some titles, but if BoS is going to get support for that, I don't have a clue. And wheather it would bring any benefits in BoS. Maybe one of the devs could share a bit of light about the issue?

 

(Disclaimer: I don't have very deep understanding of software side of things, so I might be totally mistaken here)

[361fundahl]

Yeah my SlI 680s work great. I'm not sure if it is taking full advantage of my i7 3770k overclocked to 4.2ghz per core though

[VBF-12_Snake9]

Same cpu chip. Same over clock. Cpu is still the bottleneck for my system. Gpu 970.

[361fundahl]

Do you have problems though? I haven't had any yet on 48 person server. 32gb ddr3 for ram

I think at 1600hz

[VBF-12_Snake9]

No, but frames are limited by cpu. Frames drop into forties ssometimes when card is below 60 percent. Thinking of getting liquid cooled and pushing it to maybe 4.5?

[Zettman]

No, but frames are limited by cpu. Frames drop into forties ssometimes when card is below 60 percent. Thinking of getting liquid cooled and pushing it to maybe 4.5?

 

Depends on your screen resolution I think. My 4820 is running at 4.6 GHz and can't feed my 3440x1440 Monitor to 60 FPS on the ground. GPUs are at 30-40% usage near ground. SLI is doing nothing except for weakening the CPU performance even further.

 

Zettman

[[KWN]T-oddball]

I asked about multicore optimization quite a long time ago. Then the answer was that game uses 2 cores, Operating system one, so you would get quite a performance boost going up from dual core to quad core, but no benefit from more cores than that. I have read that dx12 is supposed to bring big boost in performance for 6 and 8 core processors, at least in some titles, but if BoS is going to get support for that, I don't have a clue. And wheather it would bring any benefits in BoS. Maybe one of the devs could share a bit of light about the issue?

 

(Disclaimer: I don't have very deep understanding of software side of things, so I might be totally mistaken here)

 

I got  crash course in this during the ARMA3 alpha.

 

 

up vote 0 down vote accepted

1) I am assuming that by multiprocess you mean splitting up one program into multiple processes (which is usually done with the fork() method in POSIX systems) as opposed to multitasking. The usual definition of multiprocessing or multitasking is that of allowing a single CPU to run multiple processes seemingly in parallel by switching between them very often. Multicore means that a single CPU can have more than one core capable of executing tasks and has all execution units connected in some way (depending on the architecture of the processor). These things usually have no influence on how you program, unless you are writing very low level software, such as an operating system. Forking processes (which I believe you mean) is usually done through some method provided by the operating system and is actually huge topic, which you can find a lot of books about.

2) Operating systems often provide abstraction to the underlying CPU architecture, meaning that you as the programer do not have to concern yourself with whether you PC does have a multicore CPU or a single core, so if you use multithreading you still have to share your data just the same way. If you are writing the operating system for a multicore CPU then things become way harder and you will have to read up on the instructions which the CPU supports.

3) If your program only has a single process (and spawns no threads) then execution will be (nearly) identical no matter whether it is run on a single core or a multicore system. The only speed benefit it could gain on a multicore system could occur from the operating system pushing it onto a core on which not many other tasks are running, thus giving it more CPU time with fewer task switches. A real gain from using multicore systems can only be achieved by using parallel programming.

http://stackoverflow.com/questions/19244452/multicore-programming-tips-and-techniques-required

 

 

Parallel computing is a form of computation in which many calculations are carried out simultaneously,^[1] operating on the principle that large problems can often be divided into smaller ones, which are then solved concurrently ("in parallel"). There are several different forms of parallel computing: bit-level, instruction level, data, and task parallelism. Parallelism has been employed for many years, mainly in high-performance computing, but interest in it has grown lately due to the physical constraints preventing frequency scaling.^[2] As power consumption (and consequently heat generation) by computers has become a concern in recent years,^[3] parallel computing has become the dominant paradigm in computer architecture, mainly in the form of multi-core processors.^[4]

http://en.wikipedia.org/wiki/Parallel_computing

 

Amdahl's law

 

The speedup of a program using multiple processors in parallel computing is limited by the time needed for the sequential fraction of the program. For example, if a program needs 20 hours using a single processor core, and a particular portion of the program which takes one hour to execute cannot be parallelized, while the remaining 19 hours (95%) of execution time can be parallelized, then regardless of how many processors are devoted to a parallelized execution of this program, the minimum execution time cannot be less than that critical one hour. Hence the speedup is limited to at most 20×.

http://en.wikipedia.org/wiki/Amdahl%27s_law

 

there is much more information to be found but be warned the reading is so dry you may start drooling on yourself

[-TBC-AeroAce]

Just to let people know, I recently upgraded a 770 to a 970 and ran some tests.

 

I have a fx 8350 (default), 8gb Kingston hyper x 1600, gtx 970 Galax exoc, 750w corsair, Asus (mid range) mb, corsair h110 cooler, Kingston 120gb ssd (where BOS install is) and cooler master hafx (modified to fit the h110)

 

I tested the game in ultra full aa, in mp and sp in a variety of missions, monitoring, fps, GPU+CPU+ram usage.

 

First off there is no real max fps gain noticed at all between the 770 and 790 and min fps only increased slightly. It on average is around 40 fps, often falling to 30 and in some cases such as with lots of players, ai or especially with special effects such as spot light or fire it will drop as low as 15 to 20 fps.

 

GPU usage on average is 50% riseing to a max of around 85% when low to the ground, zoomed or when looking at lots of objects.and the vram was not really close to being used.

 

CPU usage did only seem to be on three cores and was never higher on a single core than 80%.

 

All other games I played had a significant performance boost such as far cry 3 that went from 25 fps maxed out to 60 fps.

 

My conclusion (nothing unexpected) is that BOS performance is heavily based on cpu and that doing what was a big gpu upgrade for most other games had unnoticeable results.

 

Hope this was interesting I'm now kinda bummed out that I will have to probably have to get a new mobo and Intel cpu to get the full experience

 

Ow and this was at 1920 1200 60 hz