Optimizing your Experience with Forged Alliance - It works

[-_V_-]

The previous topic was locked because of wtv reason. But this really deserves to be spread.

I'm quoting VoiceOfReason and Kat below on how to try this.

I just watched a replay on setons , big setons, 50 min game+, shit loads of units.

First case : normal setup no tweaking , game took 54 min IRL to reach the 51 min game time.

Second case : I disabled CPU0 on the exe affinity, and set high as priority , game took 46min IRL to reach the same 51 min game time



For setons. Use it.

EDIT : I have a 2600K, so Sandy Bridge if i'm not wrong. Not sure if the CPU architecture actually matters in this.

Setting the following in the task manager yields a +2 sim speed difference in 2k unit tests.

CPU: AMD FX8350

Affinity: Core 0 unselected

Priority: High

I'm 100% positive you misunderstand what I've said, or what is going on.

Yes, CORE Maximizer isn't going to do you much unless you are running Windows XP. What I've said to do, is alter the way Windows decides your FA runs on your CPU.

It's already spreading the resources across all cores/threads for you; but mainly FA hogs up core0, running it to 100% most the time, and putting maybe 3 to 7% load on the rest. When you switch the affinity, and tell FA to run on all cores EXCEPT core0, you'll watch the load on the rest of the cores skyrocket, and core0 becomes mostly free. Whatever is actually happening I know not - but I do recognize results when they slap me in the face.

I believe the post prior to this confirms it is not useless, and far from questionable.

Also, yes - It's something you need to do every time. For short games, it really isn't needed - but if you intend to play a long action packed game, it sucks you have to change the affinity and priority every time - but it can really be worth it :s

[galacticfear]

Hmmmm I will try this then and compare sim speeds at different minutes... not that I am really ever below +0

[Col_Walter_Kurtz]

You might see an improvement if you have a CPU with relatively bad single thread performance such as an AMD octacore and the OS is not managing the processor workload well.

But let's make some things clear. It is technically impossible to divide a thread over multiple cores. Therefore, the all important sim thread will still run on one core and be limited by single threaded performance. After all you can't just set manual affinity and expect your 8 cores @ 4Ghz to perform as a 24Ghz CPU.

Here's an interesting ppt from Intel / GPG devs about threading of the FA engine.

http://twvideo01.ubm-us.net/o1/vault/gd ... 6589i2.ppt

Also this:

Supreme Commander simspeed is dependent on only 1 thread: the sim thread. Once that thread completely saturates a logical core (or more like 85% of a core's capacity due to the way the engine works), the simspeed will slow down. The reason you are seeing only 2 logical cores worked to the max is because there are only 2 main threads in SupCom: the sim thread and the render thread. SupCom FA actually has 22 threads, but the other 20 do very minor tasks that have little in the way of CPU usage. So at best you can see 2 and a half logical cores being used. With Hyperthreading, one physical core will appear as 2 logical cores so what you are seeing with your hyperthreaded i7 looks fine to me.

Before you go around talking trash about how the Moho engine is poorly optimized, remember that the engine was built in 2005. Back then, dual-cores were the new thing. No one expected quad-cored CPUs to be released so quickly. So making a dual-threaded engine where the game simulation and graphical rendering could be processed on separate cores was very innovative for the time.

http://forums.gaspowered.com/viewtopic.php?f=2&t=61250

Not only is this manual affinity method not for fast dualcores suchs as the G3258 which reportedly runs FA at +1 an hour into a Setons game but you might also disturb the OS managing threads on a quad. After all you're telling the OS to use 3 instead of 4 cores for FA decreasing the available resources. It's worth noting these things instead of spreading this method as some kind of general performance increase.

[galacticfear]

Well it doesn't matter what the theory says if it actually works for whatever reason Should just be something to test on a person by person basis, and then they can determine whether its beneficial or not.

[Col_Walter_Kurtz]

Well it doesn't matter what the theory says if it actually works for whatever reason Should just be something to test on a person by person basis, and then they can determine whether its beneficial or not.

That's the point exactly.

[-_V_-]

VODKA, I don't care for the technical stuff. Impossible or not in theory.

I tried if on a huge replay. It made a big difference. I didn't use a simple comparison of sim speed displayed. I use a chronometer and saw that using the "tweaks", the replay ran faster by 8 min , out of 54, which is significant improvement right ?

Have you even tried on a heavy replay ?

[-_V_-]

Hmmmm I will try this then and compare sim speeds at different minutes... not that I am really ever below +0

Im not sure if that would be conclusive. Better try timing a replay like I did.

[galacticfear]

Hmmmm I will try this then and compare sim speeds at different minutes... not that I am really ever below +0

Im not sure if that would be conclusive. Better try timing a replay like I did.

Yeah I will probably just run a replay on +10 with and without those settings, that will tell me for sure if there is any sim improvement over time.

[Sheeo]

ed CPU0 on the exe affinity, and set high as priority , game took 46min IRL to reach the same 51 min game time

You made two changes.

High priority will make a difference as well, so you should perform tests with and without affinity and process priority individually.

For it to really work, you will need to run multiple tests and do normal statistical calculations.

[-_V_-]

I tried with another replay.

Setting both actually makes a difference , compared to setting just HIGH priority but no change to affinity, or simply changing affinity without priority.

Basically use both . The difference was a couple of minutes this time, but I stopped the replay at 25min mark (lack of time :p). My guess is the longer the game gets, the more you gain from those tweaks.Rephrasing since galanoob didn't like my neutral approach : If you play those eco whore bitches, making para shit and all, use it. You won't waste as many mins!

Also this replay was far from being so heavy on air units.

In any case, for a seton that supposed to last, I don't see any good reason not to use this.

Now need to find a tool that automatically sets those settings to the FA.exe