Lag with 5000+ troops?

[Intranetusa]

What exactly causes lag when you have a ton of troops?

I usually start noticing lag with ~6000 troops on normal speed, around 4000-5000 on 2x speed, and around 4000 on 3x speed.

I once fought a battle where both sides had only around 2000 troops, but when all the soldiers clumped onto the bride, the game started to lag horribly even at regular speed.

My system is an i7 Quad (CPU usage is less than 25%), 6GB of RAM (RAM usage is less than 40%), 4870 1GB (overkill for RTW), and I installed RTW/EB on my SSD, so I know it's not a slow HDD access problem.

So why does the game still lag? Engine/coding limitations? O_o

[Skullheadhq]

Same here, do you use Windows 7?

[Tellos Athenaios]

Oh the irony: your i7 is hardly used but it is precisely your i7 that is the bottleneck. You see, RTW is hardly multi-threaded; and during the battle both the engine and the logic (AI) must share a thread. Since the load on the system increases as the variables in the scenario (i.e. objects, and complexity of the physics of those objects) do you will quickly max out one core and simply run an underpowered single processor system as far as RTW is concerned from then on. Hence only 25% CPU load, but a lagging game nonetheless.

You may be able to run RTW/EB a bit smoother if you look into setting both CPU affinity of the cleanup and the RTW main processes/threads via the task manager (Processes tab).

[Skullheadhq]

So, this is for all quad-cores and multi threaded thingies or only for the i7?

[Tellos Athenaios]

More than 2 cores in general. RTW consists of exactly 2 processes: (a) RTW.exe; (b) e~005 or something which appears to be a kind of runtime or garbage collector.

So it is better for RTW to have 2 cores at higher clock speeds than 4 cores at lower speeds because it is not going to put more than 2 cores to any good use. So a high end Core2Duo is going to at least match if not outperform any Quad Core on this game (especially since the cheap Quad Cores sacrifice on the cache size which means that you are going to waste more clock cycles fetching memory from slower memory, namely RAM).

And some Quad Cores are more Quad Core than others.

[Skullheadhq]

So, does adding 2 additional gigs of RAM help, since I've got only 2?

[Intranetusa]

Same here, do you use Windows 7?

Yep.

Oh the irony: your i7 is hardly used but it is precisely your i7 that is the bottleneck. You see, RTW is hardly multi-threaded; and during the battle both the engine and the logic (AI) must share a thread. Since the load on the system increases as the variables in the scenario (i.e. objects, and complexity of the physics of those objects) do you will quickly max out one core and simply run an underpowered single processor system as far as RTW is concerned from then on. Hence only 25% CPU load, but a lagging game nonetheless.

You may be able to run RTW/EB a bit smoother if you look into setting both CPU affinity of the cleanup and the RTW main processes/threads via the task manager (Processes tab).

But the thing is, in my task manager, I see the 2 cores being used at less than 50%, so even if RTW only uses 2 cores, it's not using my cores to its potential. Furthermore, an i7 is faster clock-per-clock than Core2s due to better architecture, and it has turbo boost that increases the speed from 2.66GHz to 2.GHz when 2 threads are being used and to 2.9GHz when 1 thread is being used.

In either case, I see my CPU usage for the cores actually being used at less than 50%

[Tellos Athenaios]

No. Not relevant. With RTW you are running fairly generic x86 code that is not optimized for i7. You are then running this within a compatibility layer (VM-lite) on what is presumably a generic AMD64 OS. Perhaps some SSE support is required to run this game I don't know (would have to look at the specs) but you are getting no more than SSE 2 as far as optimizations go I think. Again: you are running Pentium IV/equivalent code that is basically the architectures of many years ago. You are not even getting the benefit of improved number crunching performance of the AMD64 architecture, because you are not even using that.

A few percent clock for clock gain is nullified by (a) not using it; (b) the intermediary compatibility layers; (c) the significantly higher clock speed of a high end Dual Core out-of-the-box. And that is without considering cache misses. What is interesting is to see if your set up would benefit from (1) coercing Windows into lumping these processes onto a specific core; (2) looking at the actual clock speed as reported by Windows; (3) looking into this Turbo Boost feature and see whether Intel ships utilities so you can manually enable/disable it.

In other words: that there is this new great feature on the i7 that makes code run 5 times as fast, bakes cookies and brings happiness to the world is not of any use unless you actually run a program that takes advantage of it. And RTW certainly doesn't.

[Intranetusa]

holy bawlz...that is totally meh. I'll try overclocking my i7 in hopes that increasing the raw power will somewhat make up for the lack of coding support/architecture utilization.

[SwissBarbar]

Windows 7, i7 Quad Processor, 6 GB Ram, Radeon 4890 ----> NO PROBLEM with up to 7000 Troops and most times no lag at all even with 3x speed

[Skullheadhq]

Upgrading RAM works? I've got 2 atm.

[Cullhwch]

This caused a bit of lag. Windows 7 Beta, 4 GB RAM, 2GB VRAM

[Intranetusa]

Peasantry? :D

[Cullhwch]

Nah, there were quite a few units of Triarii and Pedites Extraordinarii. When an army as badass as my own has only 1:3 battle odds it's not fighting peasantry. I just killed all the generals and used the Gaesatae's fear effect to its absolute maximum.

[Arutima]

his troops are three silver gallons + he have a ton of cavalry bum rushing the ennemy generals. The best way to give the Ai a fighting chance is to have maximum 3 cav units (two normal cav + general) and not bum rushing the generals

[Cullhwch]

Hey, I had to blast through 6 full stacks in northern Italy. Giving them a fighting chance was my last priority. You win enormous battles any way you can, especially when you're the attacker.

[Intranetusa]

Hey, I had to blast through 6 full stacks in northern Italy. Giving them a fighting chance was my last priority. You win enormous battles any way you can, especially when you're the attacker.

Wait, I thought your pic was from a custom battle/custom skirmish?

[SwissBarbar]

Wait, I thought your pic was from a custom battle/custom skirmish?

Don't think so, you only have "captains" in custom battles, no FM

[Cute Wolf]

Maybe he modify the EDB to gain missile bonus just like me... (but I only give +1 bonus, not +3 bonus)

[Cullhwch]

Yeah, I've made some edits to my game. I edited the tavern buildings to boost missile attack and I gave Triarii eagles, among other things.

[SwissBarbar]

Uhoh, that's historically not correct ;-)

[Tellos Athenaios]

And there I was, thinking the alcohol reduced accuracy?

[Intranetusa]

My question is...how in the hell was the AI smart enough to build 6 full stacks in the campaign...and how the hell were they smart enough to attack you with all 6 stacks at once!!! O_o

[Tellos Athenaios]

He attacked _them_. So it is no coincidence that this is a 1 vs 4 battle.

[Cullhwch]

My question is...how in the hell was the AI smart enough to build 6 full stacks in the campaign...and how the hell were they smart enough to attack you with all 6 stacks at once!!! O_o

The AI had six full stacks around Arretium because I bottled them up there. You see, if you spend the first few years of an Aedui campaign building forts with token garrisons in the mountain passes of northern Italy, the Romans can't expand into Cisalpine Gaul. If you're not at war with them and the forts are in Eleutheroi territory, the Romans won't attack them (no common border with the Aedui) and they'll HAVE to go to Sicily. It makes the game a lot more interesting.