Testing HOI late game slowdowns

bholstege · Jun 9, 2008

I'm running my e2140 at the same clock speed as an e8400, and its still painful in the mid 50's. I have yet to play a game to the end date.

unmerged(103753) · Jun 9, 2008

the ai aggressiveness may be a problem as the ai thinks more i bet and thus slows down the game. try modding middle america into 1 country. maybe fold chile and uraguay (or whatever that 1 province country is) into argentina. maybe put bolivia and the minor next to it into brazil? in asia make china into 1 country. thats all the minor folding i can think of. perhaps a swiss merging with Germany after a successful barbarossa event. then merge new zealand with australia. about all i can think of for minor merges. also, as Germany, make only ships after barbarossa and maybe a few strats (to carry ur nukes) and v2a10s. u wont need them anyways. careful about spamming garrisons and if japan lost to china (SEEN IT HAPPEN IN ARMA 1.1 NO FUCKING JOKE. I had to go in there and got into a fucking stalemate and japan was kicked from the alliance through event. then i had a blitzkrieg in asia...which worked and they offered decent terms...after they were divided in half and nuke 3 times)

bbasgen · Jun 10, 2008

Okay, time for some help with your empircal data bholstege. I have an AMD Athlon 64 3500+, with 2GB DDR400 RAM on a WD Raptor 10,000 RPM SATA hdd. From what I can find in terms of benchmarks, your CPU is somewhere around three to four times faster than mine.

HOI2 ARMA 1.2 with GIF installed, playing as Tibet for the month of October, Doomsday on N/N with counters on:

3 minutes, 28 seconds.

That is 21 seconds longer than your results. So, your CPU which is 3 years newer than mine and around 300-400% faster, provides an 8% faster result in HOI2. That 8% is within your original margin of error, but I only think we can establish a good margin of error with other machine results. The MythBusters might think that means this myth is busted.

In any event, folks who think increased CPU speed makes a tangible difference -- post your results! We can put this age-old HOI2 debate to rest, and I promise to post the final results in the Wiki!

P.S. Just to be clear; obviously CPU speed maters if we are talking about a 486 versus a Core 2 Duo playing HOI2. No one disputes that. The point is whether or not a certain CPU threshold is reached, after which significant increases in CPU power are required to achieve increasingly small gains. Many people have made this hypothesis, and if this hasn't been done before, I think we can do everyone a service by running this simple test to help support or disprove the hypothesis.

bholstege · Jun 10, 2008

I will also try under clocking my CPU this weekend, to see if that has any effect.

6354201 · Jun 10, 2008

bholstege said:
I'm running my e2140 at the same clock speed as an e8400, and its still painful in the mid 50's. I have yet to play a game to the end date.

I guess it depends on what you define as painful, but if you are unable to finish the game with a e2140 you either don't have enough RAM or you have too much stuff running the background. Try opening task manager (CTRL ALT DEL) and seeing how many processes you have running. I usually have it around 30, if its greater than 35 you should try to end some unnecessary ones (AIM, security programs, etc.) at least for testing purposes and see what happens.

sbr · Jun 10, 2008

I ran the same test with Tibet and it took 2:26.

Couldn't tell you exactly what CPU I have though. I think it is pretty good.

Tanesis · Jun 10, 2008

How about testing music. I've not tested myself but I think that effects game speed, as it seems whenever I remove the music folder the game goes quicker.

Tanesis,

Lennartos · Jun 10, 2008

A interesting test.

i would think that the slowdown is AI related and the amounty of units and provinces it has to manage.
As such the RAM IO speed + latency would be very important in conjunction with the cpu speed. (after a while cpu speed wont help, but is limited by RAM acess / cache size)

Would be fun to find a link between sisandra benchmarks and the slwodown.
Then everybody can find out what performance they should expect.

I have made a test on my q9450@3.72Ghz, 4GB RAM...
Doomsday scenario, N/N , with sprites, on fastest speed(counters dont seem to speed up)
1rst run: 1min and 20 sec
2nd run: 1min and 30 sec

bbasgen · Jun 10, 2008

Lennartos said:
.
I have made a test on my q9450@3.72Ghz, 4GB RAM...
Doomsday scenario, N/N , with sprites, on fastest speed(counters dont seem to speed up)
1rst run: 1min and 20 sec
2nd run: 1min and 30 sec

Wow! Those are impressive results. You are getting around 3 seconds per day! What kind of RAM? Double the ram is interesting... can you check how much your HOI2 was using?

This makes me pretty curious about bholstege's results, since his CPU is reasonably comparable to yours. It is nice to have 3 data points now, but we would be much better off if we got input from more people!

bbasgen · Jun 10, 2008

sbr said:
I ran the same test with Tibet and it took 2:26.

Couldn't tell you exactly what CPU I have though. I think it is pretty good.

Right click on "My Computer", then select properties. That window will show you your CPU. Let us know what it is, and how much RAM you have.

Lennartos · Jun 10, 2008

By the way, i am running WinXP

My current Ram usage is around 1.1GB (with HOI2 running)
1GB is around the minimum demand for HOI2 in a winXP environment.(with no speed bumbs along the way)
More than 2GB RAM will not(should not) net you any performance in a normal XP system.

Vista demands a whole lot more ram, but will also decrease repeated loading time dramatically the more RAM you give it acess to.

Ram is dual channel DDR2 running at 465Mhz. (some hardware verndors *2 to compensate for DDRs advantage over SDR, so it will be 930Mhz)

unmerged(73478) · Jun 10, 2008

Amazing!!!!

Lord Finnish said:

Code:

# Kill AI units

event = {
	id = 888880
	random = no
	persistent = yes

	name = "AI Unit terminate"
	desc = "You have been.. TERMINATED."
	style = 0
	picture = "VIC"

	action_a = {
		name = "OK"
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
		command = { type = delete_unit which = -1 }
	}
}

Just fire this to a random country, in example:

Kills lot of units, fastens the game.

I've tried this event in my current game and it's unreal! I've deleted all the units from minor countries (not involved in conflict) and my game in 1943 is fast!

QUESTION: Rather than having to type the names of each country, how can this be modified to included the names of the countries? That would make it much better.

THanks for the event.

Lennartos · Jun 10, 2008

Doing some further tests....

It seems that HOI2 is multi core optimized (amazingly)
Doomsday scenaro:
Core 1 = 0% usage
Core 2 = 20% usage
Core 3 = 15 % usage
Core 4 = 45% usage
around ~8Hours / sec (1 month = 90sec)

1936 scenario:
Core 1 = 0% usage
Core 2 = 25% usage
Core 3 = 3 % usage
Core 4 = 10% usage
around ~29 Hours / sec (1Month = 25sec)

So with this basic observation, it looks as if we have 3 threads running for HOI3.
But at the same time it seems as if we are not cpu power limited. (Doomdsday scenario is only using 80% of a singe core)
I will test memory speed vs CPU speed tomorrow.

bholstege · Jun 10, 2008

It seems the q9450 outperforms mine significantly, even though the cores are not maxing out. I also have 4gb of RAM.

Based on this, I think CPU cache may have a large effect as well, since the new quads have tons.

lambrf · Jun 10, 2008

Wouldn't it be more efficient to use the same saved game file to test performance on different configs?

bholstege · Jun 10, 2008

The doomsday scenario is standard. Another thing to look for is that I was running ARMA, which might slow it down.

lambrf · Jun 10, 2008

bholstege said:
The doomsday scenario is standard. Another thing to look for is that I was running ARMA, which might slow it down.

I agree it is standard.
But games is getting slower with time, not in the beginning. Number of countries, of units, history,... all this will have impact on the game. So the data in the save game will impact your calculations and benchmarks. In order to compare performance on different configuration, you need to work on some same data. So using the same save game would bring even more accuracy to your tests.

Another point, I noticed playing DD that lowering the game speed one or twho levels was improving performance. It can be strange, but in the late years, 46, 47 and after, game was going faster in "above normal" or "fast" game speed than in "extermly fast". I'll soon find out if I notice the same with Arma.

bbasgen · Jun 11, 2008

lambrf said:
But games is getting slower with time, not in the beginning. Number of countries, of units, history,... all this will have impact on the game. So the data in the save game will impact your calculations and benchmarks. In order to compare performance on different configuration, you need to work on some same data. So using the same save game would bring even more accuracy to your tests.

I think it is pretty well accepted that the late game slows down simply because of the sheer number of units, and the kill all trick made by Lord Finnish and others over time have substantiated this hypothesis.

The Doomsday scenario is a perfect testing ground: (a) Every test starts in exactly the same way, and the number of variations that occur in the first month are reasonable limited considering the complexity; (b) there are tons of units.

unmerged(66346) · Jun 11, 2008

From wiki : http://www.paradoxian.org/hoi2wiki/index.php/Technical_Issues#Faster_Game_and_Load_Time

Faster Game and Load Time

This change will allow you to both load and play the game much faster, but it only works if you use counters instead of sprite.

1. Open HoI 2 directory\GFX\map\units.
2. Backup this folder and its subfolder just in case you want to play with sprite again later or if anything wrong happens.
3. Delete all files in HoI 2 directory\GFX\map\units except "T-BATTLESHIP A-WALK C-MIN L-1 D-NW.spr".
4. Open HoI 2 directory\GFX\map\units\bmp.
5. Delete all files in HoI 2 directory\GFX\map\units\bmp except "T-BATTLESHIP A-WALK C-MIN L-1 D-NW.bmp".
6. Start the game and feel the speed.

Note that the file to be left alone in those folders do not have to be the one mentioned. Any of the other files will do as long as their file name match.

I tested this recently, the game loads and runs faster, trying it on Armagedon scenarios when a lot of units are engaged made the game running in a pretty decent and stable speed.

bbasgen · Jun 11, 2008

Narakir said:
From wiki : http://www.paradoxian.org/hoi2wiki/index.php/Technical_Issues#Faster_Game_and_Load_Time

I tested this recently, the game loads and runs faster, trying it on Armagedon scenarios when a lot of units are engaged made the game running in a pretty decent and stable speed.

Right, we've got some tried and true fixes (sprites), and we have some tried and true sources for the problem (units) -- but we've never really figured out how hardware comes into the equation.

Testing HOI late game slowdowns

bholstege

Guest

Banned

Field Marshal

bholstege

Guest

General

Field Marshal

My Melancholic Muse

BL-Logic

Field Marshal

Field Marshal

BL-Logic

Second Lieutenant

BL-Logic

bholstege

Guest

Lt. General

bholstege

Guest

Lt. General

Field Marshal

Major

Field Marshal