Performance survey.

General discussion related to the game goes here.

Moderators: Balthagor, Legend, Moderators

ZedPoirier
Private
Posts: 1
Joined: Jun 08 2020
Human: Yes

Re: Performance survey.

Post by ZedPoirier »

First Test using 9.2.3

Data captured using phone stopwatch:
Day 1 = 30.67sec
Sept 6th = 4min 40.63sec
Avg = 28.063sec/day

System Info:
CPU: i5 -7500 @3.4GHz
GPU: RTX 2060
RAM: 16GB
OS: Win 10
lane g17
Private
Posts: 1
Joined: Jun 08 2020
Human: Yes

Re: Performance survey.

Post by lane g17 »

Version 9.2.4
High settings zoomed in on a coast.


OS: Windows 10
CPU: Intel I5-6500 3.2ghz (4 physical cores)
RAM: 16GB
GPU: GTX 1070


Day One: 31s
September 6th: 4:21s
Average: 24s
StephaneDurette
BattleGoat Team
Posts: 2
Joined: May 01 2020
Human: Yes

Re: Performance survey.

Post by StephaneDurette »

1680 x 1050 Graphics Resolution HIGH Fullscreen
Fully Zoomed out over North America

CPU: AMD Ryzen 5 3600X@3.8GHz
GPU: AMD Radeon RX 5700
RAM: 16.0GB DDR4 3600MHz
OS: Win 10

Using build 9.2.4: 1 test
First day, 17s
Average of the first 10 days, 21.5s
Rosalis
Lt. Colonel
Posts: 213
Joined: Sep 07 2019
Human: Yes

Re: Performance survey.

Post by Rosalis »

I5-4460 3.2 GHZ
16GB RAM
GTX1050

Test1 zoomed in on southpole
First day: 25.6
september 6: 3.55.97
average: 21.3 seconds per day

Test2 max zoomed out 1280,1024 resolution
september 6: 3.49.91
average: 20.99 seconds

I expected alot worse. As soon i zoom in on UK the planes are going slowly from pixel to pixel tho.

Whats the conclusion from this? :roll:
If you want to buy a new computer 32 GB ram is your best investment? I see quite big difference between 16 and 32 GB ram, but they might just be newer computers? Anyway if you claim the game is 12 seconds or 15 seconds, might wanna update the system requirements.
Nerei
General
Posts: 1113
Joined: Jan 11 2016
Human: Yes

Re: Performance survey.

Post by Nerei »

32GB RAM is indicative of those computers having high end components. That is what makes the game run faster not the RAM itself although naturally a higher clock speed on the RAM is better and likely they also have that.

The game last I checked was 32bit. That means it physically cannot use more than 4GB RAM. Try and use more and it runs out of address space, trows an OOM error and crashes.

If I where to guess at where money are best spend them I would say a a CPU with high single core performance. The game is multithreaded but it will not spread evenly across all cores. From my experience the bottle-neck for the game is it maxing out a single core and the rest waiting for it to play catch-up.
Basically a CPU that will get as much performance as possible on say the first 8 cores.

Frame-rate issues are likely also down to a CPU bottleneck. A modern graphics card can render the 30-50M polygons that 30.000 unit models have. The DX9 graphics API however have a snowballs chance in hell of making sufficient draw calls to keep a high, stable framerate with that many unit models milling about. I have personally yet to see the game max my RTX card despite having had framerate issues in the past. That to me points towards the graphics card being less critical than the CPU.
m04sWYEGzS
Corporal
Posts: 6
Joined: Jun 19 2020
Human: Yes

Re: Performance survey.

Post by m04sWYEGzS »

Threadripper 3970X
64gb 3600mhz CL-14-14-14-34
Sabrent 2TB Rocket Nvme PCIe 4.0
GTX 1060 6GB
1080p, Borderless, Max Settings, Max Game Speed
Version 9.2.4

Max zoom out Atlantik:
1st Day : 18,5 seconds
10 Days Avg. : 13,43 seconds

Max zoom in X156 Y755:
1st Day : 17,3 seconds
10 Days Avg. : 13,53 seconds

New Game, Sandbox, 1914 Ethiopia, max zoom in X1096 Y375
1st Day : 2,5 seconds
10 Days Avg. : 2,523 seconds

Would definitely be better if the game could use more RAM and threads...
Last edited by m04sWYEGzS on Jun 20 2020, edited 1 time in total.
User avatar
Balthagor
Supreme Ruler
Posts: 20504
Joined: Jun 04 2002
Human: Yes
Location: BattleGoat Studios

Re: Performance survey.

Post by Balthagor »

The game is multi threaded and we use as much ram as is available to a 32 bit application.

What is that last number, 2,523? is that 2mins 52 seconds? You're first day and 2.5 secs is the fastest I've seen.
Chris Latour
BattleGoat Studios
chris@battlegoat.com
evildari
Colonel
Posts: 492
Joined: Aug 10 2017
Human: Yes

Re: Performance survey.

Post by evildari »

New Game, Sandbox , 1914 (Great War - Not Brinkmanship) Ethiopia, max zoom in X1096 Y375
1st day almost 1.9 seconds (thats right less than 2 seconds!)
10 days rounded up 19 seconds total...
average would be 1.9 seconds
his number of 2,523 seconds per day average seems ok (i guess i got lucky with my system for this game)
but thats the "peaceful" sandbox if i remember right
my mods
http://www.bgforums.com/forums/viewtopi ... 79&t=25932 (even techs and units for everyone - AI will own you too)
http://www.bgforums.com/forums/viewtopi ... 79&t=29326 (MARSX2)
m04sWYEGzS
Corporal
Posts: 6
Joined: Jun 19 2020
Human: Yes

Re: Performance survey.

Post by m04sWYEGzS »

Balthagor wrote:
Jun 20 2020
The game is multi threaded and we use as much ram as is available to a 32 bit application.
I know but the max RAM for 32bit is 4gb, not much...
And I wouldn't call this good multi threadding (Max Zoom out Atlantik and I was also converting images while I took the screenshot...):
                                          Current | Minimum | Maximum       | Avarage
Image
Balthagor wrote:
Jun 20 2020
What is that last number, 2,523? is that 2mins 52 seconds? You're first day and 2.5 secs is the fastest I've seen.
2,523 seconds. I haven't truncated the 25,23 seconds for 10 Days after dividing by 10.

In general the game is awesome and I really hope there will be a successor with 64 bit architecture, a real 3D planet as a map (not some distorted 2D map which resizes Russia to 3x it's original land mass...) and more political optioan eg. proghibit Religon types, languages, etc.
User avatar
Balthagor
Supreme Ruler
Posts: 20504
Joined: Jun 04 2002
Human: Yes
Location: BattleGoat Studios

Re: Performance survey.

Post by Balthagor »

A spherical planet won't happen any time soon, that's a massive redesign of the whole engine. We are going with 64 bit for our next titles. If you have ideas for the next Supreme Ruler game, feel free to add those to the thread that was started for Next Gen;
viewtopic.php?f=78&t=31379

2523 seconds is 42 mins. Are you saying it took 42 mins to complete day days? That' can't be right.

And the multi threading isn't specifically bad/good, it's that there are some calculations that are bottlenecks, one calculation ends up being needed before many other threads can start, so the game only goes as fast as the most critical thread. That will always be the case in a game as complex as this.
Chris Latour
BattleGoat Studios
chris@battlegoat.com
evildari
Colonel
Posts: 492
Joined: Aug 10 2017
Human: Yes

Re: Performance survey.

Post by evildari »

Balthagor wrote:
Jun 20 2020
2523 seconds is 42 mins. Are you saying it took 42 mins to complete day days? That' can't be right.
thousand delimiter issue or rather decimal point
2,523 in one region = 2.523 = 2523 in other regions! (thousand delimiter)
2,5 = 2.5 (decimal)
my mods
http://www.bgforums.com/forums/viewtopi ... 79&t=25932 (even techs and units for everyone - AI will own you too)
http://www.bgforums.com/forums/viewtopi ... 79&t=29326 (MARSX2)
User avatar
Balthagor
Supreme Ruler
Posts: 20504
Joined: Jun 04 2002
Human: Yes
Location: BattleGoat Studios

Re: Performance survey.

Post by Balthagor »

Oh, I see it now, thanks.

2.5 sec/day is well above our target speed, I don't see any reason we would try for faster than that. I can see us trying to make late game closer, under 10sec/day all the time would be great, but not sure if we'll ever see that until computer hardware gets much more powerful.
Chris Latour
BattleGoat Studios
chris@battlegoat.com
m04sWYEGzS
Corporal
Posts: 6
Joined: Jun 19 2020
Human: Yes

Re: Performance survey.

Post by m04sWYEGzS »

Balthagor wrote:
Jun 20 2020
2.5 sec/day is well above our target speed, I don't see any reason we would try for faster than that. I can see us trying to make late game closer, under 10sec/day all the time would be great, but not sure if we'll ever see that until computer hardware gets much more powerful.
The fastest speed setting should be something like a skip ahead function. It's not a speed you play in, it's the speed you set, if you don't want to play and have to wait a couple of days. Therefore the fastest game speed should take around a second per day or even less.
User avatar
Balthagor
Supreme Ruler
Posts: 20504
Joined: Jun 04 2002
Human: Yes
Location: BattleGoat Studios

Re: Performance survey.

Post by Balthagor »

I can't imagine any way we could "skip ahead". Even if your region isn't actively doing a lot of stuff, all the other countries are doing stuff, maybe even wars. Our goal is for players to have something worth doing "every game day". The only way we get to 1sec/day is to reduce game complexity, which is at the heart of what makes the game so engaging.
Chris Latour
BattleGoat Studios
chris@battlegoat.com
Rosalis
Lt. Colonel
Posts: 213
Joined: Sep 07 2019
Human: Yes

Re: Performance survey.

Post by Rosalis »

Excactly, i wonder how many topics were in SR2020 about speed. It didnt mather cause you were always busy expanding your empire and make sure you had enough defence to counter attacks. Now it does mather cause you can just annex regions 1 by 1 where some ai have no problem keep declaring well over +10 wars. Meanwhile you have these regions who will do nothing but produce outdated equipment. Entrenching a couple of units in border cities is often all you need to stop them. You can easily see what can comes to you. If you have nothing to do, then make "peace" in the world and declare on all the regions that are agressive.

To improve speed i wouldnt think about removing features or whatever, more in reducing different ammount of units and models. It isnt for all, but thats why options are invented. Or just some simple tricks, that the player might not see immediately. Why give 8 garrisons in a city spotting, when a city, IC or MC could do the same thing. Or a more simple map mode, where all the useless facilities in siberia are removed and the usefull improved by supply, so the AI doesnt have to send as much units to reach their destination. Those are some of the things i did for my mod atleast. No point in having WW2 artillery in 2020, some famous and good ones can stay, but after a while they should be spread to other regions as well which dont have any artillery design then. If there is competition regions gonna loose alot of troops, if there is not the problem will only grow.
This save is actually a descent one, it could be worse only if South Korea owned Japans area in Siberia. They dont have to cross water from mainland. US got involved, which you rarely see, so yeah. Atleast there are troop losses.
Post Reply

Return to “General Discussion - SRUltimate”