terrible performance with 4.167 TEST 1

[Badger]

I did some testing and here are my results

2gig 5770 card  8gig ram 4core AMD

[Ssnake]

Aside from the fact that this was a decent card in 2009 and that it's now 2020, the key issue probably is the limited amout of video RAM on this card. So you may want to reduce the road rendering detail further (although IMO averything under 3 looks crap), but above all, reduce the cache size for the terrain itself (because it's all in a texture now).

That would not be found in the Graphics settings (Alt+G) but in the Detail settings (Alt+D); note that you can open both dialogs while you're running a scenario (in single player mode), so you can experiment with the settings at runtime with the frame rate counter active.

[Badger]

I have to ask but did you notice the tests were on different systems?  the second test was on a card thats only 1-2 years old.launched in April 2017. And which do you mean  when you say "reduce the cache size for the terrain"...terrain distance or dynamic cache?

The purpose of the test was to show the difference between two different systems that basically showed the same result.That being a major difference between 4.023 and 4.167.Note that on test-1 reducing the slider from50% to 25% only show a difference in 1 frame per second.

Edited September 4, 2020 by Badger

[Badger]

This is a screenshot of terrain settings set to minimal on the newer system.It's only 6 frames above 24 frames per second.keep in mind also this is the tank range where there's nothing really going on.

Edited September 4, 2020 by Badger

[Jartsev]

1) Can you upload your test scenario?

2) What are system specs?  Sorry, but '8gig ram 4core AMD' is not quite descriptive, and sounds more like some integrated graphics, relying on shared memory.

[Badger]

There is no senerio...its the tank range in the M1a2. System1 is   AMD FX 8350 4.0 Ghz CPU , ATI RX 560 4 gig GPU ,32 gig ram on Win7 64bit,1TB HDD. System2(older)  is Amd quadcore  3.6ghz with 8 gigram,radeon 5770 graphics 2gig card, 500mb HDD Remember this test was on 2 different systems.The point being even on the newer system 4.167 runs like a dog.Where as 4.023 runs like a dream on BOTH old and new systems.

Edited September 4, 2020 by Badger

[Jartsev]

Ok, are you running SB in  full screen mode? If yes, I'll suggest to consider switching to windowed mode, which in, particular case may give you extra 5-10 frames per second.  I'm writing this as a person, whose primary system has GTX1050Ti graphics card, which is more or less equivalent of  RX560.  Other option is to reduce resolution somewhat, but blurry appearance of the image in this case may not be suitable for everyone.

[Badger]

Yes I can turn a lot down in different aspects...BUT...the real question is WHY am I only getting 29 fps in a scenario where nothing is really happening on a  AMD FX 8350 4.0 Ghz CPU , ATI RX 560 4 gig GPU ,32 gig ram on Win7 64bit,1TB HDD. system under 4.167????...on minimal terrain settings???

Edited September 4, 2020 by Badger

[Jartsev]

8 minutes ago, Badger said:

Yes I can turn a lot down in different aspects...BUT ...the real question is WHY am I only getting 29 fps in a scenario where nothing is really happening on a  AMD FX 8350 4.0 Ghz CPU , ATI RX 560 4 gig GPU ,32 gig ram on Win7 64bit,1TB HDD. system under 4.167????

Answer is very simple- you are staring through magnified optics at nearby bunch of trees, and rendering of those trees  causes impact on performance, since as you may notice trees models are not quite the same as they were in 2.5-4.0.

[Badger]

So there's something wrong with the trees then...As I was also staring at them in 4.023...and getting 40-50ish vs 6-29ish, maybe you need to do something about optimizing the trees??

[Jartsev]

As I wrote above- trees have completely new more detailed models, and to render them requires more GPU resources, despite of all possible optimizations

[Badger]

Is it possible to swap the trees out for older models??...better yet make it an option...then we could run 4.167 on older machines...just sayin.

Edited September 4, 2020 by Badger

[Badger]

BTW windowed mode...small change,3fps

Edited September 4, 2020 by Badger

[Badger]

52 minutes ago, Jartsev said:

As I wrote above- trees have completely new more detailed models, and to render them requires more GPU resources, despite of all possible optimizations

Is it the trees??? watch what happens when I look to the left...52fps.BTW thats not a lase to the trees...its the range of a previous tank kill

 

Edited September 4, 2020 by Badger

[Ssnake]

I don't think it's the trees, but it's simple to test that. If you look in the same scenario into a direction where there are no or nearly no trees and the framerate doesn't improve significantly (like, double, not just one or two frames), then it's clear that it's not the trees.

 

With the very low frame rates I'm relatively confident that the terrain cach size is the culprit, as it may exceed the amount of video RAM that you have on your sdystem, so the graphics card may use regular (system) RAM which is much, much slower. The cache size must be adjusted to match the amount of video RAM available. 

4 hours ago, Badger said:

And which do you mean  when you say "reduce the cache size for the terrain"...terrain distance or dynamic cache?

The dynamic cache size.

[Badger]

All sliders were minimized too far left in those last few screens...so 4gig of graphic card ram isn't enough?That doesn't seem right.

In test 2...those screens were not minimized (cache was set to full)and the difference is only a few frames,6 to be exact(the above shot on the target panel is more as its windowed). Plus the direct above shot is trees also and 52 fps.Doesn't make sense.

Edited September 4, 2020 by Badger

[Ssnake]

Well, we're trying to find out what's going on. So we need to test different things, and this appeared to be the most plausible explanation. 4GBytes of video memory should be good for relatively generous settings. That you're writing about "test 1" and "test 2" and "system 1" and "system 2" isn't exactly helping with clarity, especially when opening two different threads. So let's try and be a bit more systematical in the way we try to get through this. I fully agree that at least on your modern machine the overall performance should be better and I'm confident that in the end we'll find out what causes the trouble.

 

Since you already minimized all your settings and managed to get what, 30, 31 frames per second on your new machine (?) in a very simple scenario suggests that maybe it's not the Steel Beasts settings as such that are the problem. Version 4.0 was comparatively inefficient in its rendering; 4.1 is much better tailored to take advantage of your new graphics card's capabilities. So, maybe you're running some video related 3rd party software in the background? Or could it be that your system applies thermal throttling to the CPU? Switching to Windowed mode will allow you to have the task manager running with the performance tab selected while SB Pro would run in the background. This might reveal something.

 

Maybe you can find a new driver, maybe we should look at the driver settings to see if there is something conflicting with how Steel Beasts handles things.

[Badger]

I do run trackIr...gpu is around 70-77 Celsius which is in normal ranges,drivers are up to date.I will try it without trackIR. However I also run DCS at high settings and get 30-40 in cockpit higher in external...which says alot. BTW...external view in sb is alot lower than test screenshots FYI which adds insult to injury,

[Badger]

EVERYTHING  turned down trackIR off at 1680 x 1050 on tank range target panel 29 fps.Gonna try different resolution now

opps wrong res...corrected

Edited September 4, 2020 by Badger

[Badger]

going down to 1280 res made it worse 19-20fps but 1100 res hit high 50's

Edited September 4, 2020 by Badger

[Badger]

Tried to override settings in radeon control panel AA,Ansio etc....I "upped everything" so it should have dropped from 29fps...It didn't

[IamKirk]

Just throwing in some of my experience, in case it helps...

 

My system:

Acer PC, Intel i-5 Generation 8 with 6 cores and with max CPU clock of 3.9Ghz, 32GB RAM, with an Nvidia GeForce GTX 1080 SC with 8GB GDDR5X.

Acer montior, 27" diagonal 16x9 aspect ratio run at 1600x900, on account my eyes not as good as they used to be, but the monitor will support a higher res.

 

Performance:

I get 29-30 fps when I have vsync turned on, and 59-62 fps with it turned off.  Thus I suspect my monitor, since it is 7 years old, is only 30 fps.

I get these performance specs regardless of what I am looking at.  However, sometimes it will dip down to 40-50 fps and game play will bog a little when I first-time encounter many other units on the screen during Internet play, but it is momentary like it is caching up something.

SB never uses more than like 7.9GB or RAM since I have 32GB.

 

O/S Config:

Windows 10 Home 64bit.

I do not run anything not necessary when running SB.

Also, for Internet play, I go into the advanced power settings and set it for "High performance" instead of "Balanced".  This runs the CPU at max speed of 3.9Ghz, otherwise it is constantly fluctuating based upon what it perceives as the load.  "Balanced" is fine for single player.

 

SB Config:

My SB graphics settings are in the attached pic.

I run windowed with a screen resolution of 1600x900.

 

I am satisfied with this performance for the hardware that I have.

 

Before I got the GTX 1080, and upgraded from 8GB to 32GB of RAM, I was using the Intel HD 630 integrated graphics and it was adequate for single player with every graphic setting turned way down, with a fps of 20-30 fluctuating.  For multiplayer though it struggled.  The fps was around 15-30 with high fluctuation, and frequent stuttering.  Playable, but not fun.

 

Suggestions:

Set your Windows advanced power settings to "High performance".

 

There may be some tweaks you can adjust in your video card settings/drivers that can improve the performance, but I cannot tell you exactly what, so you will have to investigate those yourself.  Anti-aliasing and shadows and detailed tracks can be a large performance hit.  I have tweaked the card settings in its config software other applications and the difference was night and day in terms of performance.  For my SB setup though, I configured it for "Let the software decide" and I do not know how I could improve beyond what I already have.

 

Do not run any other applications that are not required for SB to run.  For multiplayer, I run the voice software on another computer.

 

There are some videos on youtube that can help with windows performance settings for game play.  Just do a youtube search for "windows 10 for gaming".  This is what I did to help configure my windows 10 home 64bit and it helped.

 

[Badger]

I'm Windows 7 and I'm now going to try those power settings ,it was set to balanced

[Badger]

Power settings no change BUT with vsync off AND in windowed mode putting the sight on the target panel in tank range bounced it to around 35-37 ( it fluctuated) please note panning anywhere else in tank range would go as high as 66fps...its just I wanted to use the same aiming point (the target panel )for reference and its where it falls to half FPS.

 

edit...my mistake...power settings didn't take...retried and got to 41fps. So it did something.

So I think I'm at 40 ish on that panel now.gonna do more tests

edit 2

DAMNIT NO IT DIDN'T WORK ,TRIED IT BY ITSELF STILL 29 so its the vsync and window combo that got me to 40

edit 3

oh fer fuck sake...it was the combo of all 3 that gets me to 40...confirmed.

 

Edited September 4, 2020 by Badger

[Badger]

Just to prove I Ain't lie'n  vsync off and windowed with power to performance combo that got me to 40...and this is WITH graphics and terrain settings turned back up including terrain cache 

At 1680 x 1050 and with aa and ansio and shadows.TrackIR on.GPU Stayed cool 66 Celsius

Edited September 4, 2020 by Badger