Early Graphics Setting Experiments.
Posted: Mon Mar 11, 2013 11:13 am
Being a notorious adjuster of settings, I spent a couple of happy hours playing with the graphics in A3.
Before going any further, I found SideStrafe's YouTube guide invaluable and provides not only a very sensible way of testing, but also a good set of starting values.
http://m.youtube.com/watch?v=hnmCyP5Ufu4
TBPC Specs: i5 2600K OC'd to 4 GHz / ASUS P8P67 / 16 GB of DDR2 1866 / 2 GB HD7870
Top Tip: Don't try to adjust settings in on of the SP Showcases. Use the Editor as shown in SideStrafe's video to drop your player into an area of mixed terrain and then move around to test FPS in different environments.
Note: I haven't set any Launch parameters or made any changes to the A3 config files.
Catalyst driver: 13.10
No additional graphical enhancements and everything set to "Application Controlled" in Catalyst Control Centre.
Basic Settings:
Resolution: 1920x1080x32 (native res of monitor)
Interface Size: Small (personal preference)
Aspect: 16:9 -Wide
Visibility - This is where A3 starts to need a little tweaking..
Overall: I set this by looking at a middle distance hill through the scope on the rifle and reducing the value from the resource eating 3000 that A3 gives you out of the box until large features start to disappear from the ridge line and then raise it slightly. I find around 1800 works for me.
Object: Links to the above and controls the draw distance of smaller objects on the far hills. I found around 1250 enough to show the smaller rocks etc without the game wasting resources drawing stuff way beyond what you'd focus on during combat.
Shadow: An odd one this as it's very dependant on your rendering settings. I found 80 to look about right in terms of how much shadow you can see in the middle distance.
Rendering:
Your chance to decide on fast, but ugly or slow, but very pretty. And 1 badzillion points in-between.
Resolution: 1920x1080x32 (100%) didn't play with this to see if you can achieve a pseudo-AA effect without switching AA on as per A2.
Vertical Sync: Disabled. Enabling this is supposed to stop image "tearing" when your GPU is generating more FPS than the monitors refresh rate, so above 60 FPS for a 60Hz monitor. What it actually does with my 60Hz 24" Viewsonic 2mS monitor is lock the frame rate to 30 FPS. Not good.
HDR: Standard. Purely a personal preference as I couldn't see any impact on FPS at any HDR setting.
Anisotropic Filtering (AF): Very High. Again, no impact to FPS that I could see and Very High just makes the whole thing look a little more "real" to my old eyes.
PIP (Picture In Picture - images in mirrors etc.): As I didn't drive any vehicles during this test, I left this at High.
Dynamic Lighting: As I didn't test in low-light/night conditions, I left this at High.
Anti-Aliasing (AA): The big frame rate killer. I have this disabled in A2:OA and DayZ as it eats FPS and actually makes it look worse in my opinion.
This is not the case in A3. Setting AA to X2 is basically pressing the "make it look natural" button on my machine and worth the 15-20 FPS hit in complex environments. Setting AA to X4 doesn't to make any additional improvement I could spot and it does eat another 10 FPS.
PPAA - Additional AA added during Post Processing: I found FXAA crisped things up nicely and SMAA actually introduces a little blur on sharp edges such as the accessory rails on the standard BLUFOR rifle. Didn't spot much difference between Disabled and FXAA Standard, but moving to FXAA High has a very nice effect. No obvious frame rate hit either. Very High and Ultra didn't improve things over High as far as I could tell.
ATOC: On my machine, All Trees & Grass is visually well worth the 1-2 FPS it costs as it gets rid of the "grid" effect you see on grass that's lit from behind. No change from how I have it set on A2. Unlike A2, ATOC is disabled when Post Processing is disabled.
Post Process Quality: How much depth of focus blur you see. I find the higher settings make things look somewhat artificial as my eyes don't focus like an SLR camera. Which is just as well as this feature is as hungry for FPS as AF.
I found Low to be the most visually realistic setting and it only costs 2 FPS and enables ATOC.
Quality:
Here's where you start turning things down for easy FPS gains. The Auto Detect feature set most of these to Ultra for my 7870. The weird thing is that I can't see any difference between Ultra and Very on these settings.
Here's what I ended up with:
Texture: Very High - Only a small FPS hit over High, but looks much nicer.
Object: Very High
Terrain: Very High - Turning this down gains performance at the expense of bare hills in the middle distance
Cloud: Only tested during a sunny day, but couldn't see much performance difference between Low and Very High. High looks great, so I left it there.
Shadow: Not the big FPS hit it was in A2. High looked the most realistic on my machine.
Particle: Again, I couldn't see a significant visual or performance difference above High despite throwing smoke grenades around me.
As SideStrafe pointed out in his video, settings chosen during mucking about in the Editor are only an indication of how the game might perform during MP play as there's obviously a very low CPU load using the Editor with just one character model being generated.
Hopefully, this technodrivel might prove useful for others.
Before going any further, I found SideStrafe's YouTube guide invaluable and provides not only a very sensible way of testing, but also a good set of starting values.
http://m.youtube.com/watch?v=hnmCyP5Ufu4
TBPC Specs: i5 2600K OC'd to 4 GHz / ASUS P8P67 / 16 GB of DDR2 1866 / 2 GB HD7870
Top Tip: Don't try to adjust settings in on of the SP Showcases. Use the Editor as shown in SideStrafe's video to drop your player into an area of mixed terrain and then move around to test FPS in different environments.
Note: I haven't set any Launch parameters or made any changes to the A3 config files.
Catalyst driver: 13.10
No additional graphical enhancements and everything set to "Application Controlled" in Catalyst Control Centre.
Basic Settings:
Resolution: 1920x1080x32 (native res of monitor)
Interface Size: Small (personal preference)
Aspect: 16:9 -Wide
Visibility - This is where A3 starts to need a little tweaking..
Overall: I set this by looking at a middle distance hill through the scope on the rifle and reducing the value from the resource eating 3000 that A3 gives you out of the box until large features start to disappear from the ridge line and then raise it slightly. I find around 1800 works for me.
Object: Links to the above and controls the draw distance of smaller objects on the far hills. I found around 1250 enough to show the smaller rocks etc without the game wasting resources drawing stuff way beyond what you'd focus on during combat.
Shadow: An odd one this as it's very dependant on your rendering settings. I found 80 to look about right in terms of how much shadow you can see in the middle distance.
Rendering:
Your chance to decide on fast, but ugly or slow, but very pretty. And 1 badzillion points in-between.
Resolution: 1920x1080x32 (100%) didn't play with this to see if you can achieve a pseudo-AA effect without switching AA on as per A2.
Vertical Sync: Disabled. Enabling this is supposed to stop image "tearing" when your GPU is generating more FPS than the monitors refresh rate, so above 60 FPS for a 60Hz monitor. What it actually does with my 60Hz 24" Viewsonic 2mS monitor is lock the frame rate to 30 FPS. Not good.
HDR: Standard. Purely a personal preference as I couldn't see any impact on FPS at any HDR setting.
Anisotropic Filtering (AF): Very High. Again, no impact to FPS that I could see and Very High just makes the whole thing look a little more "real" to my old eyes.
PIP (Picture In Picture - images in mirrors etc.): As I didn't drive any vehicles during this test, I left this at High.
Dynamic Lighting: As I didn't test in low-light/night conditions, I left this at High.
Anti-Aliasing (AA): The big frame rate killer. I have this disabled in A2:OA and DayZ as it eats FPS and actually makes it look worse in my opinion.
This is not the case in A3. Setting AA to X2 is basically pressing the "make it look natural" button on my machine and worth the 15-20 FPS hit in complex environments. Setting AA to X4 doesn't to make any additional improvement I could spot and it does eat another 10 FPS.
PPAA - Additional AA added during Post Processing: I found FXAA crisped things up nicely and SMAA actually introduces a little blur on sharp edges such as the accessory rails on the standard BLUFOR rifle. Didn't spot much difference between Disabled and FXAA Standard, but moving to FXAA High has a very nice effect. No obvious frame rate hit either. Very High and Ultra didn't improve things over High as far as I could tell.
ATOC: On my machine, All Trees & Grass is visually well worth the 1-2 FPS it costs as it gets rid of the "grid" effect you see on grass that's lit from behind. No change from how I have it set on A2. Unlike A2, ATOC is disabled when Post Processing is disabled.
Post Process Quality: How much depth of focus blur you see. I find the higher settings make things look somewhat artificial as my eyes don't focus like an SLR camera. Which is just as well as this feature is as hungry for FPS as AF.
I found Low to be the most visually realistic setting and it only costs 2 FPS and enables ATOC.
Quality:
Here's where you start turning things down for easy FPS gains. The Auto Detect feature set most of these to Ultra for my 7870. The weird thing is that I can't see any difference between Ultra and Very on these settings.
Here's what I ended up with:
Texture: Very High - Only a small FPS hit over High, but looks much nicer.
Object: Very High
Terrain: Very High - Turning this down gains performance at the expense of bare hills in the middle distance
Cloud: Only tested during a sunny day, but couldn't see much performance difference between Low and Very High. High looks great, so I left it there.
Shadow: Not the big FPS hit it was in A2. High looked the most realistic on my machine.
Particle: Again, I couldn't see a significant visual or performance difference above High despite throwing smoke grenades around me.
As SideStrafe pointed out in his video, settings chosen during mucking about in the Editor are only an indication of how the game might perform during MP play as there's obviously a very low CPU load using the Editor with just one character model being generated.
Hopefully, this technodrivel might prove useful for others.