Star Swarm - Oxide bench thread

andybird123 · 12 Jun 2014 at 12:51

Turning motion blur off improves my fps by what likes like about 50%, doing another 3 runs to check consistency

edit: still doing further runs, to check consistency, but so far;

Custom Scenario as above - Extreme profile - ~72fps
Deferred Contexts On - ~72fps
DC ON, Motion Blur Off - ~102fps 30% gain
DC off, motion blur off - ~102fps

do whatever the engine updates / driver updates have done, they have massively improved FPS in general (I no longer see big dips like before either), but DC no longer seems to make any difference on or off, either because the engine makers have decided to remove them, or the driver updates

I'm also about to do a couple of timed runs at 1080s to see if a longer time period has any effect

_Alatar_ · 12 Jun 2014 at 13:01

LtMatt said:
Yes it does, it's not the same at all it's easy to spot. You can tell because the fps does not drop anywhere near as low as they do in extreme follow mode. The whole screen blurs where in the custom mode there is nothing. This benchmark was designed to showcase the advantages of Mantle (draw calls) all that custom csv does is remove that advantage making it pointless. It makes it consistent but it removes vital parts of the sequence.

lmao

absolutely clueless.

Let me tell you a secret about game development. Assuming an effect or a mesh is on the screen being rendered it doesn't matter how well the player sees it. That exact same effect/mesh still uses the exact same draw call/CPU time budget because you as a developer have told it to be rendered in the scene.

Imagine I want a million 2d butterly meshes on my screen at the same time. All moving constantly. Now that I have those added I have a crapton of draw calls to deal with. In my engine editor I can set the player camera very close to the butterlies or very far away from them. Being closer to the meshes the player can see them better. However this does not change the amount of draw calls that have to be handled.

Just because the gamer can't see X as well as before due to distance, camera movement etc. does not mean that it's not being rendered exactly the same way as it was before. (usually this causes GPU time issues which is why most of the time you want different LOD meshes for your models)

If you wanted to remove a ton of draw calls from this star swarm demo you'd have to change the motion blur implementation (which doesn't change unless you change settings, motion blur implementation is the exact same regardless of scenario, otherwise you'd see the RTS scenario being the easiest to run and follow being the hardest). Or alternatively you could group a ton of individual meshes together into single big meshes, which obviously isn't happening either.

LtMatt · 12 Jun 2014 at 13:41

I get much lower fps on the follow mode when the blurring occurs, that never happens in the custom scenario. I see no blurring on the the custom mode, so i conclude its not occurring otherwise i would be feeling the fps hit from it. If it is happening but we can't see it, then we are not feeling the fps hit from it.

Now ive just tried the RTS Extreme timed mode. It has a still camera but seems to be quite demanding. I get similar results in terms of fps, batching units etc to this 780TI result posted on youtube. I've not done loads of runs so no idea how consistent this is.

Martini1991 · 12 Jun 2014 at 13:48

I distinctly remember the Crysis 2 FPS hits that were talked with the rendered water that you couldn't see.

LtMatt · 12 Jun 2014 at 13:52

Martini1991 said:
I distinctly remember the Crysis 2 FPS hits that were talked with the rendered water that you couldn't see.

What does that have to do with Star Swarm? Seeing you brought it up it was 31-38% on Radeon cards and 17-21% on Nvidia cards. A bit more than 2 fps, eh?

As a very high-profile title, Crysis 2 has gotten lots of support from Nvidia in various forms. In and of itself, such support is generally a good thing for PC gaming. In fact, we doubt the DX11 patch for this game would even exist without Nvidia's urging. We know for a fact that folks at Nvidia were disappointed about how the initial Crysis 2 release played out, just as many PC gamers were. The trouble comes when, as sometimes happens, the game developer and GPU maker conspire to add a little special sauce to a game in a way that doesn't benefit the larger PC gaming community. There is precedent for this sort of thing in the DX11 era. Both the Unigine Heaven demo and Tom Clancy's HAWX 2 cranked up the polygon counts in questionable ways that seemed to inflate the geometry processing load without providing a proportionate increase in visual quality.

Unnecessary geometric detail slows down all GPUs, of course, but it just so happens to have a much larger effect on DX11-capable AMD Radeons than it does on DX11-capable Nvidia GeForces. The Fermi architecture underlying all DX11-class GeForce GPUs dedicates more attention (and transistors) to achieving high geometry processing throughput than the competing Radeon GPU architectures. We've seen the effect quite clearly in synthetic tessellation benchmarks. Few games have shown a similar effect, simply because they don't push enough polygons to strain the Radeons' geometry processing rates. However, with all of its geometric detail, the DX11 upgraded version of Crysis 2 now manages to push that envelope. The guys at Hardware.fr found that enabling tessellation dropped the frame rates on recent Radeons by 31-38%. The competing GeForces only suffered slowdowns of 17-21%.

Source
http://techreport.com/review/21404/crysis-2-tessellation-too-much-of-a-good-thing/6

Martini1991 · 12 Jun 2014 at 13:56

LtMatt said:
What does that have to do with Star Swarm? Seeing you brought it up it was 31-38% on Radeon cards and 17-21% on Nvidia cards. A bit more than 2 fps, eh?

Source
http://techreport.com/review/21404/crysis-2-tessellation-too-much-of-a-good-thing/6

Oh wow Matt.
"Crysis 2" the game.............

FPS hits as in "FPS hits", no figure given.

And it was an example of FPS hits without visual evidence on their rendering (As is being touted the case here)

As an aside, how can I get solid numbers just off percentages with no other numbers?
You can't debate for toffee.

LtMatt · 12 Jun 2014 at 14:00

Martini1991 said:
Oh wow Matt.
"Crysis 2" the game.............

FPS hits as in "FPS hits", no figure given.

And it was an example of FPS hits without visual evidence on their rendering (As is being touted the case here)

You can't debate for toffee.

Sorry, misread lol. Was expecting another bait. You get used to it here after a while.

Martini1991 · 12 Jun 2014 at 14:02

But the example obviously renders the argument moot about not seeing something means we're not feeling the FPS hit.

andybird123 · 12 Jun 2014 at 14:07

Official Entry

4930K @ 4.5ghz
Single Titan @ 1110mhz (boost clock) 962mhz base clock

99.64FPS

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\Andy\Documents\Star Swarm\Output_14_06_12_1359.txt
Version 1.10
06/12/2014 13:59
===========================================================

== Hardware Configuration =================================
GPU: NVIDIA GeForce GTX TITAN
CPU: GenuineIntel
Intel(R) Core(TM) i7-4930K CPU @ 3.40GHz
Physical Cores: 6
Logical Cores: 12
Physical Memory: 17095229440
Allocatable Memory: 8796092891136
===========================================================

== Configuration ==========================================
API: DirectX
Scenario: ScenarioFollow.csv
User Input: Disabled
Resolution: 1920x1080
Fullscreen: True
GameCore Update: 16.6 ms
Bloom Quality: High
PointLight Quality: High
ToneCurve Quality: High
Glare Overdraw: 16
Shading Samples: 64
Shade Quality: Mid
Deferred Contexts: Enabled
Temporal AA Duration: 16
Temporal AA Time Slice: 2
Detailed Frame Info: Off
===========================================================

== Results ================================================
Test Duration: 360 Seconds
Total Frames: 35873

Average FPS: 99.64
Average Unit Count: 4440
Maximum Unit Count: 5579
Average Batches/MS: 1247.54
Maximum Batches/MS: 4209.56
Average Batch Count: 14360
Maximum Batch Count: 95665
===========================================================

_Alatar_ · 12 Jun 2014 at 14:10

Matt you're just plain wrong. Your rudimentary understanding of game engines rendering something is completely wrong.

Let's prove that.

Follow [Extreme settings]

74fps

Follow [Extreme settings without motion blur]

120fps

Custom Scenario [Extreme settings]

51fps

Custom Scenario [Extreme settings without motion blur]

80fps

Follow blur to no blur fps increase: 62%

Custom blur to no blur fps increase: 57%

It's the exact same fps hit between the modes. The same motion blur is being rendered every time. And as you can see the custom scenario is actually the more stressful one out of the two. The small variation in percentages can be attributed to the inconsistency of the follow preset.

Gregster · 12 Jun 2014 at 14:11

Nicely done and on the scoreboard Andy

andybird123 · 12 Jun 2014 at 14:22

Official Entry

sorry, nother one

Single Titan; 1020mhz base clock, 1176mhz boost
4930k is at 4.5ghz as well G

109.09fps

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\Andy\Documents\Star Swarm\Output_14_06_12_1412.txt
Version 1.10
06/12/2014 14:12
===========================================================

== Hardware Configuration =================================
GPU: NVIDIA GeForce GTX TITAN
CPU: GenuineIntel
Intel(R) Core(TM) i7-4930K CPU @ 3.40GHz
Physical Cores: 6
Logical Cores: 12
Physical Memory: 17095229440
Allocatable Memory: 8796092891136
===========================================================

== Configuration ==========================================
API: DirectX
Scenario: ScenarioFollow.csv
User Input: Disabled
Resolution: 1920x1080
Fullscreen: True
GameCore Update: 16.6 ms
Bloom Quality: High
PointLight Quality: High
ToneCurve Quality: High
Glare Overdraw: 16
Shading Samples: 64
Shade Quality: Mid
Deferred Contexts: Enabled
Temporal AA Duration: 16
Temporal AA Time Slice: 2
Detailed Frame Info: Off
===========================================================

== Results ================================================
Test Duration: 360 Seconds
Total Frames: 39272

Average FPS: 109.09
Average Unit Count: 4355
Maximum Unit Count: 5368
Average Batches/MS: 1132.69
Maximum Batches/MS: 4436.60
Average Batch Count: 11892
Maximum Batch Count: 72331
===========================================================

Gregster · 12 Jun 2014 at 14:53

Awesome

Will add now and change the CPU score

LtMatt · 12 Jun 2014 at 15:13

_Alatar_ said:
Matt you're just plain wrong. Your rudimentary understanding of game engines rendering something is completely wrong.

Let's prove that.

Follow [Extreme settings]

74fps

Follow [Extreme settings without motion blur]

120fps

Custom Scenario [Extreme settings]

51fps

Custom Scenario [Extreme settings without motion blur]

80fps

Follow blur to no blur fps increase: 62%

Custom blur to no blur fps increase: 57%

It's the exact same fps hit between the modes. The same motion blur is being rendered every time. And as you can see the custom scenario is actually the more stressful one out of the two. The small variation in percentages can be attributed to the inconsistency of the follow preset.

I just said what i see. I never claimed to have complete understanding of the game engine. I just said i didn't think it was running because i couldn't see it and didn't have the fps hit. I was sure that was the case. If i was wrong then i was wrong. All i know is i get massive motion blur on one and none on the other.

andybird123 · 12 Jun 2014 at 21:15

Gregster said:
Awesome

Will add now and change the CPU score

it is a shame that no one with a mantle capable card seems to want to showcase mantle's performance by posting results

Gregster · 13 Jun 2014 at 10:13

andybird123 said:
it is a shame that no one with a mantle capable card seems to want to showcase mantle's performance by posting results

Rules still too hard ya reckon?

tommybhoy · 13 Jun 2014 at 12:05

andybird123 said:
it is a shame that no one with a mantle capable card seems to want to showcase mantle's performance by posting results

Gregster said:
Rules still too hard ya reckon?

There's plenty of Mantle results in the thread.

The vexatious nature of the thread combined with-for example:

tommybhoy said:
290(X)@1200/1450/[email protected]

MANTLE-Average FPS: 67.92

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
E:\Documents\Star Swarm\Output_14_02_02_0323.txt
Version 1.00
02/02/2014 03:23
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon R9 200 Series
CPU: GenuineIntel
Intel(R) Core(TM) i7-3770K CPU @ 3.50GHz
Physical Cores: 4
Logical Cores: 8
Physical Memory: 17122693120
Allocatable Memory: 140737488224256
===========================================================

== Configuration ==========================================
API: Mantle
Scenario: ScenarioFollow.csv
User Input: Disabled
Resolution: 1920x1080
Fullscreen: True
GameCore Update: 16.6 ms
Bloom Quality: High
PointLight Quality: High
ToneCurve Quality: High
Glare Overdraw: 16
Shading Samples: 64
Shade Quality: Mid
Deferred Contexts: Disabled
Temporal AA Duration: 16
Temporal AA Time Slice: 2
Detailed Frame Info: Off
===========================================================

== Results ================================================
Test Duration: 360 Seconds
Total Frames: 24453

Average FPS: 67.92
Average Unit Count: 4809
Maximum Unit Count: 5969
Average Batches/MS: 1229.23
Maximum Batches/MS: 3033.72
Average Batch Count: 21846
Maximum Batch Count: 119116
===========================================================

^^^^^^^
INVALID result apparently due to no 'official entry' moniker.

andybird123 said:
sorry, nother one

Single Titan; 1020mhz base clock, 1176mhz boost
4930k is at 4.5ghz as well G
109.09fps

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\Andy\Documents\Star Swarm\Output_14_06_12_1412.txt
Version 1.10
06/12/2014 14:12
===========================================================

== Hardware Configuration =================================
GPU: NVIDIA GeForce GTX TITAN
CPU: GenuineIntel
Intel(R) Core(TM) i7-4930K CPU @ 3.40GHz
Physical Cores: 6
Logical Cores: 12
Physical Memory: 17095229440
Allocatable Memory: 8796092891136
===========================================================

== Configuration ==========================================
API: DirectX
Scenario: ScenarioFollow.csv
User Input: Disabled
Resolution: 1920x1080
Fullscreen: True
GameCore Update: 16.6 ms
Bloom Quality: High
PointLight Quality: High
ToneCurve Quality: High
Glare Overdraw: 16
Shading Samples: 64
Shade Quality: Mid
Deferred Contexts: Enabled
Temporal AA Duration: 16
Temporal AA Time Slice: 2
Detailed Frame Info: Off
===========================================================

== Results ================================================
Test Duration: 360 Seconds
Total Frames: 39272

Average FPS: 109.09
Average Unit Count: 4355
Maximum Unit Count: 5368
Average Batches/MS: 1132.69
Maximum Batches/MS: 4436.60
Average Batch Count: 11892
Maximum Batch Count: 72331
===========================================================

^^^^^^
A VALID result despite no 'official entry' moniker.

Which equates to an unfriendly chafe thread that is to bothersome for most to give a **** about any more, mission accomplished, job's a good un.

LtMatt · 13 Jun 2014 at 12:07

Lol. It's a fair point, i see no official entry on andy's post.

andybird123 · 13 Jun 2014 at 12:15

I'm sure Greg will make a special exception for you now you've highlighted that you wanted yours to be an official entry tommy

to be fair to greg, it was confusing which were entries and which were not as people were throwing up all sorts of tests without saying they had run the test as per the OP, I think we can cut G some slack for updating my second score when it was very clear it was an official entry posted just after one that did have the title set

I also wouldn't have gotten upset with Greg if he'd missed it, or pulled me up on it, I would have edited my post to have the correct title - which I have, thanks tommy for pointing it out

your post that you've quoted, directly before and after are posts with custom scenario's, so it hardly the same situation at all, if you want your score added then surely just repost it with the title added, no?

or you could just start another thread and add the scores yourself?

Gregster · 13 Jun 2014 at 12:39

I have had to use common sense in all the bench threads I maintain and with all the random posts of results from God knows what test, I thought it would be simple to add "Official entry" to the post when you want a score added. I am sorry if the rules are complicated but that's the way it is Tommy and you are quite correct, I did add Andy's score even though it didn't say "official entry" and I was just using common sense.

Feel free to have a word with a mod if you feel I am being harsh and start your own Star Swarm thread Tommy. Maybe it will run longer than the last bench thread you was running, which you couldn't be bothered to update so gave up on it