Star Swarm - Oxide bench thread

[triss]

Ah sorry my bad

 

[andybird123]

(The CPU is Bottlenecked) By the DX API because it cannot process for the Drawcalls the CPU can calculate, so while The GPU usage is dropping through lack of CPU performance, its not the CPU 'sitting on the other side of the DX API' that is the problem, its the API getting in its way.

The GPU asked DX for 8K parallel Drawcalls, The API said; i can only give you 5K, thats all i can process. the CPU on the other end calculating those Drawcalls is saying to DX, 5K? is that all you want? *yawn* the GPU back to the DX API. ok, i will have to slow things down at this end so you can keep up.

DirectX, or rather DirectX 11, is Multi-threaded. "DX Deferred Path" Google it its used in BF3 and 4, the problem with DirectX is DirectX, DX Deferred or not.

They dropped it from BF4 remember, the quote from johan on twitter was that they saw no improvement with it, which is odd as civ5 and the starswarm demo see around 50% increase in fps, and things like AC3/4 see 25-30%

Your summation is also not correct, DX gets cpu bottlnecked when it isnt threading well and a single core hits 100%, not "just because i say so"

 

[Gregster]

Don't waste your time Andy, it is like talking to a brick wall

 

[LtMatt]

They dropped it from BF4 remember, the quote from johan on twitter was that they saw no improvement with it, which is odd as civ5 and the starswarm demo see around 50% increase in fps, and things like AC3/4 see 25-30%

Your summation is also not correct, DX gets cpu bottlnecked when it isnt threading well and a single core hits 100%, not "just because i say so"

Don't waste your time Andy, it is like talking to a brick wall

Naw don't put yourself down Greg.

 

[humbug]

They dropped it from BF4 remember, the quote from johan on twitter was that they saw no improvement with it, which is odd as civ5 and the starswarm demo see around 50% increase in fps, and things like AC3/4 see 25-30%

Your summation is also not correct, DX gets cpu bottlnecked when it isnt threading well and a single core hits 100%, not "just because i say so"

Check your user settings file.




If my summation is not correct then so it is for this guy and everyother professional who has commented on it.

As a PC guy who knows hardware to the metal, I spend most of my days in frustration knowing damn well what I could do with the hardware, but what I cannot do because Microsoft and IHVs wont provide low-level GPU access in PC APIs. One simple example, drawcalls on PC have easily 10x to 100x the overhead of a console

(The CPU is Bottlenecked) By the DX API because it cannot process for the Drawcalls the CPU can calculate, so while The GPU usage is dropping through lack of CPU performance, its not the CPU 'sitting on the other side of the DX API' that is the problem, its the API getting in its way.

 

[andybird123]

Check your user settings file.


If my summation is not correct then so it is for this guy and everyother professional who has commented on it.

Johan Anderson posted on twitter saying that they'd dropped command lists because there was no benefit and in fact a detriment in some cases (e.g. AMD) in using them... you use deferred contexts to build command lists so using one without the other is a bit pointless... I'm not sure whether to believe anything Johan says on twitter in that case, as he seems to not realise his own game still uses DC

 

[humbug]

Johan Anderson posted on twitter saying that they'd dropped command lists because there was no benefit and in fact a detriment in some cases (e.g. AMD) in using them... you use deferred contexts to build command lists so using one without the other is a bit pointless... I'm not sure whether to believe anything Johan says on twitter in that case, as he seems to not realise his own game still uses DC

It was added to BF4 in the last patch, along with the CPU thread allocation string you see there a couple of lines up, i don't know what it does on AMD GPU's, but DX11 batching seems to work just fine.

The CPU, DX11 CPU threading is used for many different things, various calculations and rendering options are done with it; and on the CPU, some aspects it may help while in others it may not.

Its not just me saying DX is the Bottleneck, its the professionals who use it.

DX thread batching on AMD;

http://www.behardware.com/articles/770-2/report-amd-radeon-hd-5870-and-5850.html

DC should work fine ?
seems it was tested on a 4890 even
We tested this on a Radeon HD 4890 with a demo supplied by Microsoft in its SDK and which works on all Direct3D 10 cards. This represents a scene limited by the CPU,

- Rendering in immediate context mode: 55 fps
- Rendering in deferred context mode, not multithreaded: 48 fps
- Rendering in deferred context mode, multithreaded: 78 fps

might be since way back when this was tested DC when it wasnt multithreaded (assuming a single core cpu) was slower would be a valid reason for the Demo now it to be disabled by default

 

[andybird123]

In what way is a dx10 demo related to star swarm?

Yes, in some case DX is clearly a bottleneck, however in THOSE cases it is very clear because at least 1 Cpu core hits 100%, the thread on that cpu cant do anything else and it prevents all the other threads and the gpu from carrying on

With star swarm, nothing hits 100%, links to BF4 or an sdk demo dont detract from that

tell you what, why don't you actually test SS with DC on yourself and tells us what results you get

 

[humbug]

tell you what, why don't you actually test SS with DC on yourself and tells us what results you get

I will when i have time later, i already promised Triss i would.

 

[los]

[email protected] 7970 1200/1800 13.12 driver

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\Los\Documents\Star Swarm\Output_14_02_19_1737.txt
Version 1.10
02/19/2014 17:37
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon HD 7900 Series
CPU: GenuineIntel
Intel(R) Core(TM) i5-4670K CPU @ 3.40GHz
Physical Cores: 4
Logical Cores: 4
Physical Memory: 8543887360
Allocatable Memory: 140737488224256
===========================================================


== 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: Disabled
Temporal AA Duration: 16
Temporal AA Time Slice: 2
Detailed Frame Info: Off
===========================================================


== Results ================================================
Test Duration: 360 Seconds
Total Frames: 18311

Average FPS: 50.86
Average Unit Count: 4263
Maximum Unit Count: 5565
Average Batches/MS: 732.37
Maximum Batches/MS: 1869.13
Average Batch Count: 15867
Maximum Batch Count: 82322
===========================================================

 

[los]

same as above but 14.1 running mantle

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\Los\Documents\Star Swarm\Output_14_02_19_1749.txt
Version 1.10
02/19/2014 17:49
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon HD 7900 Series
CPU: GenuineIntel
Intel(R) Core(TM) i5-4670K CPU @ 3.40GHz
Physical Cores: 4
Logical Cores: 4
Physical Memory: 8543887360
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: 21130

Average FPS: 58.69
Average Unit Count: 4603
Maximum Unit Count: 5691
Average Batches/MS: 1061.99
Maximum Batches/MS: 3753.29
Average Batch Count: 20887
Maximum Batch Count: 153311
===========================================================

 

[los]

same as above but 14.1 running D3D

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\Los\Documents\Star Swarm\Output_14_02_19_1757.txt
Version 1.10
02/19/2014 17:57
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon HD 7900 Series
CPU: GenuineIntel
Intel(R) Core(TM) i5-4670K CPU @ 3.40GHz
Physical Cores: 4
Logical Cores: 4
Physical Memory: 8543887360
Allocatable Memory: 140737488224256
===========================================================


== 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: Disabled
Temporal AA Duration: 16
Temporal AA Time Slice: 2
Detailed Frame Info: Off
===========================================================


== Results ================================================
Test Duration: 360 Seconds
Total Frames: 14188

Average FPS: 39.40
Average Unit Count: 4232
Maximum Unit Count: 5510
Average Batches/MS: 682.88
Maximum Batches/MS: 1844.94
Average Batch Count: 19316
Maximum Batch Count: 100613
===========================================================

 

[thefogo]

^ So there's a huge hit to performance in D3D with 14.1?

 

[los]

There was for me. All the settings stayed the same. I Then reinstalled 13.12.

 

[humbug]

GPU: 7870XT (Tahiti LE) @ 1253 / 1637
CPU: Phenom II x6 1090T @ 4.3Ghz

DirectX: 32.375
Mantle: 49.798

Direct X + Deferred Contexts: 30.261
Mantle + Deferred Contexts: 50.316


Direct X

Spoiler

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\DeskTop\Documents\Star Swarm\Output_14_02_19_1839.txt
Version 1.10
02/19/2014 18:39
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon HD 7800 Series
CPU: AuthenticAMD
AMD Phenom(tm) II X6 1090T Processor
Physical Cores: 6
Logical Cores: 6
Physical Memory: 8549396480
Allocatable Memory: 140737488224256
===========================================================


== 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: Disabled
Temporal AA Duration: 16
Temporal AA Time Slice: 2
Detailed Frame Info: Off
===========================================================


== Results ================================================
Test Duration: 360 Seconds
Total Frames: 11531

Average FPS: 32.375
Average Unit Count: 3830
Maximum Unit Count: 5612
Average Batches/MS: 387.64
Maximum Batches/MS: 943.44
Average Batch Count: 19794
Maximum Batch Count: 136675
===========================================================

Direct X + Deferred Contexts

Spoiler

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\DeskTop\Documents\Star Swarm\Output_14_02_19_1855.txt
Version 1.10
02/19/2014 18:55
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon HD 7800 Series
CPU: AuthenticAMD
AMD Phenom(tm) II X6 1090T Processor
Physical Cores: 6
Logical Cores: 6
Physical Memory: 8549396480
Allocatable Memory: 140737488224256
===========================================================


== 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: 10896

Average FPS: 30.261
Average Unit Count: 3811
Maximum Unit Count: 5730
Average Batches/MS: 456.74
Maximum Batches/MS: 887.57
Average Batch Count: 28185
Maximum Batch Count: 144205
===========================================================

Mantle

Spoiler

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\DeskTop\Documents\Star Swarm\Output_14_02_19_1846.txt
Version 1.10
02/19/2014 18:46
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon HD 7800 Series
CPU: AuthenticAMD
AMD Phenom(tm) II X6 1090T Processor
Physical Cores: 6
Logical Cores: 6
Physical Memory: 8549396480
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: 17927

Average FPS: 49.798
Average Unit Count: 4316
Maximum Unit Count: 5619
Average Batches/MS: 670.47
Maximum Batches/MS: 7451.53
Average Batch Count: 22736
Maximum Batch Count: 153435
===========================================================

Mantle + Deferred Contexts

Spoiler

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\DeskTop\Documents\Star Swarm\Output_14_02_19_1902.txt
Version 1.10
02/19/2014 19:02
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon HD 7800 Series
CPU: AuthenticAMD
AMD Phenom(tm) II X6 1090T Processor
Physical Cores: 6
Logical Cores: 6
Physical Memory: 8549396480
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: Enabled
Temporal AA Duration: 16
Temporal AA Time Slice: 2
Detailed Frame Info: Off
===========================================================


== Results ================================================
Test Duration: 360 Seconds
Total Frames: 18114

Average FPS: 50.316
Average Unit Count: 4116
Maximum Unit Count: 5551
Average Batches/MS: 719.33
Maximum Batches/MS: 6839.18
Average Batch Count: 22735
Maximum Batch Count: 123409
===========================================================

 

[andybird123]

And what was Cpu usage like, per core?
(deferred contexts enabled wont be doing anything with mantle also enabled as it is a DX tech, the difference is just the demo's variability)

That is really interesting los same DX performane on 13.12 as with mantle on 14.1
The plot thickens

 

[humbug]

And what was Cpu usage like, per core?

On Direct X + Deferred Contexts?

The same as it was on DX, about ~50% on each core.

 

[MyBrainz]

Nevermind.. after reading up on this bench it all seems pointless.

 

[Kaapstad]

Single 290X @1280/1625

4930k @4.8

===========================================================
Oxide Games
Star Swarm Stress Test - ©2013
C:\Users\Keve\Documents\Star Swarm\Output_14_04_27_1627.txt
Version 1.10
04/27/2014 16:27
===========================================================

== Hardware Configuration =================================
GPU: AMD Radeon R9 200 Series
CPU: GenuineIntel
Intel(R) Core(TM) i7-4930K CPU @ 3.40GHz
Physical Cores: 6
Logical Cores: 12
Physical Memory: 17116684288
Allocatable Memory: 8796092891136
===========================================================


== 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: 26915

Average FPS: 74.76
Average Unit Count: 4670
Maximum Unit Count: 5907
Average Batches/MS: 1202.87
Maximum Batches/MS: 10692.37
Average Batch Count: 18883
Maximum Batch Count: 112214
===========================================================

There is no consistency in this bench.

 

[vsg28]

You would need to create a custom scenario to have any sort of replication and consistency.