Flaws on Oxide

[Shivansps]

And when not on screen too, i get 15fps even if there only 1 ship on screen that has lower quality than 1999 Freespace 2 ships.

woow, just, woow, why anyone could defend this? whiout mantle we couldt render a ugly ship with a black background with white dots on it at more than 15 fps?

Guys im gona make this simple, if we need to go insane doing really dumb stuff to show off Mantle, them we dont need it, plain and simple, for a benchmark, first the thing has a horrible visual quality, just horrible, 1999 original Freespace 2 may look even better than this, it has better weapon, shield and explosions effects for sure.

Second, instead of making a limited but realistic escenario to show up Mantle potential they went in full retard mode, it its also a awfull show up for the engine itselft, it lacks even the most basic and essential optimisations every game need! even if you gona use mantle you cant do anything like that.

I was expecting something A LOT better than this.

 

[james1701]

14.1 Driver with Mantel

== Results ================================================
Test Duration: 360 Seconds
Total Frames: 19796

Average FPS: 54.99
Average Unit Count: 4516
Maximum Unit Count: 5727
Average Batches/MS: 926.61
Maximum Batches/MS: 2755.79
Average Batch Count: 18719
Maximum Batch Count: 120986
===========================================================

14.1 3D settings

== Results ================================================
Test Duration: 360 Seconds
Total Frames: 12803

Average FPS: 35.56
Average Unit Count: 4045
Maximum Unit Count: 5341
Average Batches/MS: 716.94
Maximum Batches/MS: 1415.06
Average Batch Count: 21808
Maximum Batch Count: 170191
===========================================================

 

[Adampa1006]

Havent read whole thread, sorry if this has been posted. This is not directx crippled. From their FAQ

Q. This is just a marketing tool for AMD; you’ve obviously crippled the DirectX version!

A. We really haven’t; to be perfectly honest we’ve spent more time optimizing for DirectX than we have on Mantle. The fact is that DirectX is conceived and implemented as a single-threaded API, and so a lot of the more significant gains we see thanks to the Nitrous engine’s aggressive multithreading are badly limited by API overhead when we’re using it.

We obviously can’t prove this to the satisfaction of everyone on the Internet, but understand that our primary goal with Nitrous is to make the best engine we can so that we can open the door to the new kinds of games that we want to make (and play ourselves!). An awful lot of the gamers we hope to entertain don’t and won’t have access to Mantle-enabled hardware any time soon, so we’d be making a huge mistake as entertainers and as businesspeople by not supporting or poorly supporting non-Mantle hardware.​

 

[DreamMystic]

DX11

== Hardware Configuration =================================
GPU: AMD Radeon R9 200 Series
CPU: AuthenticAMD
AMD FX(tm)-8350 Eight-Core Processor
Physical Cores: 4
Logical Cores: 8
Physical Memory: 17072631808
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: 10025

Average FPS: 27.84
Average Unit Count: 4101
Maximum Unit Count: 5505
Average Batches/MS: 403.37
Maximum Batches/MS: 953.93
Average Batch Count: 16806
Maximum Batch Count: 134047
===========================================================


Mantle

== Hardware Configuration =================================
GPU: AMD Radeon R9 200 Series
CPU: AuthenticAMD
AMD FX(tm)-8350 Eight-Core Processor
Physical Cores: 4
Logical Cores: 8
Physical Memory: 17072631808
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: 15949

Average FPS: 44.30
Average Unit Count: 4402
Maximum Unit Count: 5382
Average Batches/MS: 814.14
Maximum Batches/MS: 2019.60
Average Batch Count: 20942
Maximum Batch Count: 169978
===========================================================

 

[Carfax83]

@BrightCandle and Sontin, what drivers are you guys using? I'm using the 334.67, and the benchmark is crashing on start up with DC enabled..

 

[csbin]

computerbase.de did some benching.

http://www.computerbase.de/artikel/grafikkarten/2014/erste-eindruecke-zu-amds-mantle/3/

I asked them already to rerun the GTX 780 Ti with deferred contexts enabled. If you have an account you should ask too as they seem to interact in english in their Twitter.

It's actually enjoyable. :biggrin:

 

[biostud]

System below @1920x1080

DX:
== Results ================================================
Test Duration: 360 Seconds
Total Frames: 8141

Average FPS: 22.61
Average Unit Count: 3948
Maximum Unit Count: 5528
Average Batches/MS: 487.91
Maximum Batches/MS: 843.32
Average Batch Count: 21235
Maximum Batch Count: 139556
===========================================================

Mantle:
== Results ================================================
Test Duration: 360 Seconds
Total Frames: 17092

Average FPS: 47.48
Average Unit Count: 4564
Maximum Unit Count: 5610
Average Batches/MS: 790.52
Maximum Batches/MS: 2096.19
Average Batch Count: 19330
Maximum Batch Count: 112872
===========================================================

So 110% increase on my system

System below @2560x1440

DX:
== Results ================================================
Test Duration: 360 Seconds
Total Frames: 6968

Average FPS: 19.35
Average Unit Count: 4023
Maximum Unit Count: 5625
Average Batches/MS: 481.56
Maximum Batches/MS: 851.87
Average Batch Count: 24663
Maximum Batch Count: 102041
===========================================================

Mantle:
== Results ================================================
Test Duration: 360 Seconds
Total Frames: 15068

Average FPS: 41.85
Average Unit Count: 4663
Maximum Unit Count: 5936
Average Batches/MS: 656.08
Maximum Batches/MS: 2151.62
Average Batch Count: 17937
Maximum Batch Count: 116914
===========================================================

So 116% @ 2560x1440

 

[BrightCandle]

I am using 334.67 also.

 

[Carfax83]

I am using 334.67 also.

Weird, and we both have similar rigs as well. You said you have Windows 8 Pro though if I recall, and sontin said he had Windows 7..

I'm using 8.1, so that might be why I'm guessing..

 

[DarkKnightDude]

That might not be it Carfax, because the benchmark works with my GTX760 and that rig is on Windows 8.1 pro with drivers 334.67

Have you tried reinstalling it or launching it in compatibility mode?

 

[Carfax83]

That might not be it Carfax, because the benchmark works with my GTX760 and that rig is on Windows 8.1 pro with drivers 334.67

Have you tried reinstalling it or launching it in compatibility mode?

Launching it in Windows 8 compatibility mode fixed the problem thanks..

Doesn't seem like a very good benchmark though..

 

[Noctifer616]

Doesn't seem like a very good benchmark though..

It's not a benchmark, it's a stress test.

Hopefully Oxide should give us info soon on how to set it up to run as a benchmark.

 

[Erenhardt]

I don't know how mantle rendering works, but I found a funny thing.
1. Start Starswarm (mantle in my case) demo with input enabled.
2. Move the camera so that not a single ship is in sight.
3. Observe CPU utilization in task manager
4. ???
5. Smell the roasting dust on CPU heatsink
6. profit

 

[psolord]

I did a test run with Star Swarm, which was updated today. I used my 2500k+single 7950.

I run Attract/Custom/1080P/Mantle (2500k @ 4Ghz) where custom was the extreme settings without the motion blur. Motion blur is extremely ugly in this game. You can't practically see anything.

[System]
Fullscreen=1
Resolution=1920,1080
vsync=0
FixedFrameMS=16.6

[Quality]
Bloom=HIGH
PointLights=HIGH
ToneCurve=HIGH
Glare=16
ShadingSamples=64
ShadeQuality=MID
DeferredContexts=0
EnableMotionBlur=0
MaxMotionBlurFrameTime=16
MotionBlurInterframeTime=2
RenderResolution=1920,1080

[TimedRun]
Scenario=ScenarioFollow.csv
Seconds=360
TimedRun=0
FrameDump=0

The results for Mantle were

== Results ================================================
Test Duration: 360 Seconds
Total Frames: 11841

Average FPS: 32.89
Average Unit Count: 3902
Maximum Unit Count: 5491
Average Batches/MS: 208.53
Maximum Batches/MS: 754.76
Average Batch Count: 6953
Maximum Batch Count: 33638
===========================================================

The results of D3D were:

== Results ================================================
Test Duration: 360 Seconds
Total Frames: 22850

Average FPS: 63.47
Average Unit Count: 4615
Maximum Unit Count: 5793
Average Batches/MS: 295.47
Maximum Batches/MS: 989.34
Average Batch Count: 5429
Maximum Batch Count: 32917
===========================================================

Um what the hell? Is it not optimized for Tahiti's yet?

 

[SiliconWars]

I'm sure Repi has forgotten more about programming than most (or any) of us here, but anyone that has a free thinking brain will find issues with that statement.

If DCL is fundamentally broken, why did it work in Civ 5? The last benchmark of Civ 5 on Anandtech was in the GTX 770 review, and you can see the GTX 780, the most CPU limited GPU, is nearly 50% ahead of the 7970 Ghz, despite being on average 20% faster:

It didn't work in CiV, it worked in the CiV "late game demo benchmark" (and incidentally AMD closed most of the gap without using multi-threading...)

The actual gameplay benchmarks were completely different.

 

[sontin]

Launching it in Windows 8 compatibility mode fixed the problem thanks..

Doesn't seem like a very good benchmark though..

I installed Windows 8.1 and the demo is not really working on it.
DC is even slower than on Windows 7.

This thing is murks. Says nothing about DX.

/edit: Deferred Context doesn't work under 8.1 with a nVidia card.

 

[Carfax83]

It didn't work in CiV, it worked in the CiV "late game demo benchmark" (and incidentally AMD closed most of the gap without using multi-threading...)

The actual gameplay benchmarks were completely different.

You have no idea what you're talking about. The late game benchmark according to the designers themselves:

Late Game View Benchmark. This benchmark is designed to simulate a late game workload. This scenario exercises all aspects of the game engine pipeline since all simulation and renderable object types are represented at a frequency consistent with a game that has been in progress for 300+ turns. To run this test, run the application with the command line option “-benchmark lateGameView”. Case insensitive. Results are reported in the “LateGameViewBench.log” file. Full Render Score reflects the active rendering settings of you application. No Shadow render score turns the shadow pass off for the application. No Render score emulates an infinitely fast GPU and will bypass most driver overhead. This serves as a baseline for Civ 5’s engine performance vs GPU and Driver performance. ( note while in this mode, the display will flicker ). All results are reported as frames rendered over 60 seconds. Additionally the number of draw call made per frame is reported. Note: Mousewheel zoom in some benchmarks are enabled for inspection, however using this will affect the end results. For consistent results the user should avoid altering the zoom level.

Also, deferred context is a major feature of the LORE engine. They built their entire engine around deferred context rendering, so the pathway is highly optimized for that type of rendering.

 

[Carfax83]

I installed Windows 8.1 and the demo is not really working on it.
DC is even slower than on Windows 7.

This thing is murks. Says nothing about DX.

/edit: Deferred Context doesn't work under 8.1 with a nVidia card.

Yeah, I figured Windows 8.1 had something to do with this. Or maybe it's just the application itself..

 

[krumme]

Havent read whole thread, sorry if this has been posted. This is not directx crippled. From their FAQ

Q. This is just a marketing tool for AMD; you’ve obviously crippled the DirectX version!

A. We really haven’t; to be perfectly honest we’ve spent more time optimizing for DirectX than we have on Mantle. The fact is that DirectX is conceived and implemented as a single-threaded API, and so a lot of the more significant gains we see thanks to the Nitrous engine’s aggressive multithreading are badly limited by API overhead when we’re using it.

We obviously can’t prove this to the satisfaction of everyone on the Internet, but understand that our primary goal with Nitrous is to make the best engine we can so that we can open the door to the new kinds of games that we want to make (and play ourselves!). An awful lot of the gamers we hope to entertain don’t and won’t have access to Mantle-enabled hardware any time soon, so we’d be making a huge mistake as entertainers and as businesspeople by not supporting or poorly supporting non-Mantle hardware.​

Its incredible they have to write this. They have a business. Ofcource they optimize foremost for dx. Its a stresstest showing excactly where dx is weakest vs mantle but its doesnt mean all gameplay will be like that in the games using the engine but rts is just plain simply drawcall limited and it show in the games we already have and it shows here. No big surprice.

I am pretty sure nv will do the same for engines like this as it have proven to be quite cheap. And then all this crying can hopefully stop.

 

[SiliconWars]

You have no idea what you're talking about. The late game benchmark according to the designers themselves:

Also, deferred context is a major feature of the LORE engine. They built their entire engine around deferred context rendering, so the pathway is highly optimized for that type of rendering.

Really? You should go tell SKYMTL as it was him who made the claim. And you know seeing as he was benchmarking it over and over...

 

[stahlhart]

This thread is about StarSwarm, not Civilization 5. Take it offline and stop derailing the discussion.
-- stahlhart

 

[ShintaiDK]

Swarm demo got updated today on steam. No changelog info. DX performance decreased 10-20% in the few runs I did. DC option also got little to no impact now.

 

[Pottuvoi]

Did couple of tests yesterday while tweaking .INI file.

For those who are interested in their shading pipeline try set shading samples to 1.
This makes object based shading low resolution enough to be seen on ships.

Will try to do tweaks today on resolutions if I have time.
Resolution shouldn't have big impact on rendering speed. (even when not CPU limited.)

 

[jj109]

Havent read whole thread, sorry if this has been posted. This is not directx crippled. From their FAQ

Q. This is just a marketing tool for AMD; you’ve obviously crippled the DirectX version!

A. We really haven’t; to be perfectly honest we’ve spent more time optimizing for DirectX than we have on Mantle.
​

You mean these perfectly honest optimizations?

And when not on screen too, i get 15fps even if there only 1 ship on screen that has lower quality than 1999 Freespace 2 ships.

Or these?

 

[Shivansps]

Freespace 2 (1999), improved by community @1080P

http://www.youtube.com/watch?v=1mTfktLn-mk
around 300 fighters combat

http://www.youtube.com/watch?v=r5LscCmUQ58
fleet battle

each ship has its unique AI working on them, each turret has its unique AI working on them, they find and attack targets independientely, plus subsystem destruction, the game is fully single threaded, means, the game is doing rendering, AI, sound and everything else on the same CPU core, it does not even set the affinity to more than 1 core. All of that on Windows, using the ancient OpenGL 1.4 + GLSL extensions and a HD5850.
Fraps max fps is capped at 30, and you can see that the fps increases if you are looking at nothing.

Lets compare to this just released in 2014:
720P

-5000 ships... yeah a useless number, how you manage to run AI on all those individually, its kind of pointless, also 99% of the time im not seeing them.
-AI is nearly non-existant, the ships are in wings following waypoints whiout any real AI on them, there are MAYBE 80, at best, diferent AI in the whole tech demo, can you easily see this because all ships in a wing do the same thing at the same time, firing at nowhere in imposible angles.
-For a tech demo is plain AWFULL, horrible effects, horrible low poly models models, horrible shaders, blur, explosions, horrible everything.
-No real optimisations, even if you look at nothing, fps are on the ground and AI is not the bottleneck.

Im pissed off at Star Swarm because i was waiting something like "wooow look how much more you can do", adding detailed models, in big numbers with its own unique AI that you can get by freeing CPU, and all kinds of optimisations that whould allow me to see a better game overall, instead i got this that, to me, does not stand up to a 1999 game engine, and the only two remarkable thing that does is waste gpu in rendering things that are not needed (that is not optimisation) and add lots of mindless ships that works as one, if you "melt" the wings intro 1 big ship you whould no see any difference, instead that the number whould say "100 ships" instead of 5000, placebo effect people.