• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Star Swarm - Oxide bench thread

spankingtexan said:
My thoughts too, I don't know what percentage usage AMD owners are getting but my 780 is well under 50%, also got a better score when disabled SLI, it does appear this bench is geared up for mantle while DX is crippled.
Having said that, I hope AMD / Mantle users are getting nearer to NVidia performance, as for me I'll stick with NVidia / DX for now. :D
Click to expand...

It's not crippled at all, this is a game that was in development for a long time under DX only. It was 100% being worked on for DX till september and a single person has been working to add the Mantle code since then, Mantle is unoptimised and DX is optimised.

The issue is unit count/draw calls, DX simply can't scale beyond a certain level so 99% of games simply design to this limit and don't go past it or they'd suffer the horrible performance you're getting. This is the first engine really designed to be able to scale up to that level.

If the benchmark ran say a 500 unit count maximum, Mantle would offer a lower performance boost and Nvidia gpu usage would be higher as the DX thread would be significantly, massively reduced.

The gap in performance will likely grow as Mantle is properly optimised for, xfire in dx vs Mantle will be an even bigger bloodbath.

Game design is currently held back by dx overhead, this demo is merely showing, remove the API overhead... this is the performance you can get, these are the games we can make without the overhead in place.
 
Gregster said:
Mantle looking good. It is just a shame that this is the only game that doesn't allow the GPU past 50% using DX. Results obviously looking great for Mantle and poor for DX.
Click to expand...

Its the nature of this bench.

Evey ship, Turret and point of Light is a CPU Physics calculation. in this it keeps building and building up until the CPU gets completely bogged down.

In a Game like BF4 the Physics calculations are fixed, and they are generally fixed to a limit that DX can cope with, so mantle isn't going to make a massive difference as the render is designed to run with the DX weakness in mind.
That is, unless you have a very weak CPU, that very weak CPU will suddenly become a powerful CPU in Mantle BF4

What it shows, IMO, is the possible future, now Devs have much more CPU power to play with if they use Mantle. :)
 
Last edited:
humbug said:
Its the nature of this bench.

Evey ship, Turret and point of Light is a CPU Physics calculation. in this it keeps building and building up until the CPU gets completely bogged down.

In a Game like BF4 the Physics calculations are fixed, and they are generally fixed to a limit that DX can cope with, so mantle isn't going to make a massive difference as the render is designed to run with the DX weakness in mind.
That is, unless you have a very weak CPU, that very weak CPU will suddenly become a powerful CPU in Mantle BF4

What it shows, IMO, is the possible future, now Devs have much more CPU power to play with if they use Mantle. :)
Click to expand...

If that was the case, with me clocking my 3930K to 5Ghz would result in more frames but it didn't, it actually gave me lower frames. I am claiming this is purposefully crippling anything using DX to make Mantle look good. I don't have a problem with this at all and Oxide know what they are doing.
 
Gregster said:
If that was the case, with me clocking my 3930K to 5Ghz would result in more frames but it didn't, it actually gave me lower frames. I am claiming this is purposefully crippling anything using DX to make Mantle look good. I don't have a problem with this at all and Oxide know what they are doing.
Click to expand...


If DX is capable of 100 performance then overclocking your CPU to 300 performance DX will still only yield 100 performance, it may even reduce performance as DX is trying to deal with even more Physics handles that DX already can't deal with.
 
Last edited:
Gregster said:
If that was the case, with me clocking my 3930K to 5Ghz would result in more frames but it didn't, it actually gave me lower frames. I am claiming this is purposefully crippling anything using DX to make Mantle look good. I don't have a problem with this at all and Oxide know what they are doing.
Click to expand...

I agree on this.They crippling DX on purpose.So its easy to see the difference in tech,but its still pretty much unfair comparison.
 
Gregster said:
If that was the case, with me clocking my 3930K to 5Ghz would result in more frames but it didn't, it actually gave me lower frames. I am claiming this is purposefully crippling anything using DX to make Mantle look good. I don't have a problem with this at all and Oxide know what they are doing.
Click to expand...

Something is wrong as over clocking the CPU should bring more performance in DX if that is what is causing the bottleneck between the CPU & GPU. It appears gimped for DX to show the performance of Mantle. Or maybe coded badly on DX as it is a mantle title. Guess more will come to light over the course of time.
 
You guys are ignoring the definition of a "Bottleneck"

The CPU is the Bottle, the GPU is the Bucket, the bit in-between, the bit that has the opening to feed the bucket is DX, DX is the Bottles neck

The Bottles neck is the limiting factor, it does not matter how big the bottle (CPU) is, the opening in the bottle (DX) is what the contense of the bottle has to flow though. thats fixed and the flow limited to it.
 
humbug said:
You guys are ignoring the definition of a "Bottleneck"

The CPU is the Bottle, the GPU is the Bucket, the bit in-between, the bit that has the opening to feed the bucket is DX, DX is the Bottles neck

The Bottles neck is the limiting factor, it does not matter how big the bottle (CPU) is, the opening in the bottle (DX) is what the contense of the bottle has to flow though. thats fixed and the flow limited to it.
Click to expand...

And the link between the bucket and the bottle, the neck is governed by the pressure at which it flows between the two. I.e more pressure from the source will give a faster flow. :)
 
Gregster said:
If that was the case, with me clocking my 3930K to 5Ghz would result in more frames but it didn't, it actually gave me lower frames. I am claiming this is purposefully crippling anything using DX to make Mantle look good. I don't have a problem with this at all and Oxide know what they are doing.
Click to expand...

Did GPU usage go up? Could be a case of giving the cpu more work to do brought frames down but the gpu may have been rendering more. Worth a check.
 
No it was your analogy, so I explained in your terms.

What I am saying is that most other DX driven software sees an improvement in OC'ing the CPU, this does not. Why?
 
Stanners said:
No it was your analogy, so I explained in your terms.

What I am saying is that most other DX driven software sees an improvement in OC'ing the CPU, this does not. Why?
Click to expand...

Because this measures parallel handles through put, how many Physics calculations the API software can process at any given time, not how fast the CPU can process them.
 
AMD FX 8350 stock and Sapphire Tri-X R9 290x OC (1040/1300) stock

Dx Results
== 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: 17138696192
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: 10406

Average FPS: 28.90
Average Unit Count: 3861
Maximum Unit Count: 5572
Average Batches/MS: 470.16
Maximum Batches/MS: 884.30
Average Batch Count: 17680
Maximum Batch Count: 136611
===========================================================
lowest was 4.8 FPS and highest memory use 1510 MB from gpu-z

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: 17138696192
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: 17884

Average FPS: 49.68
Average Unit Count: 4381
Maximum Unit Count: 5466
Average Batches/MS: 882.06
Maximum Batches/MS: 1965.38
Average Batch Count: 21540
Maximum Batch Count: 172636
===========================================================
lowest was 11.2 FPS and highest memory use 2434 MB from gpu-z
 
eddyr said:
Did GPU usage go up? Could be a case of giving the cpu more work to do brought frames down but the gpu may have been rendering more. Worth a check.
Click to expand...

No it didn't, it still sat at 40% ish. 4.625Ghz is plenty to run a pair of Titans in any other game and keep them at 99% but not this Oxide demo. I am not fussed but does seem strange that only this one game does that.
 
drunkenmaster said:
You should get a better score without sli. It's cpu limited due to the amount of work DX is doing(ridiculously inefficiently), adding SLI, you're effectively trying to send out even more data by increasing the amount of work DX is doing. Higher cpu time for DX, less cpu time for the game.........

Thing about BF4 is, some speed gains, sure but Rroff, me, plenty of others, Oxide... Mantle's place to shine is in games which are actually trying to draw more than DX currently allows. Backporting it to a game already out is a good proof of concept and there are gains, ranging from small to huge on weaker hardware(which the majority of gamers have), also great.

But this Oxide game is the first really next gen engine capable of drawing well beyond what DX realistically can, and we can see clear as day that a better API is required for such levels of draw calls.

I wouldn't expect any optimisation that will allow this demo to run faster on SLI or Xfire under DX, because adding another card adds more work for DX, which simple adds additional overhead to a highly limited system.

Many completely cpu limited games show negative scaling with sli/xfire for that very reason.
Click to expand...

You do know I have no problems with BF4 performance since last patch, I'm getting consistent 120fps with vsync and significantly higher than that with vsync off. Obvious 120hz monitor.

As for star swarm demo, I do not believe it is optimised for DX and think if it was, NVidia users would see better results, see earlier in thread where slight changes were made to swarm giving NVidia big boosts in frame rates, don't get me wrong though, I do believe Mantle is a step in the right direction and is more efficient than DX and this is a good thing for all, it may well lead to better versions of DX.
I do understand mantle API is capable of higher draw calls and any future games benefiting from that will also be optimised for DX with little or no difference to the way the game looks. DX wont stand still because of Mantle though and could be more efficient when it's needed. DX12

SLI / Crossfire will be coming in a future update so we'll see if the demo runs faster or not.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom