Star Swarm - Oxide bench thread

_Alatar_ · 12 Jun 2014 at 11:02

LtMatt said:
Has anyone found a consistent and reliable way to bench this yet?

Yes. Custom scenarios.

Easy to see when a run is bugged (*cough* humbug *cough*) and otherwise most runs are within 3fps of eachother due to the consistent camera angle and identical ship squadron spawns.

LtMatt · 12 Jun 2014 at 11:05

_Alatar_ said:
Yes. Custom scenarios.

Easy to see when a run is bugged (*cough* humbug *cough*) and otherwise most runs are within 3fps of eachother due to the consistent camera angle and identical ship squadron spawns.

Where can you find these custom scenarios? Or do you mean you just select custom in the menu?

andybird123 · 12 Jun 2014 at 11:07

DamnedLife said:
Skip this one Get this one: http://forums.overclockers.co.uk/showpost.php?p=25762826&postcount=94

GUYS I think I'm on the right track!

So What I did is: The camera doesn't change to any unit but stay at the exact same unit during whole run, the camere MAY be still random (2 runs matche camera movement but help my find if it is yrue) but I think I'm close to figuring that one out too!!

MY SCENARIO CUSTOM .csv FILE! OPEN WITH NOTEPAD AND COPY PASTE

Command,Subject,Player Number,Start Time,Max Reps,Unit Rep Rate,Unit Spawn Rate (the time between units spawned into this squadron once it has materialized (in seconds)),X,Y,z
SpawnBackgroundObject,TheCosmos,,0,1,1,0,0,0,0
SpawnBackgroundObject,Planet,,0,1,1,0,0,-12000,-2300
SpawnBackgroundObject,PlanetAtmosphere,,0,1,1,0,0,-12000,-2300
SpawnBackgroundObject,StarSphere,,0,1,1,0,0,0,0
SpawnBackgroundObject,StarSphere,,0,1,1,0,0,0,0
SpawnBackgroundObject,StarSphere,,0,1,1,0,0,0,0
SpawnBackgroundObject,StarSphere,,0,1,1,0,0,0,0
SpawnBackgroundObject,StarSphere,,0,1,1,0,0,0,0
SpawnBackgroundObject,TheSuns,,0,1,1,0,0,0,0
CameraToUnit,,,0.1,1,0,0,0,0,0
CameraToRandomAngle,,,0.1,1,0,0,,,
CameraToRandomUnit,,,9999999,1,0,0,,,
SetGrids,,,0,1,1,0,0,0,0
SpawnSquadron,Bench FighterA1-0-0,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-1-0,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-2-0,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-3-0,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-4-0,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-5-0,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-6-0,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-7-0,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-0-1,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-1-1,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-2-1,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-3-1,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-4-1,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-5-1,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-6-1,0,0,3,12,0,,,
SpawnSquadron,Bench FighterA1-7-1,0,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-0-0,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-1-0,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-2-0,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-3-0,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-4-0,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-5-0,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-6-0,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-7-0,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-0-1,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-1-1,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-2-1,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-3-1,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-4-1,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-5-1,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-6-1,1,0,3,12,0,,,
SpawnSquadron,Bench FighterB1-7-1,1,0,3,12,0,,,
SpawnSquadron,Bench BomberA1-0,0,0,24,100,0.2,,,
SpawnSquadron,Bench BomberA1-1,0,0,24,100,0.2,,,
SpawnSquadron,Bench BomberB1-0,1,0,8,200,0.2,,,
SpawnSquadron,Bench BomberB1-1,1,0,8,200,0.2,,,
SpawnSquadron,Bench CarrierA1-0,0,0,1,0,0,,,
SpawnSquadron,Bench CarrierA1-1,0,0,1,0,0,,,
SpawnSquadron,Bench CarrierB1-0,1,0,1,0,0,,,
SpawnSquadron,Bench CarrierB1-1,1,0,1,0,0,,,
SpawnSquadron,Stress00,0,1,250,12,0.01,,,
SpawnSquadron,Stress01,1,1,250,12,0.01,,,
SpawnSquadron,Stress02,1,1,250,12,0.01,,,
SpawnSquadron,Stress03,0,1,250,12,0.01,,,
SpawnSquadron,Stress10,0,1,250,12,0.01,,,
SpawnSquadron,Stress11,1,1,250,12,0.01,,,
SpawnSquadron,Stress12,1,1,250,12,0.01,,,
SpawnSquadron,Stress13,0,1,250,12,0.01,,,
ResetScenario,,,2700,99999999,,,,,

in this thread Mathew

if you remember, Greg tried to get everyone to use a custom scenario by posting it in the OP, but then people complained that it was too difficult so Gregster was forced to change it back to one of the standard ones

LtMatt · 12 Jun 2014 at 11:10

andybird123 said:
in this thread Mathew

if you remember, Greg tried to get everyone to use a custom scenario by posting it in the OP, but then people complained that it was too difficult so Gregster was forced to change it back to one of the standard ones

Cheers. Has anyone posted results using this apart from Diamond?

andybird123 · 12 Jun 2014 at 11:15

Yes, me and Greg did as well, then everyone got uppity that Greg wasn't posting the results because people were posting Follow instead of Custom

LtMatt · 12 Jun 2014 at 11:16

andybird123 said:
Yes, me and Greg did as well, then everyone got uppity that Greg wasn't posting the results because people were posting Follow instead of Custom

Got a link to them?

Gregster · 12 Jun 2014 at 11:25

Rules are too hard and I took too much stick, so I will leave this thread as it is. Run it and put "official entry" and your score will be added.

_Alatar_ · 12 Jun 2014 at 11:46

andybird123 said:
in this thread Mathew

if you remember, Greg tried to get everyone to use a custom scenario by posting it in the OP, but then people complained that it was too difficult so Gregster was forced to change it back to one of the standard ones

Don't use that, this: http://forums.overclockers.co.uk/showpost.php?p=25762826&postcount=94

is the best one.

andybird123 · 12 Jun 2014 at 11:47

LtMatt said:
Got a link to them?

nope, can't find it, but if I get a chance later I'll re-download and run it again

I'll also set it to run for much longer and see how that affects things

_Alatar_ said:
Don't use that, this: http://forums.overclockers.co.uk/showpost.php?p=25762826&postcount=94

is the best one.

thanks!

LtMatt · 12 Jun 2014 at 12:00

_Alatar_ said:
Don't use that, this: http://forums.overclockers.co.uk/showpost.php?p=25762826&postcount=94

is the best one.

What score do you get using that and what clocks?

_Alatar_ · 12 Jun 2014 at 12:07

Can't remember exactly, I'll run it later. Though I only have stock cooling on my CPU atm...

I think with a stock 4770K and an OC'd Titan I was in the high 50s. Timed run extreme settings ofc.

that scenario is more stressful than the standard follow mode.

LtMatt · 12 Jun 2014 at 12:09

_Alatar_ said:
Can't remember exactly, I'll run it later. Though I only have stock cooling on my CPU atm...

I think with a stock 4770K and an OC'd Titan I was in the high 50s. Timed run extreme settings ofc.

that scenario is more stressful than the standard follow mode.

This run has no motion blur though...

That was the most demanding part, it renders the scene 9 times.

LtMatt · 12 Jun 2014 at 12:17

DamnedLife's results - Custom CSV

Average FPS: 22.80
Average Unit Count: 3126
Maximum Unit Count: 4551
Average Batches/MS: 662.45
Maximum Batches/MS: 838.84
Average Batch Count: 31417
Maximum Batch Count: 71644

This is still not consistent at all.

== Results ================================================
Test Duration: 360 Seconds
Total Frames: 16355

Average FPS: 45.43
Average Unit Count: 4230
Maximum Unit Count: 5755
Average Batches/MS: 1548.61
Maximum Batches/MS: 3130.04
Average Batch Count: 34420
Maximum Batch Count: 90209
===========================================================

That is using that custom.csv. Look at how much more units i have, how much more batches i have. Still as inconsistent as ever.

andybird123 · 12 Jun 2014 at 12:35

LtMatt said:
This run has no motion blur though...

That was the most demanding part, it renders the scene 9 times.

extreme settings still uses motion blur - that is in settings, not the scenario file

I've just run DL's scenario 3 times and gotten 71-72fps all 3 times with average batches around 22k

LtMatt · 12 Jun 2014 at 12:37

andybird123 said:
extreme settings still uses motion blur - that is in settings, not the scenario file

I've just run DL's scenario 3 times and gotten 71-72fps all 3 times with average batches around 22k

It does not, not like the original extreme follow mode. I just sat and watched the two different modes and there is no motion blur as the camera does not move. I give up with this bench.

_Alatar_ · 12 Jun 2014 at 12:40

LtMatt said:
It does not, not like the original extreme follow mode. I just sat and watched the two different modes and there is no motion blur as the camera does not move. I give up with this.

The moving ships are still blurred just like they are on follow mode. Just because you can't see the effect being hugely exaggerated due to fast camera movements doesn't mean that it's not there.

And to top it off this is harder on the CPU than follow. Meaning that mantle and CPU overhead improvements in general will see bigger gains here than in follow.

And as others have pointed out the results using that file are extremely consistent.

LtMatt · 12 Jun 2014 at 12:42

_Alatar_ said:
The moving ships are still blurred just like they are on follow mode. Just because you can't see the effect being hugely exaggerated due to fast camera movements doesn't mean that it's not there.

And to top it off this is harder on the CPU than follow. Meaning that mantle and CPU overhead improvements in general will see bigger gains here than in follow.

And as others have pointed out the results using that file are extremely consistent.

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.

triss · 12 Jun 2014 at 12:47

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.

So what your saying is the whole benchmark was just set up in mantles favour ? seems pointless tbh,
Who were they bench marking against if its purposely set up in mantles favour?

LtMatt · 12 Jun 2014 at 12:49

triss said:
So what your saying is the whole benchmark was just set up in mantles favour ? seems pointless tbh,
Who were they bench marking against if its purposely set up in mantles favour?

That's the whole idea of it triss. It's not even a benchmark, it's a tech demo designed to show how much better Mantle can handle high draw calls than DirectX. Nothing more, nothing less.

triss · 12 Jun 2014 at 12:51

LtMatt said:
That's the whole idea of it triss. It's not even a benchmark, it's a tech demo designed to show how much better Mantle can handle high draw calls than DirectX. Nothing more, nothing less.

Ah ok thanks matt