DirectX 10.1 Clear Sky personal benchmarks

[daletkine]

I heard that Clear Sky added directX 10.1, don't really know if it looks better, but since I have DirectX 10.1 card and Stalker Clear Sky, I was curious how it performs. THere are no reviews out there on the performance for this setup, so I applied 1.5.07 patch and run a couple of benchmarks with Fraps. Post your results here as well, or if you find an article with performance of DirectX 10.1, post it.

Setup: I am using ATI Catalyst 8.11 drivers. 1024*768 res, under advanced tab all sliders set to max settings, all effects enabled, the level is NPP power plant, using a quicksave from when you are dropped to kill Strelok and given new guns, I start benchmark before camera zooms in on Strelok, then walk back on the bridge, look right at power plant, look left at entrance to the map, go down on the sloped columns, go towards fighting monolith and clear sky, zoom in, shoot someone, then keep moving towards the firefight. Here are the results:

DirectX 10, no AA:
Frames 4066
Time (ms) 60000
Min 31
Max 118
Avg (fps) 67.77

DirectX 10, 4X AA:
Frames 3591
Time (ms) 60000
Min 31
Max 98
Avg 59.85

DirectX 10.1, no AA:
Frames 3831
Time (ms) 60000
Min 33
Max 109
Avg 63.85

DirectX 10.1, 4X AA:
Frames 3880
Time (ms) 60000
Min 33
Max 116
Avg 64.67


Hmm, the only interesting thing that jumps at me is that Minimum fps went from 31 to 33 with transition from 10.0 to 10.1, so it seems to yield faster performance. I didn't notice NPP area looking any different on either version. This is done on Vista x64 with 8GB RAM. My specs should be below.

 

[error8]

Great work you've done here, but why 1024X768??? You have a 15 inch CRT or something?

 

[cmdrdredd]

Maybe because he likes to keep the FPS near 60? Some people have a soft limit where they will turn down the rez 1 notch, or turn down AA to keep the FPS going.

 

[daletkine]

Hehe, no I have a 24". Resolution makes little difference in Stalker, I feel the game still doesn't look very good. Well, compared to Crysis that is.
Also, I decided to run different resolution tests and the results are completely all over the place. Here they are:


Resolution: 1280*800

DirectX 10, no AA:
Min 20
Max 49
Avg (fps) 31.45

DirectX 10, 4X AA:
Min 21
Max 54
Avg 32.95

DirectX 10.1, no AA:
Min 19
Max 51
Avg 32.25

DirectX 10.1, 4X AA:
Min 21
Max 51
Avg 32.17


Resolution: 1440*900

DirectX 10, no AA:
Min 27
Max 76
Avg (fps) 45.62

DirectX 10, 4X AA:
Min 26
Max 89
Avg 47.17

DirectX 10.1, no AA:
Min 24
Max 74
Avg 42.63

DirectX 10.1, 4X AA:
Min 26
Max 82
Avg 39.82

Resolution: 1680*1050

DirectX 10, no AA:
Min 24
Max 63
Avg (fps) 38.22

DirectX 10, 4X AA:
Min 21
Max 61
Avg 35.75

DirectX 10.1, no AA:
Min 22
Max 68
Avg 37.12

DirectX 10.1, 4X AA:
Min 19
Max 60
Avg 34.45


Resolution: 1920*1200

DirectX 10, no AA:
Min 23
Max 53
Avg (fps) 34.08

DirectX 10, 4X AA:
Min 23
Max 58
Avg 37.38

DirectX 10.1, no AA:
Min 21
Max 56
Avg 36.1

DirectX 10.1, 4X AA:
Min 24
Max 51
Avg 34.77


So there. A couple of things to note: 1) I ran all these tests only once, while official reviews usually run the test 5 times and then average. I didn't want to bother. 2) The variation is quite wild here. The resolution that is the slowest here is the 1280*800, even slower than 1920*1200, on all the settings! This probably suggests some gimmick with the game or the drivers. 3) In 1440*900 and 1680*1050 resolutions I noticed that the game artifacted pretty heavily (during rain), but not on the first runs on those resolutions (DX10 and no AA), only on the rest 3 runs. Also, this problem came up on another forum here at Anandtech - someone complained their GTX 260 was artifacting during rain, but not their 8800 GTX. I am not overclocking videocard beyond it's factory stock and I am not experiencing any other artifacts in any other games, so, it's a Clear Sky glitch.
Ohh, well, all resolutions are fairly playable.

 

[thilanliyan]

Originally posted by: daletkine
So there. A couple of things to note: 1) I ran all these tests only once, while official reviews usually run the test 5 times and then average. I didn't want to bother.

The results might be all over the place because you only ran it once (level loading and such comes into play I think).

 

[MarcVenice]

Game doesn't look very good? Wow, you're a first tbh. I love the look of the game, it creates an atmosphere that's really immersive. Crysis might look better on screenshots, I really prefer the way clear sky uses it's graphics to create atmosphere, instead of just pretty pictures. It has to be said though, the game does have quite a few hiccups with all sorts of videocards.

 

[Sylvanas]

The game is great on my 4870X2 in DX10.1, I have noticed though that it is VERY CPU dependant. As I OC'd my rig in sig I saw at least a 5fps min difference between 2.8ghz and 3.4ghz.

 

[BFG10K]

How is it some 4xAA benchmarks are faster than no AA? Have you verified AA is working with screenshots?

I don't think you're getting any AA at in all those situations but if you are, there's probably something wrong with your test methods.

 

[daletkine]

Originally posted by: thilan29

Originally posted by: daletkine
So there. A couple of things to note: 1) I ran all these tests only once, while official reviews usually run the test 5 times and then average. I didn't want to bother.

The results might be all over the place because you only ran it once (level loading and such comes into play I think).

Hey again, you're right, the results are like that cause I ran it only once.

Originally posted by: BFG10K
How is it some 4xAA benchmarks are faster than no AA? Have you verified AA is working with screenshots?

I don't think you're getting any AA at in all those situations but if you are, there's probably something wrong with your test methods.

Hey, no I didn't verify that AA is working with screenshots. I was hoping performance difference would be enough, but there wasn't any. And you're right, my methods were kind of off. I also realized after restarting the game that the game will save those AA settings, but it won't use them until you restart the game - but it doesn't tell me that when I hit 'use'. Anyway, so that's why 4X AA took no performance hit, since I ran it without restarting the game.
So, I'll try to re-run the benchmarks when I have a bit more time during the week. For now, I created a special save point near a wall in the same level, I'll benchmark by walking towards the power plant in a straight line. 60sec, takes me through the whole field, with fighting going on on the right, no obstacles in the path, except at the last second I bump into the crashed helicopter right in front of the power plant. This should be pretty repeatable, so that my benchmark is not affected by how I turn or how much sky is visible on screen.
I used Fraps again and ran DX10.1, all highest settings, no AA, 1920*1200, 5 times, then get the average for all values. Here are the results from 5 runs and their average. I think this is satisfactory. And yes, before running other tests, I will restart the game.


Run 1:
Min 22
Max 38
Avg 29.73

Run 2:
Min 22
Max 39
Avg 29.97

Run 3:
Min 24
Max 41
Avg 31.23

Run 4:
Min 24
Max 41
Avg 31.37

Run 5:
Min 24
Max 39
Avg 31.45

Average across 5 runs:
Min 23.2
Max 39.6
Avg 30.75


So, how does this look? Pretty repeatable? Any other ideas what I should do before I re-run my benches? I think I'll only do 2 resolutions though, since anything more would take up a crazy amount of time.


By the way, Sylvanas, I see you have that water cooling set up. I was considering something very close to yours, except only one MCW 60, all else the same as yours. How is that working for you? Do you find two pumps overkill? Ohh, wait, you have 3 pumps!? ^_^ Did you find one or two not enough?

Thanks for everybody's interest!

 

[daletkine]

Ok, I completed the benchmarks again. I chose 2 resolutions and benchmarked each run 5 times, walking diagonally from cement perimeter wall to the NPP power plant. Here are my results:

Resolution: 1680*1050

DirectX 10, no AA:
Min 25.8
Max 45.6
Avg (fps) 34

DirectX 10, 4X AA:
Min 8.6
Max 22.4
Avg 16.44

DirectX 10.1, no AA:
Min 24.6
Max 43.8
Avg 33

DirectX 10.1, 4X AA:
Min 9
Max 22
Avg 16.03


Resolution: 1920*1200

DirectX 10, no AA:
Min 22.2
Max 39.6
Avg (fps) 30.29

DirectX 10, 4X AA:
Min 8
Max 19
Avg 13.44

DirectX 10.1, no AA:
Min 23.2
Max 39.6
Avg 30.75

DirectX 10.1, 4X AA:
Min 6.6
Max 14.2
Avg 9.4