Benchmarked Dxtory/FFsplit/Xsplit

[LlamaNL]

--- FFSplit / Xsplit Encode settings ---
5000
Superfast
1024x576
30f fps
192 Stereo Audio (disabled)

PC Specs:
AMD Phenom II 960T Unlocked 6 Cores @ 3.2Ghz
AMD HD 6850 Base
8GB DDR3

--- Half Life 2: Lost Coast Benchmark Tool ---

Everything maxed, no vsync, Average FPS:
200,96 - Base - Fullscreen
181,40 - Base - Windowed

***FFSPLIT***

DXTORY:
FPS - Capture > Output - Screen state
119,45 - 1080p > 1080p - Fullscreen
117,64 - 1080p > 1080p - Fullscreen + Aero
89,83 - 1080p > 720p - Fullscreen + Aero
114,09 - 720p > 720p - Fullscreen + Aero
110,95 - 576p (same across dxtory and ffsplit) - Fullscreen + Aero

--- DXTORY Notes ---
DXTORY scaling causes massive frame drops so have it capture at the game resolution and select 100% at the scaling section.

Windowed mode with FFsplit Overlay:
148,00 - 1080p - Windowed + Aero
82,37 - 1080p - Windowed

--- FFSplit Overlay Notes ---
While it gives 30 frames per second more the screen stutters like crazy and has horizontal tearing. (might just be this benchmark)

***XSPLIT QUALITY 10***

DXTORY with optimal settings as derived from FFSplit testing
100,15 - 1080p > 1080p - Fullscreen
100,78 - 1080p > 1080p - Fullscreen + Aero

Windowed Mode Screen Region:
135,83 - 1080p - Windowed + Aero
93,47 - 1080p - Windowed

***XSPLIT QUALITY 5***

DXTORY with optimal settings as derived from FFSplit testing
105,05 - 1080p > 1080p - Fullscreen
98,00 - 1080p > 1080p - Fullscreen + Aero

Windowed Mode Screen Region:
131,38 - 1080p - Windowed + Aero
86,92 - 1080p - Windowed

--- Conclusion ---
Optimaly as few scaling step as possible, as closely to the encoder source as possible. Capture at maximum resolution with DXTORY and let ffsplit scale. Although FFsplit Overlay provides a much higher fps it also gives a rather irritating screen stutter and a LOT of horizontal screen tearing due to lack of vsync because of windowed mode. In my opinion use Screen region when you want the maximum amount of game performance and will tolerate a lot of visual noise. None of that will generally translate to the stream. In any other case use DXTORY for nice things such as no horizontal tearing and much smoother looking frame rates.

---XSplit vs FFsplit---
Framerates are 20% better overall for FFSplit which i think is caused by the preview window that xsplit uses. At the moment i'd still recommend XSplit over FFSplit because of the many many many quirks that ffsplit has where it will randomly stop streaming or decreases stream frame rate for no reason at all.

--- General Streaming Weirdness ---
Xsplit quality has no bearing on framerate because all results are well withing a 5% error margin.


Here's a graph to make sense of this mess of numbers!


EDIT: I learned after making this post that Aero caps the desktop framerate at 22fps and thus also caps capture at 22fps. All the comparisons between Aero / Basic are now moot. But that means DXTory/Gamesource is now always the better choice for both framerate and visual fidelity

[MSgtGunny]

one thing to know about aero plus screen region is that having aero on actually makes the screen region capture max out at around 18-22 fps meaning you get better in game performance because you aren't capturing as many fps, BUT it makes the stream more stuttery. comparing the 93 to the 135 fps with xsplit screen region yields about a 45% increase in in game frame rate at the cost of about a 33% drop in your screen captures framerate (more if you are trying to do 60 fps). The main benefit xsplit provides is that it has a very broad range of capabilities, and these are constantly being refined and added to. Sure some updates are bad for a small subset of the users, but thats usually due to their software configuration (usually drivers and registry/other programs/directshow filter config is the issue).

So a 20% drop in frames (regarding dxtory) for 100X more features is a great trade off for me personally.

Also to people reading this note these numbers are specific to this user's configuration, but if other people would like to run the same tests with their own setup, by all means.


Also know that reducing the quality isn't supposed to have an affect on your cpu usage/performance. setting a lower quality tells the encoder to use less bitrate in scenes that don't need it but in high motion where you would hit your bitrate limit with any quality option, then there is no actual quality difference. Setting a higher quality just gives better quality in low motion scenes at the expense of higher bitrates during that time (which may be a bad thing if your connection has a monthly cap from your isp)