The OFFICIAL FFXIV Benchmark & Beta Performance Thread

[Gokulo]

oh (general question), does windowed mode require more resources then full screen? I had no idea..

Yep, it does. Though I think someone else can elaborate on it more and better than me

[Celeras]

It doesn't, unless you're doing silly things like running Aero transparency or similar modifications with bad hardware or something. Flashing ray tracers on window borders, etc. Otherwise there's no difference between windowed and full screen.. they should perform identically. One is being hardware rendered and the other software if you want to get technical about it, but it won't change much.

[Katlan]

ahh ok, i didnt think it did but that post made me think otherwise.

[Celeras]

I mean.. it could make a difference. But it'd be at fault of the user, not windowed mode.

[Kyrra]

I mean.. it could make a difference. But it'd be at fault of the user, not windowed mode.

Default settings in Windows will make it have worse performance.

[Celeras]

Default settings in Windows will make it have worse performance.

I'd consider that user error, if your hardware can't handle Aero you shouldn't be using it.

[Shaard]

Would having those small things make THAT big of a difference for your score?

Doesn't seem like small and big icons would.

Yeah, I was very surprised but I just tested, and it is exactly the reason why my score dropped from 4000 to 2700. Weird isn't it?

Big icons = no issue at all, even when moving the task bar on and off the benchmark screen.

Small icons = instant drop when the task bar overlaps the screen (it shows clearly on the graph at the bottom left corner: instant drop and doesn't come back up as high unless I hide the bar again).

But that's not an issue anymore, since all I need to do is avoid small icons

[Katlan]

I'd consider that user error, if your hardware can't handle Aero you shouldn't be using it.

my hardware is fine, but what default settings are you talking about? just overall curius since i learn new stuff every day

[Celeras]

Afaik he was referring to Aero.

[Kurokikaze]

I'm pretty sure he means having Aero Glass on (transparent window borders) and all the other graphical changes that were introduced with Windows Vista/7.

As mentioned earlier in the thread though... If Aero Glass is giving you a noticeable performance difference - Aero Glass isn't your problem.

[Katlan]

Afaik he was referring to Aero.

oh.. lol, that makes a 25 point difference on my test... not too worried about it. Thought there was something big that I wouldn't know about since my lv of computer knowledge goes up by the day

[alecrast]

It doesn't, unless you're doing silly things like running Aero transparency or similar modifications with bad hardware or something. Flashing ray tracers on window borders, etc. Otherwise there's no difference between windowed and full screen.. they should perform identically. One is being hardware rendered and the other software if you want to get technical about it, but it won't change much.

I believe it does make a difference, whether you've got bells and whistles turned off or not. In windowed mode, applications render to an off-screen location and call Present(), then DWM is responsible for mapping those off-screen renders onto the Direct3D surface (the desktop IS the D3D surface consisting of a couple triangles) whereas in full-screen mode, the application renders directly. If you attempt to use DirectDraw, DWM gets disabled. I am guessing that this is part of why Crossfire only works in full-screen mode. So it would seem that drivers are still optimized towards applications that run in full-screen mode, not to mention that games often take control of your screen in strange and profound ways...

There are differences between how WDDM 1.0 and 1.1 work and I am not keen on all the changes but I would be shocked if Crysis at 1920x1200 windowed ran as smoothly as Crysis at 1920x1200 full-screen. I would benchmark it but I have my gaming rig torn apart at the moment. I will revisit this, however. The proof is in the pudding, obviously, so I'll start cooking when I have my system back in a usable state. Past experience has taught me, however, that when all settings are equal, full-screen runs better.

Yeah, I was very surprised but I just tested, and it is exactly the reason why my score dropped from 4000 to 2700. Weird isn't it?

Big icons = no issue at all, even when moving the task bar on and off the benchmark screen.

Small icons = instant drop when the task bar overlaps the screen (it shows clearly on the graph at the bottom left corner: instant drop and doesn't come back up as high unless I hide the bar again).

But that's not an issue anymore, since all I need to do is avoid small icons

That makes me wonder if "small icons" == DPI scaling rather than actual small icons. I didn't think that was the case but who knows.

[Celeras]

There are differences between how WDDM 1.0 and 1.1 work and I am not keen on all the changes but I would be shocked if Crysis at 1920x1200 windowed ran as smoothly as Crysis at 1920x1200 full-screen.

Ask and ye shall receive ;p Maxed out both settings, windowed only difference. Can't really get much closer than a .17 differential .. and windowed was actually the higher fluctuation. Aero/All bells enabled.

http://img338.imageshack.us/img338/2...ysiswindow.jpg

[alecrast]

Ask and ye shall receive ;p Maxed out both settings, windowed only difference. Can't really get much closer than a .17 differential .. and windowed was actually the higher fluctuation. Aero/All bells enabled.

Interesting. Do you get the same result when running in DX9 mode? DX10 and DX11 have specific optimizations that DX9 doesn't have, if I remember correctly. I could have sworn I read that in one of Greg Schechter's blogs on MSDN.

[Lovian]

Oppps i just bought my self a G73 asus notebook.

Its pretty damn fast and loving it so far. Just run the benchmark and got just under 4000 on low and just over 2000 on high, this is before i get around to formating and updating all the drivers. I'll post screen shorts later on.

[Celeras]

Interesting. Do you get the same result when running in DX9 mode? DX10 and DX11 have specific optimizations that DX9 doesn't have, if I remember correctly. I could have sworn I read that in one of Greg Schechter's blogs on MSDN.

Same with DX9 yeah.

[Warf]

I can't post pics yet I guess. But here are my specs. Thought it'd do a bit better, but shouldn't have any problems running the game.

i7 930 OC'd to 3.2
P6X58D-E Mobo
6Gb G.Skill Ripjaws 1600 in triple channel (stock speeds are 8-8-8-24)
EVGA GTX 470 OC'd a bit
Win 7 Home Premium 64bit
1TB WD Caviar Black 7200rpm 32Mb cache HDD
Antec true power 620w PSU
Wrapped in CM HAF 922 case

Low score Stock - 5223
Low score OC'd - 5521

High score stock - 3259
High score OC'd - 3581

Stock cooling, so I don't think I'll take my CPU past 3.5, any tips for increasing performance? Though reaching the next bracket of scoring for high or low is a stretch.

[Pyrix]

Well, I got a score of a bit above 1800, normal resolution, with a good framerate, so I think I'll be pretty decently well off.

[Rynix]

Cant post picture because im new...

Windows Vista 64-Bit
Intel Quad-Core 2.6Ghz Yorkfield
4GB RAM
Dual Card SLI nVidia 9800 GTX

Score: 1650

Any ideas why this is so low?!?! Any feedback is appreciated. Thanks.

[Zeh]

SLI is borked at the moment. That's about right for CPU and single GPU performance.

High:

Spoiler: show

http://img52.imageshack.us/img52/391/ffxivhigh.jpg

Low:

Spoiler: show

http://img180.imageshack.us/img180/9177/ffxivlow.jpg

AMD Phenom II X4 955 @ 3.515ghz
Sapphire 4870 HD

Which is actually pretty decent, considering I'm using a 1680x1050 monitor at the moment. Also taking final optimizations into account, I believe I'll be fine. 1920x1080 was a bit choppy, so 1680x1050 should be manageable... and at this rate, I don't believe I'll be gutting out my 4870. I'll save my money for my next ultra-high end PC project