The OFFICIAL FFXIV Benchmark & Beta Performance Thread

[shirasoma]

Also alecrast, what software are you using to record your FPS findings? I'm biting the bullet and testing this on my own machine as well, just to give another data-point. I have an AMD CPU also, but I have an ATI card, so I'm gonna see that bears any factor.

[alecrast]

Also alecrast, what software are you using to record your FPS findings? I'm biting the bullet and testing this on my own machine as well, just to give another data-point. I have an AMD CPU also, but I have an ATI card, so I'm gonna see that bears any factor.

I also tested against a Sapphire 5870 2GB Toxic edition card. Same thing, just with a slightly higher framerate. It should also be noted that to keep the tests fair, the weather has to be the same. Rain and wind affect GPU usage enough to throw off results. And it likes to rain in Limsa... a lot. I don't know how many times I had to start my benchmarking runs over again because I would get out of the city only to discover it raining.

I used MSI Afterburner 2.0 for GPU usage and framerate. You could also use eVGA Precision, probably (if I'm thinking of the right tool), though I've never tried it. I believe they're both built on rivatuner.

[shirasoma]

I also tested against a Sapphire 5870 2GB Toxic edition card. Same thing, just which a slightly higher framerate.

I used MSI Afterburner 2.0 for GPU usage and framerate. You could also use eVGA Precision, probably (if I'm thinking of the right tool), though I've never tried it. I believe they're both built on rivatuner.

Good thinking, you really went the extra mile. Thanks for that info. I do have MSI Afterburner so I'll use that.

It's hard to make sense out of this. The game is clearly getting CPU bottlenecked in someway, but it seems absurd that top-end machines would be getting this problem. I remember Celeras' pic from earlier from the thread, showing 20 FPS in Ul'Dah with a Thuban X6 and GTX 470, you would think that's plenty enough power to not get bottlenecked in anyway. Conversely I have friends in the Nehalem Intel camp getting the exact same problems.

Also IIRC there's an option to enable "Dust Effects" in the in-game menu, under Configuration. Whether that means what it means, or it's another synonym for weather, I don't know. My guess is the former though.

[alecrast]

Good thinking, you really went the extra mile. Thanks for that info. I do have MSI Afterburner so I'll use that.

It's hard to make sense out of this. The game is clearly getting CPU bottlenecked in someway, but it seems absurd that top-end machines would be getting this problem. I remember Celeras' pic from earlier from the thread, showing 20 FPS in Ul'Dah with a Thuban X6 and GTX 470, you would think that's plenty enough power to not get bottlenecked in anyway. Conversely I have friends in the Nehalem Intel camp getting the exact same problems.

Also IIRC there's an option to enable "Dust Effects" in the in-game menu, under Configuration. Whether that means what it means, or it's another synonym for weather, I don't know. My guess is the former though.

I just think there's a lot of debugging code in there that is artificially limiting the speed at which the game can run. If you take a look at the final graph here (my i7 860 at 4GHz) you can see some interesting performance data that does seem to indicate CPU bottleneck. Compare the i7 860 at 2.8GHz with 2 cores enabled versus the i7 860 at 4.0GHz with 4 cores enabled. When there are only 2 cores, the GPU usage and framerate both sag around the edges and middle of the graph (Adventurer's Guild and Bearded Rock respectively) very significantly. In "dual-core" mode, the i7 860 (even at 4GHz which is the second graph) cannot squeeze performance out of the GPU. Now look at the 4-core 4GHz graph - while the edges still sag, you can see that CPU utilization rises across all cores around Bearded Rock (the very middle of the graph) and GPU usage rises significantly as the GPU tries to maintain the 60fps cap with reasonable success.

From what I've seen, I think they can do a lot better with the client and I believe that they will - it's just a matter of how long it takes them to really optimize performance. I think some of it has to do with threads waiting on each other but my primary domain is UNIX so there's only so much useful data I can squeeze out of SysInternals without spending way more time than I want to; I'd rather be playing the game. :-p

[alecrast]

double post

[alecrast]

triple post >_<

[Rammses]

Good thinking, you really went the extra mile. Thanks for that info. I do have MSI Afterburner so I'll use that.

It's hard to make sense out of this. The game is clearly getting CPU bottlenecked in someway, but it seems absurd that top-end machines would be getting this problem. I remember Celeras' pic from earlier from the thread, showing 20 FPS in Ul'Dah with a Thuban X6 and GTX 470, you would think that's plenty enough power to not get bottlenecked in anyway. Conversely I have friends in the Nehalem Intel camp getting the exact same problems.

Also IIRC there's an option to enable "Dust Effects" in the in-game menu, under Configuration. Whether that means what it means, or it's another synonym for weather, I don't know. My guess is the former though.

Well, I could take screenshots of 11 fps with my Athlon II X3 @3.5 Ghz, and I frequently dip well below 20 in cities. 20 fps with a Phenom II X6 is within the realm of possibilities - even if the cities are CPU bottlenecked - unfortunately. X6 is not supposed to perform much better than Phenom tri- and quadcores in most gaming applications anyway. What would be useful however is a confirmation that even high-end Intel processors are dealing with the same terrible performance in cities.

Also, what framerate are the people who've ghettoed FFXIV (lowering buffer size to SNES equivalent etc) to make it run on old computers/laptops, experiencing in the cities? Cause they should REALLY have it rough if the cities are so extremely CPU reliant.

[shirasoma]

I just think there's a lot of debugging code in there that is artificially limiting the speed at which the game can run. If you take a look at the final graph (url snip) you can see some interesting performance data that does seem to indicate CPU bottleneck. Compare the i7 860 at 2.8GHz with 2 cores enabled versus the i7 860 at 4.0GHz with 4 cores enabled. When there are only 2 cores, the GPU usage and framerate both sag around the edges and middle of the graph (Adventurer's Guild and Bearded Rock respectively) very significantly. In "dual-core" mode, the i7 860 (even at 4GHz which is the second graph) cannot squeeze performance out of the GPU. Now look at the 4-core 4GHz graph - while the edges still sag, you can see that CPU utilization rises across all cores around Bearded Rock (the very middle of the graph) and GPU usage rises significantly as the GPU tries to maintain the 60fps cap with reasonable success.

From what I've seen, I think they can do a lot better with the client and I believe that they will - it's just a matter of how long it takes them to really optimize performance. I think some of it has to do with threads waiting on each other but my primary domain is UNIX so there's only so much useful data I can squeeze out of SysInternals without spending way more time than I want to; I'd rather be playing the game. :-p

Nice data. Based on your graphs, currently it looks like any form of Intel Nehalem or better running past 3.5 Ghz is "necessary" to get what I consider a smooth experience from this game. Which is really a shame, since a majority of people will definitely not be able to reach this kind of performance probably until a couple or more years down the line.

But like you said, this may not be indicative of final performance down the road. So maybe a Core i5/i7 @ 3.5 or higher would be a "worst case scenario." I'd like to be optimistic about it, but yes, it probably will come down to how "long" it will take to "improve" these issues. And that time period is totally unknown right now, if and when it actually happens.

[shirasoma]

Well, I could take screenshots of 11 fps with my Athlon II X3 @3.5 Ghz, and I frequently dip well below 20 in cities. 20 fps with a Phenom II X6 is within the realm of possibilities - even if the cities are CPU bottlenecked - unfortunately. X6 is not supposed to perform much better than Phenom tri- and quadcores in most gaming applications anyway. What would be useful however is a confirmation that even high-end Intel processors are dealing with the same terrible performance in cities.

Also, what framerate are the people who've ghettoed FFXIV (lowering buffer size to SNES equivalent etc) to make it run on old computers/laptops, experiencing in the cities? Cause they should REALLY have it rough if the cities are so extremely CPU reliant.

On my machine:

Athlon II X4 620 @ 3.12, 4gb DDR2 800, 512mb Radeon 4850 @ 685/1153, old ass 74gb raptor, Windows 7 x64, Catalyst 10.8

Both screens Windowed 1600x900

MSAA: 8xQ
Buffer Size: Resolution (Double won't run, don't have enough VRAM)
Shadow Detail: Highest
AO, DoF: On
Texture Quality: High
Texture Filtering: Highest
All In-game Configuration Options: On

img443.imageshack.us/img443/4421/highth.jpg


MSAA: No AA
Buffer Size: Quarter
Shadow Detail: Lowest
AO, DoF: Off
Texture Quality: Low
Texture Filtering: Lowest
All In-game Configuration Options: Off

img816.imageshack.us/img816/4112/lowa.jpg

[Katlan]

Set up my 2 480's. but waiting until retail since I dont care to DL the client for a few days. How is the performance w/ the Crysis hack?

[Ashtray]

I dunno with the Crysis SLI Profile, but I'm using the Nvidia Profile originally for FF14 that wasn't labeled correctly by just changing the appname in the profile.


Performance is 60 FPS nearly everywhere other than high character areas near Camp Crystals or certain parts of cities, but that seems to be a FFXIV coding issue that everyone is getting.

Edit: On 2x GTX 470
i7 930 @ 4.0Ghz
All Settings Highest, 16xQ AA, No AO or DOF. 1920 * 1080

[Siya]

OC that i7 yet?

Nay. Partly cause I don't know how, partly cause I'm running stock cooler.

[Pikarya]

Nay. Partly cause I don't know how, partly cause I'm running stock cooler.

Almost anything in the $30-$47 dollar bracket of coolers with some good TIM will take that baby home. Overclocking part isn't hard with a little research.

[Franzer]

I have a question regarding SLI. I've never used SLI before, and i just recently bought video cards to do it. Fullscreen is required for SLI correct? Also, does this apply to having a dual monitor setup? For example, If i have the game fullscreened on 1 monitor, so that i can utilize the other monitor for other things, am i only using 1 card for the screen that the game is running in? Thanks in advance.

[Shecky]

Does anyone have crossfire working??? I have a 5970 but performance is worse with both cards running. For some reason when both cards are enabled gpu usage never goes above 55% on either card. If anyone has a solution for this please help!!!!!

ARRRRG nevermind. I just deleted the Beta Client. I will try again when the actual game comes out retail. This is a piss off. An $800 card performing as bad as this. I can log into La Noscea and both cards will run at around 65% -70% each getting 60 FPS no problem...then after about 30 seconds of gaming...BOOM, right back down to 45-50% on each card dropping my FPS to around 35. This is stupid. Hope either ATI or SE gets this working.

[Cooleko]

I have a question regarding SLI. I've never used SLI before, and i just recently bought video cards to do it. Fullscreen is required for SLI correct? Also, does this apply to having a dual monitor setup? For example, If i have the game fullscreened on 1 monitor, so that i can utilize the other monitor for other things, am i only using 1 card for the screen that the game is running in? Thanks in advance.

SLI does not require full screen. there is nothing wrong with sli and dual/triple monitors. you can do anything you want to as far as configuring your windows and screens.

[Franzer]

SLI does not require full screen. there is nothing wrong with sli and dual/triple monitors. you can do anything you want to as far as configuring your windows and screens.

Great, thanks. I just wanted to make sure that i wasn't wasting performance on having two cards. I had read somewhere that if you weren't in fullscreen mode, you weren't actually in SLI mode, and that you were only using one card. I guess when i get home i'll play around with my hooking up my other monitor.

[Ashtray]

New Official Beta Nvidia Drivers just released, but anyway I was checking out if they added anything new to the SLI profile to FF14. Seems to be a new line of code.....


Profile "Final Fantasy XIV"
ShowOn GeForce
ProfileType Application
Executable "ffxivwin.exe" <--- still incorrect exe file name for beta should be "ffxivgame.exe"
Executable "ffxivwinbenchmark.exe"
Setting ID_0x0095def9 = 0x00010000
Setting ID_0x1033cec1 = 0x00000003
Setting ID_0x1033dcd2 = 0x00000004
Setting ID_0x1095def8 = 0x02c00005
Setting ID_0x10ecdb82 = 0x00000001 <--- new line of code
EndProfile


SHOULD BE:


Profile "Final Fantasy XIV"
ShowOn GeForce
ProfileType Application
Executable "ffxivgame.exe"
Executable "ffxivwinbenchmark.exe"
Setting ID_0x0095def9 = 0x00010000
Setting ID_0x1033cec1 = 0x00000003
Setting ID_0x1033dcd2 = 0x00000004
Setting ID_0x1095def8 = 0x02c00005
Setting ID_0x10ecdb82 = 0x00000001
EndProfile



Get testing!

[Katlan]

have you tested the Crysis settings? How does the game run w/that?

[Ashtray]

Never tested Crysis settings, used Nvidia's supplied settings throughout open beta.