The OFFICIAL FFXIV Benchmark & Beta Performance Thread

**Jojimbo** · 2010-06-21 01:31

Has anyone got the 3DMark06 rename to work (also saw that the person had a 3dx file in their benchmark folder). I renamed the file in \FFXIVBenchmark\data and ran it, but it's just using the same race/res settings as I had selected when running it normally.

Although I did find out that something has broken my crossfire >.> Doesn't work on anything now (with 10.6 or 10.5 I'm getting exact same issues)

EDIT:: Installed 10.4 and XFire is working on Furmark Multi-GPU and UT3 Benchmark, but not FFXIV or STO. 2nd card isn't underclocking when idle either, well it does for a few seconds randomly at times ;;

**Aurelius** · 2010-06-21 01:42

Originally Posted by Jojimbo

Has anyone got the 3DMark06 rename to work (also saw that the person had a 3dx file in their benchmark folder). I renamed the file in \FFXIVBenchmark\data and ran it, but it's just using the same race/res settings as I had selected when running it normally.

Although I did find out that something has broken my crossfire >.> Doesn't work on anything now (with 10.6 or 10.5 I'm getting exact same issues)

The d3d9.dll file is a DirectX file for DX9. I'm not sure exactly why they've got it in that folder as it's normally stored elsewhere on the PC. Unless the guy is some sort of super hacker or the JP (I'm assuming the text in his Windows Explorer pane was Japanese) benchmark app is more refined than the one we got, the only way he could have posted scores like that would have been for both of his 5970s to have been working as intended.

Also, as far as resolution/race settings, if you're getting low resolution settings try exiting the benchmark, running it from the normal executable and selecting high res, then exiting that and trying the renamed executable again. The first time I ran it from the renamed executable, it defaulted to high resolution with random race, and I had never selected random race in any of the previous tests I had run. I went back later and ran it again and it was defaulting to whatever settings I had used the last time I ran it from the normal exe.

Looking a little closer, it's got the i7 "Intel Inside" logo instead of the nVidia one, so I'm not quite sure what to make of it. I apologize if the information I offered doesn't work, but at the very least the video shows what could be possible with other multi-GPU builds when ATI/nVidia drivers support multi-GPU use in XIV. Obviously not everyone is going to be running an overclocked 980X with a pair of overclocked 5970s, but it wouldn't surprise me to see newer quad cores with dual 58xx/4xx cards pushing 6-7k quite handily.

Again, if the guy in the video was more sneaky than it appears and the solution I offered doesn't work, I apologize.

**Deadgye** · 2010-06-21 02:45

Interesting.. apparently my low score puts me at second on the scoreboards.. Guess I should redo it and take a picture.

**Aurelius** · 2010-06-21 02:58

Originally Posted by Deadgye

Interesting.. apparently my low score puts me at second on the scoreboards.. Guess I should redo it and take a picture.

Probably put you at at least third depending on when it's updated next. Still, that's not bad at all.

Soap inspired me and I've been tweaking a bit:

Spoiler: show

**Kachi** · 2010-06-21 04:54

So I'm wondering if I should look into OCing my CPU/GPU. I've never done it before, so I defer to the better informed in regards to how far I should push it. How much could I improve my performance?

w7 64 bit
AMD Phenom II X4 805, 2.5GHz
ATI Radeon HD 5750
8GB DDR2 RAM

I keep my place at an icy 65F, though I'm not sure how much a few Fahrenheit actually helps with cooling compentents in plain air. Definitely not comfortable with water cooling. Thanks for any input.

**Kryssan** · 2010-06-21 08:41

Ok, going to combine all my stuff into one post for easier localization.

Spoilers for size of pictures.

Spoiler: show

Original run data. I'm using Crossfire-enabled 5870 Sapphire Vapor-Xs (specifically this model). Also, the benchmark only seems to report up to 4Ghz. As you can see from this picture:

http://www.aethernavale.net/Media/images/cpuz.jpg

I am actually running higher than that.

All settings the same with only one card installed instead of crossfire - low then high.

Spoiler: show

So I reinstalled my second card, and ran all the tests again, this time one at a time for most of the things I specified.

Original rerun, minus AI:

Spoiler: show

Original settings [so with Catalyst AI (numbers are too close to call really, but this is the one case where AI won out over being disabled - AI is supposed to improve performance, but from what I can tell it lowers it and doesn't provide any noticeable quality improvement)]:

Spoiler: show

Original settings, with AI, with AAA enabled:

Spoiler: show

Original settings, minus AI, with AAA enabled:

Spoiler: show

Original settings, minus AI, edge-detect:

Spoiler: show

Original settings, with AI, edge-detect:

Spoiler: show

Tweaked settings, second run:

Spoiler: show

Settings used in above tests:
AA enabled (not app controlled), with Temporal AA, 8x Edge Detect (24 samples)
AAA enabled, Quality, Supersampling
AAmode Quality, Supersample AA
AF enabled (not app controlled), High Quality AF enabled, 16x per pixel samples
Catalyst AI enabled, Advanced mode
Mipmap Hiqh Quality
Geometry Instancing Enabled
Triple Buffering Enabled

Spoiler: show

The exact same settings were used in these tests, with one exception - Catalyst AI is completely disabled.

During the tests with AI enabled, though the scores are similar, the performance was not. Performance was especially hampered during the battle scenes where multiple critters came up on deck, and even more so during the Mithra's shield event with the graph dipping down to nil at several points that lasted for several seconds, with visible stutter on screen.

With AI disabled, playback is smooth to my eyes, with a few instances of graph dipping but instantaneously returning to the normal values.

Edit: Oh yeah, all tests displayed during this post were run on a single GPU card only.

Below are the results of running it with the same settings as above only with Aero disabled:

Spoiler: show

No discernible difference. Aero's so minimal that it shouldn't affect anything on a high end system. Still, some people are showing differences of a good margin so it might be worth the effort for you.

Registry tweak time~

Spoiler: show

The above pair was performed with Crossfire Enabled.

Spoiler: show

The above pair was performed with Crossfire Disabled.

The tweaks imposed with ATT:

Spoiler: show

Same tests, low grade, with crossfire and AI enabled and some more AA/AF tweaks but no reduction in quality:

Spoiler: show

GPU Overclocks. This isn't going to show as much improvement as others have been able to achieve with GPU overclocks due to the nature of the VRAMs on the Sapphire VaporX cards. Since the only way to change that is to do BIOS editing of your card, I will not include results from that seeing as the average user will not be going to such lengths to get performance. Therefore, these OCs are from using CCC and ATT only.

Spoiler: show

Above with Crossfire.

Below without Crossfire.

Spoiler: show

Different driver sets, options setup to be as stated below - Vanilla 10.6:

Spoiler: show

Installed the new Enzyme 1.4 betas:

Spoiler: show

Score increased around 300 from original score, much better than 10.6 vanilla. 1.4 betas are based off of the 10.5 hotfix drivers, where as 1.3 (the original driver set I used) are based off of the 10.3 drivers. Additionally, where before I had a decent spread between scores, the scores are now less than 10 apart. Not as great as the exact same on each run, but much tighter than before. Options were crossfire enabled, registry tweaks removed, CCC: AA x8 Edge Detect, HDAF 16X, AI Disabled, Mipmap High Quality, Vertical Refresh off, AA Mode Super-sample AA, triple buffering enabled. It would also seem the issues with AI disabled affecting AA being revised with the later driver sets.

Added note: Resized all images to help with loading. Right now using vanilla 10.6 due to an intermittent issue with the betas and my TV.

**Cautery** · 2010-06-21 10:53

So is it possible to let dual gpu card work with both gpu in this benchmark (like HD5970)?

**Gulkeeva** · 2010-06-21 10:55

Late reply but as I mentioned earlier, I didn't upgrade to cata 10.6 is because it has issues with 2D gaming, which I play alot of since hentai games are mainly 2D *ahem* >.>

I've heard of other issues such as...

Full screen gaming issues 2D/3D (crashes and or just jumps out of fullscreen)
(tearing with or without vsync, black box)
Some flash videos with hardware acceleration transparently crash/BSOD

although personally I've noticed that this months MS updates.. one of them was also causing "fullscreen" gaming errors, removing them fixed alot of issues also.

**Kryssan** · 2010-06-21 10:59

Originally Posted by Cautery

So is it possible to let dual gpu card work with both gpu in this benchmark (like HD5970)?

Not yet. People have suggested some ideas to force the program to take it but so far there is no data showing that someone has gotten multigpu/CF/SLi to work with it.

**alecrast** · 2010-06-21 11:22

Originally Posted by Aurelius

The CCC Overdrive utility is not horrible for overclocking with the stock fan, but something like MSI Afterburner gives you a bit more control and lets you push a little harder. 90C isn't horrible for an overclocked GPU but you're always best to shut it down at the point you're no longer comfortable with the temperature. E-peen isn't worth the anxiety.

Having said that, I decided to do a little messing around and renamed my executables not because I have a multi-GPU rig but just to see if any driver optimizations for other applications might give me a boost for the XIV bench, which they did not D:

And so now, I offer my final score for artificial e-peen extension and await my humiliation at the hands of you multi-GPU tyrants ;D

5746 hires; CPU @ 4.35ghz, single Radeon 5870 1.30v, 1070 core, 1290 vram

Spoiler: show

5512 is about the max that I'll get on my system with 10.6 Cats. I can't get past 4.2 stable on my i7 860 without starting to apply voltage tweaks and while my Titan Fenrir is nice, it just can't match the thermal convection of liquid cooling. 4.0GHz (20x200) with Turbo off is pretty much the sweet spot on my 860 since it fits nicely with DDR3-1600. My Sapphire 5870 2GB (at stock voltages) won't run stable past 995MHz core and 1285MHz mem. The extra memory hurts a little in that regard but makes up for it in other areas.

Now it's just a matter of seeing if we'll have a GTX485 refresh with all 512 shaders enabled or an HD6000 series card from ATI before we have the actual game in our hands.

I'm still of the opinion that SE is overestimating their figures. I'm expecting to be able to play at 2560x1600 with all the bells and whistles on while maintaining an acceptable framerate with a single card solution, be it NVIDIA or ATI without having to resort to extreme overclocking and "exotic" cooling solutions. I try to avoid water-cooling because my gaming rigs tend to get re-purposed as dedicated "on 24x7" Linux systems after about a year and, when that happens, I don't want the maintenance associated with it.

Out of curiosity, what liquid are you using in your system? Distilled water? MCT-5/40? Feser?

**HolyDragoon** · 2010-06-21 12:33

I was asked to show everyone this awesome technological feat.

Spoiler: show

Guess I need to go shopping for this set up.

**Chicomagiko** · 2010-06-21 13:21

^LOL

Anywho, here's mine. Also, I'm in the alpha but have yet to bother trying it. Whomp whomp.

Spoiler: show

**Eticket** · 2010-06-21 13:31

Originally Posted by Chicomagiko

Anywho, here's mine. Also, I'm in the alpha but have yet to bother trying it. Whomp whomp.

Good work. It's done now anyway.

**eskimobobz** · 2010-06-21 13:54

Sorry, I can't post direct links since I just registered.

High settings:
Score - 2212
Load time - 11756
img683.imageshack.us/img683/558/ffxivhighbench.jpg

Low settings:
Score - 3946
Load time - 11529
img822.imageshack.us/img822/6161/ffxivlowbench.jpg

System:
WinXP SP3
AMD Phenom II X4 955 @ 3.2 stock speed
Nvidia 9800 GTX+ 512Mb
G.Skill 2x2=4gig DDR2 800mhz

Just wanted to throw in my benchmark as the current charts are mostly Intel/high end systems, and I wanted to show my AMD/Mid tier system bench.
Looks like I'll be upgrading my Vid card when the game comes out. =)

Thanks for the awesome thread!

PS. I also want to note that on both low and high settings, my GPU was being used @100%. While my CPU was @20-25% on High, and 40-55% on low.

**Minions** · 2010-06-21 20:09

Finally did High:

Ended up getting 2795 on GTX 285 with I7 920 @2.7ghz. Guess that's good enough.

**Fade** · 2010-06-21 20:24

Has anyone posted scores from gaming laptops? Like the Asus G73, Alienware laptops, ect...

**Aurelius** · 2010-06-21 20:41

Originally Posted by alecrast

5512 is about the max that I'll get on my system with 10.6 Cats. I can't get past 4.2 stable on my i7 860 without starting to apply voltage tweaks and while my Titan Fenrir is nice, it just can't match the thermal convection of liquid cooling. 4.0GHz (20x200) with Turbo off is pretty much the sweet spot on my 860 since it fits nicely with DDR3-1600. My Sapphire 5870 2GB (at stock voltages) won't run stable past 995MHz core and 1285MHz mem. The extra memory hurts a little in that regard but makes up for it in other areas.

Now it's just a matter of seeing if we'll have a GTX485 refresh with all 512 shaders enabled or an HD6000 series card from ATI before we have the actual game in our hands.

I'm still of the opinion that SE is overestimating their figures. I'm expecting to be able to play at 2560x1600 with all the bells and whistles on while maintaining an acceptable framerate with a single card solution, be it NVIDIA or ATI without having to resort to extreme overclocking and "exotic" cooling solutions. I try to avoid water-cooling because my gaming rigs tend to get re-purposed as dedicated "on 24x7" Linux systems after about a year and, when that happens, I don't want the maintenance associated with it.

Out of curiosity, what liquid are you using in your system? Distilled water? MCT-5/40? Feser?

Right now I'm using Primochill PC Ice but next time I do a teardown I'm going to replace it with plain ol' distilled water and see if it makes all that much of a difference. Most of my fittings are silver plated and my tubing had an anti-microbial agent mixed in with the plastic before it was extruded, so I'm not too worried about that. Main thing is that your coolant temperature should never creep up over time. If it does, it's usually time to either throttle back the OCs or add another radiator and the more exotic coolant blends are highly unlikely to prevent from having to do either of those two things. I appear to be stable now at 4.15Ghz + turbo but I won't be able to really test it until I get my chipset block installed. And even if I am rock solid stable at 4.15, I think that's about as far as I'm going to be able to push it. 4.15 + turbo for an i7 920 is not too shabby, though.

**Kryssan** · 2010-06-21 21:27

Originally Posted by Aurelius

Right now I'm using Primochill PC Ice but next time I do a teardown I'm going to replace it with plain ol' distilled water and see if it makes all that much of a difference. Most of my fittings are silver plated and my tubing had an anti-microbial agent mixed in with the plastic before it was extruded, so I'm not too worried about that. Main thing is that your coolant temperature should never creep up over time. If it does, it's usually time to either throttle back the OCs or add another radiator and the more exotic coolant blends are highly unlikely to prevent from having to do either of those two things. I appear to be stable now at 4.15Ghz + turbo but I won't be able to really test it until I get my chipset block installed. And even if I am rock solid stable at 4.15, I think that's about as far as I'm going to be able to push it. 4.15 + turbo for an i7 920 is not too shabby, though.

You really ought to be able to get a higher clock than that if you're coolant/heat exchanger limited. I'm doing 4.2 on air, rock solid stable. When I watercooled in the past the only thing I used for coolant was Koolance's blend or grade A water (we make it all the time at work) - never went exotic. And definitely never silver plated. To be honest sounds like a waste unless you're using peltiers or phase change.

**Kajii** · 2010-06-21 22:25

Originally Posted by Fade

Has anyone posted scores from gaming laptops? Like the Asus G73, Alienware laptops, ect...

Asus G73 here

Just ran test. I only did Low since High didn't even fit on my screen.

http://img197.imageshack.us/img197/1...ivlowscore.png

**Aurelius** · 2010-06-21 22:29

Originally Posted by Kryssan

You really ought to be able to get a higher clock than that if you're coolant/heat exchanger limited. I'm doing 4.2 on air, rock solid stable. When I watercooled in the past the only thing I used for coolant was Koolance's blend or grade A water (we make it all the time at work) - never went exotic. And definitely never silver plated. To be honest sounds like a waste unless you're using peltiers or phase change.

It's running fine for day to day use at 4.315, but my VREG hit 98C after a few minutes the last time I ran Intel Burn Test and that was at 4.0Ghz. About my only option to go higher now would be to lower my multiplier and up my clock speed but I'm satisfied with where it's at now. I'm just holding of on stability testing until I install the chipset block because I know my VREG is going to scream ;D For all the overvolting it took to get to this point, my idle temps are 48C across all cores and 76C on the VREG which is not bad at all. It just doesn't leave the VREG much wiggle room for torture testing.

Thread: The OFFICIAL FFXIV Benchmark & Beta Performance Thread

Thread Tools

Benchmark results.

Latest Threads

Up & Coming Threads

Hottest Threads