My benchmark settings

I've been toying around with the benchmark to try to get the best picture quality for my graphics card. Since this sort of performance tuning thing is one of the things I do for work anyway, annnnnnd... well, I've done enough playing with the character creator xD

I got some ideas from this guide on reddit: http://www.reddit.com/r/ffxiv/comments/1jkm9n/lunas_guide_to_optimizing_low_end_system_for_arr/

The guide is pretty good as a starting point and if you don't want to think too hard about it, you can just set these same settings and get a pretty good balance.

I came at the problem from a different angle. My goals were:

1. Keep the score above 8000, which translates to a 60+ FPS

2. Get quality as high as possible, especially when it comes to things that I notice in the game.

#2 for me meant different settings from the guide. I notice whenever low-detail models are used on distance objects in the game, so I wouldn't check this box. I also notice when shadows have lots of "jaggies", especially when they fall on a moving character in the game as they jitter around the area. I don't notice grass quality changes, but since there's minimal performance impact to turning it to "high" for me, I just have it set like that.

A few things I've found about the benchmark configuration itself:

1. The "transparent lighting quality" setting has a significant impact on FPS. I couldn't notice any difference between high and normal, though maybe I don't know what to pay attention to
2. "Screen Space Ambient Occlusion" also has a significant impact. Due to the effect it has on lighting, I can't recommend turning it completely off, but "weak" is fine for most purposes.
3. The effect of the different shadow settings is small. Getting about a 100 point drop for increasing "Shadow Cascading" and "Shadow Softening", but I think the improvements in shadows is worth it. The two things I was paying attention to were jaggies in the shadows of the Aetherlyte as well as where shadows fell on character model armor. The resolution of the shadows makes the bigger difference, so I'm keeping it at 2048p. Since there was actually a point drop here, if I run into issues in the actual game when it releases, these would be the first things I tune down again.

Here are my final settings for the benchmark program

There are also a few other tricks outside of the game I used

1. I experimented with some overclocking of my GPU. Some really "safe" overclocking settings for my AMD HD7970 was to boost the clock speed from the stock 925Mhz up to 1Ghz and the memory clock from 1375Mhz to 1600Mhz. However, my card, the Sapphire HD 7970 OC edition is designed for overclocking with extra cooling, so your mileage may vary depending on your card. Interestingly, the memory clock speed increase gave the most increase in performance here, a good 300+ points. I guess they're pushing lots of textures.
2. I did not try overclocking my CPU. I generally don't like CPU overclocking as a 10% increase in performance results in a lot of extra heat being generated. I'd rather not roast in my room in summer xD. I did notice that many people on reddit had generally the same scores with the same GPUs, leading me to believe that the CPU is usually not a bottleneck.
3. Luna's guide says to use the Razer Game Booster to turn of unnecessary things before starting the game. I tried it, and it didn't do anything for me, probably because my FFXIV computer is already tuned for playin games to begin with. If you use your computer for other things, you might consider doing this
4. Use of a SSD vs. HDD gives virtually no performance difference. This is what I expected with the graphics being cached in ram. The main advantage that a SSD might give is shorter loading screens. I see a major difference in waiting for the load screens when running the benchmark on a computer with a SATA-III SSD and a laptop.

So final "tuned" benchmark score with the graphics as nice as I can make it is:

Overall, increasing the details in many different areas dropped the score by about 150 points compared to my original settings. It's a far step down from the 15000+ from running at 1920x1080, but I like the higher 2560x1600 resolution. It's a definite step up from the 6690 from running maximum settings at this resolution too.

I'm not sure if this would be useful to anyone else, but... drop a comment if it helps you in figuring out your own settings!

To finish things off, here are a couple of pics from the benchmark program itself with hopefully final hair color Ascy!