I picked up a Palit GTX460 Sonic Platinum 1GB last week to test out Square-Enix's recommended video card in a variety of configurations since it seems to be the soup du jour that a lot of people are looking into for FFXIV. This particular GTX460 weighs in with an 800MHz core, 1600MHz shader clock, and 4000MHz memory.
The system configurations the GTX460 was tested in are as follows:
First system:Intel Core i7 860 (2.8GHz stock, 4.2GHz OC)
Gigabyte GA-P55A-UD7
8GB DDR3-1600
Intel X25-M 80GB
Windows 7 Professional 64-bit
Intel Core 2 Extreme X6800 (2.93GHz stock, 3.7GHz OC)
eVGA 680i SLI
8GB DDR2-1066
Intel X25-M 80GB
Windows 7 Professional 64-bit
The X6800, for those that don't remember, is essentially a Core 2 Duo E7500 with an additional 1MB L2 cache (4MB total) on 65nm instead of 45nm and has an unlocked multiplier.
And lastly, our third system:AMD Phenom II x3 720BE (2.8GHz stock, 3.8GHz OC)
MSI 790FX-GD70
8GB DDR3-1600
Intel X25-M 80GB
Windows 7 Professional 64-bit
In each case, the X25-M was secure erased back to its factory new condition, Windows was then cleanly installed and patched, and all the latest drivers for each system were installed. The video card driver used for this test was NVIDIA's 258.96 with all settings left on stock.
The first table shows the results from the Official FFXIV Benchmark tool results for the various configurations with the GTX460 installed (scores at stock speeds bolded):
| 2.8GHz | 2.93GHz | 3.7GHz | 3.8GHz | 4.2GHz | |
| Intel Core i7 860 (quad-core) | 2920 | - | - | - | 2969 |
| Intel Core 2 Extreme X6800 (dual-core) | - | 2491 | 2635 | - | - |
| AMD Phenom II x3 720BE (tri-core) | 2717 | - | - | 2909 | - |
The second table shows the results of my hand-run benchmark, running the same course for each test which started in Limsa Lominsa's Adventurer's Guild, down a short ways past Bearded Rock, and then back along the same path. The course takes a little over 8 minutes per run.
The HIGH settings used were:Display Mode: Fullscreen
Window Size: 1920x1080
Multisampling: 16xQ CSAA
Buffer Size: Resolution
Shadow Detail: Highest
Texture Quality: High
Texture Filtering: Highest
AO: Off
DoF: Off
| i7 860 2.8GHz | i7 860 4.2GHz | X6800 2.93GHz | X6800 3.7GHz | 720BE 2.8GHz | 720BE 3.8GHz | |
| Average FPS | 41.7 | 49.8 | 27.9 | 30.5 | 32.6 | 41.1 |
| Minimum FPS | 21.7 | 28.2 | 11.7 | 12.3 | 13.4 | 17.5 |
| Average GPU | 77.2% | 89.8% | 51.2% | 52.3% | 59.2% | 73.3% |
| Average CPU | 53.0% | 55.1% | 90.3% | 90.8% | 70.2% | 67.2% |
Unfortunately, average CPU usage doesn't paint a real picture as it's possible for the average to hide a pegged CPU. For that, we need the graphs which will come at a later time (yes, I fail at Excel).
I can't say I was very impressed with the results for Core 2 Duo. Looking at the average framerate, it appears playable but I found the dive in framerate in various spots in Limsa and at Bearded Rock maddening. I can only imagine what a Dynamis situation might be like. If I was on C2D, I would seriously consider upgrading to i7 right about now.
I know some of you will be thinking you're going to get away with playing at lower settings to lessen the stress on the system so I tried that, too.
The HIGH settings can be found above.
The LOW settings used were:Display Mode: Fullscreen
Window Size: 1920x1080
Multisampling: No AA
Buffer Size: Resolution
Shadow Detail: Lowest
Texture Quality: Low
Texture Filtering: Lowest
AO: Off
DoF: Off
| i7 860 2.8GHz | X6800 2.93GHz | 720BE 2.8GHz | |
| Average FPS HIGH | 41.7 | 27.9 | 32.6 |
| Average FPS LOW | 40.2 | 25.6 | 33.4 |
| Minimum FPS HIGH | 21.7 | 11.7 | 13.4 |
| Minimum FPS LOW | 17.8 | 8.9 | 12.3 |
| Average GPU HIGH | 77.2% | 51.2% | 59.2% |
| Average GPU LOW | 59.1% | 37.4% | 44.3% |
Yes, you see correctly. On the two Intel systems, changing all settings to their lowest while still maintaining 1920x1080 resolution actually reduces performance. I ran both benchmarks twice before hitting my head against a wall. At least the AMD system seemed to behave more rationally in that regard.
Finally, I know some of you are going to ask about 1280x720. To be honest, I have little interest in 1280x720 resolution but I did run the benchmark a couple times on the Core 2 X6800 system to see if I could do something about the framerate. Since we see better (or at least equivalent) performance with the HIGH settings, they were used for both 1920x1080 and 1280x720. I tossed a 1280x720 test with "Buffer Size: Double" in there just for fun.
| 1920x1080 | 1280x720 | 1280x720 Buffer Size: Double | |
| Average FPS | 27.9 | 26.6 | 25.6 |
| Minimum FPS | 11.7 | 11.5 | 10.5 |
| Average GPU | 51.2% | 31.9% | 69.2% |
Overall, I am very impressed with the GTX460's performance in the i7. It does appear to be a nice little workhorse GPU but a dual-core system, be it a Core 2 Duo or an Athlon/Phenom dual-core, is going to have an incredibly tough time extracting performance from it. The i3 might faire a little better but I start to get sleepy looking at anything below i5.
There is still plenty of time between now and launch and I believe there is a lot more performance that can be squeezed out of the client if only SE's programmers are able. There are places where performance inexplicably takes a nosedive. If SE can address those, it will definitely improve everyone's experience.
Update: Added the chart below to highlight the performance of a $250 GTX460 versus a ~$500 Sapphire 5870 2GB Toxic.
| i7 860 2.8GHz + GTX460 | i7 860 2.8GHz + 5870 | i7 860 4.2GHz + GTX460 | i7 860 4.2GHz + 5870 | |
| Average FPS | 41.7 | 44.6 | 49.8 | 52.6 |
| Minimum FPS | 21.7 | 22.3 | 28.2 | 31.1 |
| Average GPU | 77.2% | 58.7% | 89.8% | 70.7% |
| Average CPU | 53.0% | 48.5% | 55.1% | 45.1% |
The overclocked GTX460 matches a stock 5870 in performance though the 5870 felt like it provided an overall smoother experience. Not bad for a card that's about half the price, however.

Reply
