Test Setup and Errata
I am currently unable to do the frame rating/frame pacing testing that Ryan does. It simply costs a lot of money to get that setup. Instead, I have used the more traditional testing of FPS across multiple applications and benchmarks. In this context, it certainly does not invalidate these results. I do not test in SLI or CrossFire, which is where we see issues primarily crop up with frame rating. Ryan covers every new GPU with the frame rating suite, so if a reader is curious about those results, they can find them here.
I recently changed from the AMD test system to the Intel based i7 3770K. This CPU certainly pushes even high end graphics cards and gives us a better look at their theoretical performance. I also started testing in only 1920×1200 and 2560×1600. 1080P is the most common resolution that games are typically run in, due to the extreme affordability of 1080P panels.
I test the cards in a rather large case, instead of an open air workstation. I believe this is more representative of most usage scenarios, and the controlled airflow does provide positive results when it comes to thermals.
I also tested the three top end cards in Eyefinity/Surround. Resolution is set to 5760×1200 and the applications used are Skyrim, GRID 2, Battlefield 3, and Bioshock Infinity.
Intel i7 3770K
Gigabyte G1 Sniper 3
GSkill 2 x 4GB DDR-3 1866
WD 1 TB Caviar Black HD
Enermax Fulmo GT Case
Corsair AX1200 Power Supply
Windows 7 64 Bit
NVIDIA 320.18 Drivers
Catalyst 13.6 Beta 2
I try to have a good balance of synthetic testing combined with real world results. I believe this is a good balance of what is theoretically possible with these chips vs. what we see in actual games. Let us start off with the synthetics first.