We tested Battlefield 3 with all of its DX11 goodness cranked up, including the "Ultra" quality settings with both 4X MSAA and the high-quality version of the post-process FXAA. Our test was conducted in the "Kaffarov" level, for 60 seconds starting at the first checkpoint.
We've gathered a tremendous amount of data during our testing. To make it more manageable, we're trying these new switchable plots. You can click on the buttons to see the GTX 660 Ti compared to various types of cards.
These are plots of the time required to render every frame in a single, representative test run. Lower frame times are preferable, and in some cases, you can see the undesirable spikes representing long-latency frames quite clearly in the plots above—particularly on the older GeForce cards.
If you're confused by our use of frame times, let me direct you to my article, Inside the second: A new look at game benchmarking, for an introduction to our latency-focused approach to graphics testing. Traditional frame rate averages don't offer enough resolution to capture the occasional stops and stutters that happen when your PC is just a bit too slow to run a game smoothly. We've found that concentrating on frame latencies allows us to evaluate performance more accurately. For reference, the table on the right offers some translations from notable frame time thresholds to FPS rates.
We won't leave out the FPS average, of course, since it's very familiar.
However, switching to the 99th percentile frame time—probably our best single-number performance summary—illustrates just how poorly the older GeForce cards fare in this test scenario. The time required to render that last ~1% of frames on the GTX 560 Ti and GTX 470 is pretty unfortunate.
A look at the broader latency curve further illuminates the problem. Frame times on the older GeForces stay happily below 40 milliseconds most of the time, but roughly three percent of the frames involve a much longer wait, spiking to 60 milliseconds or more, where you'd really notice the slowdown.
We've seen this problem in certain levels of BF3 a number of times before on Fermi-class GeForces. Fortunately, the newer Kepler cards appear to have overcome it. In fact, Zotac's GTX 660 Ti AMP! hangs with the boost-enhanced Radeon HD 7950.
Our final metric may be my favorite, because it's about avoiding those long-latency frames that disrupt gameplay. The question is: how much time does each card spend working on really long-latency frames? To answer that, we add up all of the time the cards spend rendering beyond a per-frame threshold.
We start by setting that threshold at 50 ms, which corresponds to 20 FPS, because we think the illusion of motion begins to break down for most folks somewhere around that point. (Movies run at 24 FPS, for instance.) We can then ratchet the threshold down if we want to be even pickier about performance.
As you can see, the two older GeForces are the only cards that produce really worrisome slowdowns. In fact, even at the 33-ms threshold, all of the newer cards are golden. That means they all maintain a near-constant frame rate of 30 FPS. The 16.7-ms threshold is the toughest test—equivalent to 60 FPS—and there, the GTX 660 Ti cards outperform all of the Radeons we tested, though the PNY card's margin of victory over the 7950 is slim.
|Samsung's 28'' display serves up single-tile 4K at 60Hz for $800||115|
|Good Friday Shortbread||37|
|Friday night topic: where are the good ultraportables?||80|
|Deal of the week: Radeon R9 290X cards for... more than list?||19|
|Release roundup: Bits, pieces, and whole PCs||29|
|AMD posts another loss but beats Wall Street forecast||65|
|GlobalFoundries licenses Samsung process tech, grants AMD access to FinFETs||108|
|MSI shows next-gen Intel motherboards||47|