Exploring “Frame time” measurement – Part 3 – the GTX 680 versus the HD 7970 GHz Edition

In our last two evaluations, we established that Fraps is a very good measure of frame rates (fps) as well as its ability to measure frame times. Fraps outputs results that can be charted in Excel or a similar spreadsheet program if “frametimes” is checked (along with “FPS” and “MinMaxAvg” which are usually measured, under the “FPS” tab).  And we also established that using a SSD versus a HDD for benching does not generally impact frame time smoothness although we will take care to use identical drives when benchmarking.

We looked at eleven in-game benchmarks – especially the STALKER, Call of Pripyat stand-alone benchmark and Metro 2033’s official benchmark, as well as Far Cry 2, because their game engines also measures frame times. By comparing the game engines reported timing with Fraps frame time measurements, we found that the charts – although not identical – were so close as to nearly mirror each others results.

We feel reasonably confident as we proceed that Fraps is as useful a tool in reporting frame times as it is for frame rates. We also found that using a hard disk drive (HDD) versus a solid state drive (SSD) has little impact using Fraps to measure the smoothness of a gaming experience although we shall always bench from identical drives.We are still using the stock-clocked GTX 680 in in this evaluation and we shall pit its frame time smoothness of delivery versus the HD 7970 at GHz edition clocks in eleven benchmarks. Raw frame rates simply do not convey the complete gaming experience and this has been especially illustrated with multi-GPU, where otherwise satisfactory frame rates may feel much slower than then they should due to uneven delivery of frames.  And now we see this unevenness or “jitter” on single GPUs in some games.

One thing to note is that we have overclocked our Core i7-3770K CPU to 4.5GHz to make sure that there is no CPU bottleneck. We are testing at 1920×1080 with a 120Hz display. We are testing eleven benchmarks in this Part 2, generally using 2x/4xAA or 8xAA plus 16xAF and with maximum (ultra) DX11/10.1/10/9c details whenever it is available. We are benching with GeForce 310.70 WHQL drivers and the very latest Catalyst 12-11 Beta 11 and the latest CAPs.

An Update on our Last Testing, SSD vs HDD smoothness differences – using a faster SSD

We noted few differences between the HDD and the SSD in terms of smoothness, possibly because we are using repeatable built-in benchmarks and timedemos.  Of course, there are situations where the game must stream from the disk and there can be choppiness, hitches or jitter if the device is too slow to keep up with the requests.  In that case, a faster drive could be very helpful.  Of course, game developers have naturally targeted and optimized their streaming from the HDD, not from the SSD.

Games typically load content in the background as the player progresses through a scene.  The game has to request data from the storage media and the longer it takes, the game may contain frames that are not rendered evenly as the rendering engine tries to keep up.  In a worst case, the game may stall.  According to Intel, uneven frame delivery occurs whenever 0.1 percent or less of total display pixels change for a duration of at least five frames.

In all cases, by their nature of being repeatable benchmarks, our runs were very similar with no apparent jitter differences – except for two benches.  Resident Evil 5 showed the benchmark running on the Kingston SSD to be smoother in the begining than on the OWC SSD; and Far Cry 2 showed the benchmark running on the OWC SSD to be significantly smoother than on the Kingston SSD.  Since Resident Evil 5‘s shorter (“variable”) benchmark that we used in our evaluation’s Parts 1 and 2 is more CPU dependent than GPU-dependent, we are using the much longer “fixed” benchmark in this Part 3.

Here is the Far Cry 2 (Ranch Short) bench as run with a GTX 680 on the Kingston SSD:

Here is the same bench run on the OWC SSD with the same GTX 680 and everything else, the same:

We see that the OWC SSD exhibits smoother framerate delivery and less incidents of jitter.  In this particular case, we can assume that a slower drive can contribute to uneven delivery of framerate.    At any rate, for this evaluation, we imaged our Kingston HyperX drives to our fastest SSD, the 240GB OWC Mercury EXTREME 6G, and reran all of our benches. We also imaged the HD 7970 partition to the same OWC SSD after we were done with the GeForce so our results in this comparison would be apples-to-apples and the differences would be caused by the graphics sub-system.

If you are a gamer and you want absolutely the best chance for the ultimate gaming experience and you already have a high-end PC, you might also consider using a very fast SSD and seeing for yourself if you can notice the possible differences between the HDD and a SSD in gaming.

Let’s check out our hardware and software used for this Part 3 of our ongoing evaluation testing gaming smoothness.  Please note that we are using the OWC 240GB Mercury EXTREME 6G for all of our benching.