Understanding FreeSync A Bit More
Understanding FreeSync A Bit More
At first glance, it may look like AMDís FreeSync is simply a display-bound technology which should have a drastic impact upon in-game motion fluidity but minimal effect upon actual framerates. This couldnít be further from the truth since it has the ability to both improve and reduce in-game performance depending on the way it is utilized. Remember, any system with FreeSync has three possible display modes: FreeSync ON & V-Sync ON for the best possible image quality, FreeSync ON & V-Sync OFF, FreeSync OFF & V-Sync OFF and only V-Sync ON
Iíve included a chart below which clearly highlights how these settings translate into framerates. Other than a visual representation of stuttering and tearing, I think this is the clearest visual reference of FreeSyncís benefits and shortcomings.
Disabling V-Sync is obviously the best way to go from a raw performance standpoint since the framerate is able to operate independently from the monitorís refresh rate limitations. However, it introduces noticeable tearing of the image.
Turning on FreeSync while keeping V-Sync off doesnít impact framerates at all but, as Iíll discuss in the gameplay impressions section, the visual benefits are minimal at best. From my understanding, this is the setting AMD has been using to compare their solutionís performance to G-SYNC but from a visual standpoint thereís very little to recommend it.
Looking at performance from a V-Sync enabled perspective highlights why this setting can be detrimental to onscreen motion consistency. Even on a 144Hz panel we can see how the framerates go through a series of step-downs as the frames are delivered in an order that matches the panelís refresh rate. When framerates are above 144 things match up perfectly but synchronizing below that point causes problems since the graphics card is forced to buffer ready frames as it waits for a refresh cycle to hit. This leads to those frames being delivered at ratio of 1:1, 1:2, 1:3, 1:4 and so on of the maximum refresh rate which roughly translates to 144FPS, 72FPS, 48FPS, 36FPS, etc. on this particular BenQ monitor depending on what the actual deliverable framerate is.
For example, if the GPU can deliver 100 frames per second it will sync at 72 since 144 isnít achievable or 65 would become 48. In these situations stuttering becomes increasingly apparent as the frame times jump around like a jack in the box. Meanwhile, a good portion of the graphics cardís performance is left on the table to rot.
Enabling FreeSync and V-Sync leads to a completely different situation than those described above. Its performance range under 144FPS is very similar to what a V-Sync disabled environment provides. Here, FreeSync allows framerates to vary without any impact upon visual quality (tearing and stuttering are completely eliminated) but above that point, framerates are still capped at the panelís maximum refresh rate. Granted, once again some of the GPUís raw horsepower is wasted in those few instances where it can effectively render more than 144 frames per second but users are left without the jarring step-downs that happen when just V-Sync is enabled.
Perhaps one of AMDís most controversial claims about FreeSync is its ability to actually increase performance while G-SYNC tends to reduce framerates. Supposedly this is due to the fact that Adaptive Syncís handshake protocols give the graphics card real-time knowledge of the monitorís refresh states while G-SYNC has to constantly poll the monitor, taking up some core operational cycles. We arenít talking about noticeable differences since even AMDís most optimistic charts all highlight variances that are so minor they can be chalked up to any number of variables. However, the numbers are there and I needed to test them for myself.
The Unigine chart posted above should give some hints about what FreeSync can and canít do from a performance standpoint. It looks like when FreeSync is enabled and V-Sync is turned off, thereís very little to no negative impact. With that being said, when FreeSync operates identically to G-SYNC Ėthatís to say with V-Sync enabled to eliminate tearing- thereís actually a very minor reduction in delivered frames when under 144FPS while the upper range is still capped at the vertical synchronization rate.
In order to look into this situation a bit closer, I took ten games and repeated run-throughs for each ten times. In each situation, G-SYNC and FreeSync + V-SYNC were turned on and off for a true apples to apples comparison. These 100 data points were translated to the chart below.
Contrary to AMDís claims, all of the data points towards FreeSync negatively impacting framerates more than G-SYNC does. This could be due to additional driver overhead or GPU cycles being used to insure FreeSyncís constant functionality or it could point towards FreeSyncís relative immaturity. Whatever it is, I feel like Iím splitting hairs here since thereís just no way a gamer notice any visible performance differences in either scenario.
In these scenarios I was insuring that the framerates remained above 40 and below 144 since FreeSync behaves very differently when performance drops below the monitorís minimum refresh rate. Outside of this ďsweet zoneĒ it exhibits a worrying tendency to step down to even lower levels than V-Sync does, completely destroying motion fluidity in many games. This will be discussed more on the next page.
|Latest Reviews in Video Card News|