Our testing conundrum
Sadly, testing Virtual Vsync and HyperFormance is much harder than it seems.
In a conventional graphics configuration, all frames are created equal. They all follow the same path through the graphics pipeline, and they all end up on the display. Measuring frames per second—or frame times, as we do in our inside-the-second reviews—yields accurate and valuable information about real-world performance.
Virtual Vsync and HyperFormance turn that concept on its head. Their entire raison d'être is to ensure frames are not created equal—that some frames never make it to the display, or that only certain frames are fully rendered. Statistics about frame times or frame rates instantly become misleading, because they don't discriminate between visible, invisible, and partially rendered frames. That problem is further exacerbated by the fact that Fraps, the tool we use to monitor frame rates and conduct our inside-the-second testing, operates above Lucid's driver in the graphics pipeline. Fraps monitors API calls from the game, just before the Lucid mojo kicks into gear and decides what to do with each new frame.
Here's a hypothetical scenario that demonstrates our problem:
Game X runs at an average of 80 frames per second, and all frames are sent to the display. Enabling vsync (the regular kind) caps the frame rate at 60 FPS. Pretty straightforward stuff. Now, what if we enable Virtual Vsync? The frame rate counter is going to climb back to 80 FPS, but the display will still receive only 60 frames each second. Add HyperFormance to the mix, and unseen frames will dramatically increase in number, since they'll only be partially rendered. Our FPS counter might spike up well over 100 FPS—but the user still won't see more than 60 frames each second.
The reported frame rate may roughly hint at the speed of the game loop, which will in turn gives us a tentative sense of input responsiveness. However, the numbers will give us no indication of what the player actually sees. On top of that, there's no reliable way to measure slowdowns or some of the other issues our inside-the-second testing has exposed.
Faced with no easy way to conduct empirical testing, we decided to test subjectively, instead. Over the next couple of pages, we'll use our keen senses, which have been sharpened by years of graphics hardware reviews, to determine whether Virtual Vsync and HyperFormance improve the gaming experience on the Origin EON11-S. We'll also keep an eye on Fraps' FPS counter to get an idea of what's going on behind the scenes.
Let's start with games that have been officially validated by Lucid to be compatible with Virtu MVP.
The experience in supported games
Despite its age, Infinity Ward's Call of Duty 4 is a poster child for the benefits of Virtual Vsync with HyperFormance. It suffers from the exact problems the technology seeks to resolve, and it's the game Lucid shows off in its demonstrations. We got our own demo at the Consumer Electronics Show earlier this year.
On a system like the EON11-S, Call of Duty 4 runs very fast. Fraps reports frame rates in excess of 130 FPS in the game's starting area, and that's with all the detail options maxed out. COD4 is, as you'd expect, very responsive at those settings—but there's an awful lot of screen tearing, as well. That tearing is especially noticeable during lateral strafing or quick mouse movements. With so many frames being sent to the display, object edges almost seem to shimmer when the camera moves.
We can fix the tearing by enabling regular vsync in the options, thereby limiting the frame rate to 60 FPS. However, that setting has a clear and palpable impact on input responsiveness. Mouse movements feel laggy and sluggish, a bit like sticking your hand in a tub full of water and trying to move it around as quickly as possible. You can feel resistance—something getting in the way.
Going into the Virtu MVP Mobile control panel and enabling Virtual Vsync instantly takes care of that problem. Screen tearing still doesn't occur, but the game responds much more rapidly to input. Fraps' frame-rate counter jumps back up to 130 FPS and above. Remember, though: most of those frames aren't making it to the display.
Does HyperFormance help any? A little. Frame rates as reported by Fraps more than double, rising to 270 FPS and above. Do the math, and it would seem the game loop is updating every 3.7 ms at that rate. There does appear to be a further improvement to input responsiveness, or at least, mouse movements and key presses translate instantly to action on the screen. That's good.
What isn't good is HyperFormance's impact on animation fluidity.
Animation in COD4 isn't perfectly smooth no matter what setting is used. When moving from side to side, for instance, objects don't follow a continuous lateral progression. They tend to slow down and speed up again a few times a second, in what can best be described as a jerking motion. HyperFormance visibly exacerbates that issue. It introduces what we call hitching, whereby animation stops briefly and then jumps forward several frames—sort of like when a scratched CD skips. (Kids, ask your parents.) This hitching happens at random intervals, and it tarnishes the experience enough that you'd probably want to leave HyperFormance disabled in COD4. Virtual Vsync is responsive enough on its own and doesn't make the jerking animations any worse.
Now, what about more recent titles?
Like most modern, graphically intensive games, DICE's Battlefield 3 isn't really a good candidate for Virtual Vsync on the EON11-S. Virtual Vsync only kicks in when the frame rate exceeds 60 FPS, but the EON11-S's GeForce GT 650M graphics processor isn't nearly potent enough to run BF3 that fast with the eye candy turned up. In fact, we had to step down to the "Medium" preset to keep the game playable at 1366x768. Frame rates ended up in the neighborhood of 45-55 FPS at those settings, which was enough to give HyperFormance a shot on its own. Lucid says HyperFormance by itself is effective from 45 FPS on up.
Without HyperFormance, BF3 on the EON11-S looks very fluid, with completely smooth animations and decent input response. That changes when we flip the HyperFormance toggle. Fraps reports a frame rate increase of about 10 FPS, and input responsiveness increases in a palpable way. Movements feel much more instantaneous. At the same time, animation fluidity clearly worsens. Occasional hitching is visible, and much like in COD4, lateral strafing causes objects to jerk across the screen instead of moving in a single, smooth, continuous motion. Screen tearing also seems to cause more severe shimmering around object edges during movement.
This is a tough one. In a multiplayer context—and Battlefield 3 is one of those games conducive to very heated multiplayer battles—you may want to switch on HyperFormance, animation fluidity be damned, just to enjoy quicker response. Your kill-to-death ratio might depend on it. The visual tradeoff can't be understated, though. HyperFormance gives motion a sort of grittiness, like riding a bike on gravel instead of smooth pavement. Some players may find that detracts from the experience enough to live with slightly laggier controls.
Bethesda's The Elder Scrolls V: Skyrim is also a little too demanding to run above 60 FPS at maximum settings on the GeForce GT 650M, which precludes benefits from Virtual Vsync. We settled on the game's "High" detail preset at 1366x786, which yielded frame rates upward of 45 FPS or so.
Skyrim comes with regular vsync enabled by default. Disabling it requires the addition of a parameter—"iPresentInterval=0"— to the Skyrim.ini configuration file in C:\Users\YourUserName\My Documents\My Games\Skyrim. We played the game both with and without that parameter before giving HyperFormance a shot.
With regular vsync enabled, one immediately notices cursor lag in the menus. In-game movement is buttery smooth, but there's definitely a touch of input lag there, too. Modding Skyrim.ini to disable regular vsync does away with cursor lag, and it mitigates some of the in-game input lag, as well. However, it also makes animations jerky and introduces some very obvious screen tearing.
With HyperFormance, frame rates jump to 65-119 FPS according to Fraps, and in-game input lag is reduced. Controls feel practically instantaneous. However, the jerkiness and unevenness of motion is actually worse than with regular vsync disabled and HyperFormance off. In a game like Skyrim that doesn't really prize twitch responses to in-game stimuli, the tradeoff seems too great to be worthwhile.
We tried enabling Virtual Vsync in addition to HyperFormance to see if it might alleviate the unevenness of motion, and it did... sort of. Frame rates returned to their previous level—45-61 FPS or thereabouts—and motion became smoother, but the initial input lag returned. That makes sense. Lucid clearly states that Virtual Vsync doesn't yield benefits over regular vsync below 60 FPS. When HyperFormance is paired with Virtual Vsync, rendering tasks are only skipped for surplus frames that don't make it to the display. There weren't any such frames in this case.
One last thing to note: enabling HyperFormance without Virtual Vsync automatically disables Skyrim's built-in vsync. The user doesn't need to bother with the config file edit. That's pretty convenient, at least.
|Rockchip SoC powers $149 Chromebooks, sub-$100 dongle||8|
|Corsair's M63MM RGB mouse is bringing balls back||10|
|Asus' ROG Sica cuts the gaming mouse to the bare essentials||15|
|Here's why Xeon D could make dual-socket servers scarce||29|
|The TR Podcast 173: Torquing the Titan||4|
|A fresh look at storage performance with PCIe SSDs||36|
|Leaked specs detail Intel's 14-nm Braswell SoCs||37|
|Here are our musings on the new MacBook||156|
|THIS IS THE INTERNET. THERE IS NO PLACE FOR FUN DISCUSSION.||+36|