Sabtu, 11 Mei 2019

A tale of two “inside-out” VR headsets: The $400 Oculus Rift S, $600 HP Reverb - Ars Technica

The $600 HP Reverb (left) and $400 Oculus Rift S (right).
Enlarge / The $600 HP Reverb (left) and $400 Oculus Rift S (right).

By the end of 2019, many major VR headset manufacturers seem poised to launch a new "statement" product for PCs. This month sees two such headsets reach store shelves: the Oculus Rift S (coming May 21, priced at $400) and the HP Reverb (out now, starting at $600).

In both companies' cases, the statement from each headset is a mix of upgrade and compromise. Rift S sees Oculus take two steps forward, two steps back, from its three-year-old Rift headset to establish a new "baseline" PC-VR experience, particularly with active hand tracking in mind. Meanwhile, Reverb aims to deliver the most affordable "high-res" VR headset ever made—which, as you might expect, includes a few imperfections, ranging from the obvious to the surprising.

After living with both headsets, I can report that each headset's sales pitch is totally fine, not game-changing, and both are worth scrutinizing—because neither is currently a slam-dunk recommendation.

HP Reverb: You say you want resolution

Let's start with the HP Reverb, a headset that promises to exceed the screen quality and pixel density of the already-impressive HTC Vive Pro and Samsung Odyssey. Reverb's total resolution weighs in at a combined 4,320x2,160 across its fast-switching LCD panels.

As a result, before HP agreed to send us a testing unit, we were asked to confirm a graphics card minimum of a GTX 1080 or AMD Radeon Pro WX 8200. Notice that their latter suggestion is designed for workstations, not consumer PCs. That's the point. The Reverb is squarely targeted at an enterprise or development use cases where screen quality is paramount; this is not necessarily your headset for gaming or high-speed interactivity.

This is first made apparent by an awkward headstrap. Unlike the springy flex of the original Oculus Rift or the rotary dial of the PlayStation VR, Reverb's strap is spartan. You get three velcro straps to adjust, which are all difficult to adjust while wearing the headset. This is all bound by a light-yet-wide halo, meant to fit against the back of the head, and it comes with two decidedly cheap-feeling over-ear speakers—with fuzzy material you'd expect from $5 airplane headphones. (You can't detach these, but you can push them off your ears and plug in your own headphones with a 3.5mm jack.)

Yet this spartan approach means the strap system is under-engineered, a rarity in the sector. The headset is substantially lighter in mass than the over-engineered Vive Pro, which may be worth some inconvenience. Its back-halo design keeps more weight off your face than the original Rift. Plus, fit it onto your exact head shape once, and it's simple enough to put on and take off from there.

The Reverb's flip-up functionality is flimsy and bumps into your head, so it disappoints as a "peek at the outside world" option. Its cramped interior is not particularly glasses-friendly. And if your interpupillary distance (IPD) measures outside the "average" measure of 61-66mm, Reverb's "digital IPD" adjustment option will leave you unsatisfied from a comfort level.

All of which is to say: If your expected business/enterprise use case includes mostly VR-savvy professionals of a perfect head shape and size, you'll be fine. If you're handing these to a variety of utter VR newbies, on the other hand, be ready to on-board them just to get the thing on.

Sweet spot, not-so-sweet result

As for the Reverb's screens, the pixel resolution of 4,320x2,160 is incredible for the sector, well past the 2,800x1,600 measure of the Vive Pro and Odyssey. (And that's even more than the Rift S, which we'll get to.)

Reverb's "sweet spot" visibility, in the center of a user's field of vision (FOV), is the absolute winner in its price sector. After roughly one week of Reverb testing, I'm now convinced that this is the pixel count to count as "good enough" if you want to guarantee unobtrusive small-text legibility for the sake of VR's research, education, and industrial design apps. (In Reverb's case, this selling point is buoyed by the LCD panels' dense subpixel resolution and 90Hz refresh.)

HP provided some sample "professional" apps during my testing period, and after wandering through a virtual replica of Helsinki and dissecting a frog in a classroom, I understood why. With my attention focused on front-and-center content, I could see the Reverb's sales pitch damned clearly. HP has crossed an important VR-quality line at a reasonable business price point in 2019, and the rest of this review's caveats and warnings can't erase that fact.

One of those caveats, however, is my use of "sweet spot" as a qualifier. I struggled to understand why the high-res panel, all-in-all, looked a bit smeary ("a bit" is not a very scientific descriptor, after all). That issue became clearer once I set up a "VR desk" of a headset, a single hand-controlled WMR controller, a physical keyboard, and a floating VR replica of my PC's 2D desktop.

Doing this confirmed my suspicions: the HP Reverb, like many other VR headsets, offers a clear view in its central FOV but is less successful at translating its peripheral pixels. Trying to read text in the periphery was a struggle compared to the same text appearing front-and-center. This was particularly easy to notice as I examined details spread across 16:9-ratio desktop windows and Web browsers. Without a physical IPD slider to work with, and only a "virtual" IPD adjustment option, I had no idea how to remedy this apparent blurriness. Countless attempts to re-fit the headset didn't help.

To be fair, the fact that I could comfortably read Windows desktop content is its own VR revelation. That's almost impossible to comfortably do on the first wave of 2016 VR headsets. But the Reverb's arrangement of pixels and lenses does no favors to peripheral-view content, which leads to an uncanny valley-like issue: once some of the content is so damned crisp, why can't all of it be? Additionally, why must the headset be so demanding of PC hardware if it blurs its peripheral pixels by default?

Some headset manufacturers are toying with foveated rendering, which reduces pixel resolution depending on what's being displayed or how a user's eyes are tracked. But nothing so efficient is happening here. The Reverb instead renders, then wastes, at least one fourth of its pixels, which I can tell by budging the headset awkwardly around until only its corner pixels become clear.

LCD concerns

That divide in clarity means all the comfort you might hope for from a higher-res headset dwindles in longer-term use. This is a shame because the fast-switching LCD panel lives up to a 90Hz refresh with fuller subpixel resolution than comparable OLED panels. However, the "halo" effect from its fresnel lenses is particularly noticeable within this headset.

Additionally, HP's choice of fast-switching LCD panels means the Reverb simply suffers from imprecise color calibration—at least, compared to the rich, RGB-perfect results you can expect from a calibrated OLED panel. Part of that Vive Pro $1,100 price tag is an understanding that whatever content you bring in will enjoy nearly uniform color reproduction. But the Reverb's "cold" blue-green wash, which is nigh indiscernible when looking at a standard CMYK color sheet, becomes apparent across a wider scene, particularly the pastoral, faux-outdoor environs of the hub spaces in Windows Mixed Reality and SteamVR.

In VR experiences that rely on moody color mapping (like the storybook-adventure of Moss), the color reproduction borders on problematic. Everything in that game looks darker and less alive than on the OLED-fueled HTC Vive Pro. A visible "mura" effect on our testing headsets resulted in uneven color reproduction across wide fields of pixels, as well. It was arguably the most intense mura effect I've ever seen on a consumer-grade VR headset, in fact. (To best explain this, think of a large website background color, which is supposed to be totally uniform, having an uneven smudginess to it. Next, imagine that smudginess moving in relation to where your head is aimed in VR.)

The usual Windows Mixed Reality caveat

If you're fine with somewhat imprecise color reproduction, high system demands, and an asterisk on Reverb's high resolution, you have one more pill to swallow: its merely adequate room-tracking powers. This is identical to most inside-out Windows Mixed Reality headsets, which rely on two forward-facing cameras and guarantee decent tracking, so long as you keep your hands generally in front of your face.

The short version: milder apps like TiltBrush and Vacation Simulator work just fine. (So did the educational apps that HP provided.) Highly active apps like Beat Saber and Space Pirate Trainer, on the other hand, have to compensate for Reverb's poorly tracked hands on a pretty regular basis—as in, every 45 seconds in a high-speed Beat Saber song, a hand will noticeably disappear for a moment. Meanwhile, any games that rely on above- and behind-the-head hand action are out the window, as most WMR headsets (including the Reverb) don't have upward-facing tracking sensors.

In good news, most popular VR fare expects lighter tracking, and as a result, WMR-style tracking will work in a pinch. But if the idea of randomly disappearing hands is too VR-breaking for you, then you'll want to pony up for a fuller tracking experience.

Let's block ads! (Why?)


https://arstechnica.com/gaming/2019/05/a-tale-of-two-inside-out-vr-headsets-the-400-oculus-rift-s-600-hp-reverb/

2019-05-11 13:00:00Z
52780291535732

Tidak ada komentar:

Posting Komentar