Benchmarks and Conclusion
The setup process for Windows MR headsets is about as simple as VR calibration gets. You need a Windows 10 PC with at least version 1809 installed, a free USB 3.0 port and the appropriate display output.
To begin the setup, plug in the USB and DisplayPort cables, and Windows should automatically detect that there is an MR headset present and launch the Windows Mixed Reality Portal software. If the software isn't installed, it should prompt an automatic Windows update.
Once the software opens, it asks you to pair the controllers. As noted above, the controllers that ship with the Reverb come pre-paired to the headset, but if you ever need to pair a new set, you'll find the pairing button under the battery cover on each controller.
Windows MR works in room-scale and standing / seated configurations, but room-scale is required for some games and applications, so choose that option if you have space. Like with the original Windows MR headsets, you must carry it around your desired play area to map the borders for safety.
Once the borders are set, Microsoft drops you into a quick training session that teaches you how to interact with the virtual Windows MR environment. The process takes a few minutes, and you'll learn how to target objects, teleport through the environment and call upon Cortana for aid. When it completes, you’re transported to Microsoft's home environment, where you can launch apps.
For performance testing, we used a PC equipped with 4.8-GHz Intel Core i7-8700K paired with 16GB of DDR4-3666 RAM and an Nvidia GeForce RTX 2080 Founders Edition graphics card. We performed our tests with Nvidia's latest GeForce driver, version 430.39.
SteamVR lets you adjust the rendering resolution of your headset and automatically detects your hardware before recommending a setting based on your system's capability. Curiously, SteamVR assigned a resolution of 1609 x 1576 as the 100% value for the Reverb, which is a far cry from the actual panel resolution of 2160 x 2160. SteamVR suggested 200% rendering scale for our hardware. That worked out to be 2275 x 2229, which is much closer to the native resolution of the panels.
In addition to the Reverb headset, we also included a Vive Pro for comparison. We set the Vive Pro's resolution as close to the Reverb's as possible. The two headsets have different aspect ratios, so you can't match the resolution perfectly. Instead, we tried to match the pixel count of each headset's rendering resolution.
At 200%, SteamVR delivers 5,070,975 pixels to the Reverb headset, whereas our Vive Pro receives 5,059,714 pixels when we set the render resolution to 112% or 2134 x 2371.
It's worth noting that SteamVR suggested 154% for the Vive Pro, which translates to 2582 x 2869 per panel, or 7,407,758 pixels. To compare against that resolution, I set the Reverb resolution to 2749 x 2693 (292%), which works out to 7,403,057 pixels rendered.
Hand Tracking Issues
Microsoft's tracking system worked well enough to determine the headset's position, and it captured the controllers when they were directly in front of the headset. But the tracking FOV is narrow, and unlike Oculus' new Insight tracking technology, the Windows MR software doesn't do an outstanding job of estimating the controllers' position when they're out of range.
After less than a second of holding my hand out of the camera's range, I found my controller drifting all over the screen. That was a regular occurrence because there are only two cameras on the headset, and they aren't even on the outer corners of the device. The Reverb's tracking cameras are positioned almost a full inch from the edge of the faceplate, which prevents the headset from seeing anything beside it.
A standalone VR headset like the Oculus Quest can handle a fast-paced round of Beat Saber without a hint of tracking issues, so it's embarrassing when a PC-connected system can't keep up. Microsoft really needs to update its camera system or enable hardware makers to use more than two of them so we can have better hand tracking with Windows MR devices.
The test suite included Space Pirate Trainer, Serious Sam VR: The Last Hope, Arizona Sunshine and Beat Saber. We intended to include performance figures for Skyrim VR and Project Cars 2, but both of those games failed to launch. Skyrim refused to open while FCAT VR was active, and Project Cars 2 wouldn't start while the Reverb was plugged in. I don't have another Windows MR headset on hand to rule out Microsoft's VR platform as the issue.
Space Pirate Trainer
Space Pirate Trainer is a fast-paced game with a lot of bright colors over a dark environment, which is a great test for VR displays. The LCD panels in the HP Reverb produced vivid bright colors and did an adequate job with the dark sky in the background. Thanks to the high resolution, every object on the screen appeared crisp and clear.
Space Pirate Trainer appears to be optimized better for the Reverb than it is for the Vive Pro. The FCAT VR results show that the unconstrained frame rate on the Vive Pro was around 250 frames per second (fps), whereas our system sent up to 290 fps to the Reverb.
With the resolutions set to their higher values, we observed nearly identical unconstrained frame rates in the range of 195 fps. At no point did our frame rate drop below the required 90 fps threshold.
Beat Saber isn't a very graphically demanding game, but it requires you to move your hands at a rapid pace. The harder difficulty levels demand swift movement, and poor graphics or display performance could have a meaningful impact on your potential score. The Reverb's ultra-low latency displays did a great job with this kind of usage.
The RTX 2080 in our test system had no trouble delivering enough frames to keep up with the refresh rate of the display. We observed an unconstrained average of over 160 fps, which is nearly double the necessary 90 fps minimum.
With the higher resolution settings, both headsets again saw virtually equal average frame rates. The Reverb enjoyed 127.39 fps, while the Vive Pro saw 127.6 fps on average.
Interestingly, the Reverb’s average delivered fps was a touch below the desired 90 fps with an average of 88.71 fps. This may denote an optimization issue with the Windows MR platform, as the unconstrained performance numbers suggest our system had ample performance to handle to job.
Serious Sam VR: The Last Hope
Serious Sam VR takes place in a bright desert environment with a lot of colorful explosions and firepower, which again lends well to the RGB LCD panels in the Reverb headset. This game highlighted how much better the display in the Reverb compared to the Vive Pro. The higher resolution was extremely evident in the menus of this game.
Oddly, Serious Sam VR: The Last Hope is the only game where frame rate fell at the recommended resolution. Our test system didn't have any trouble keeping the frame rate steady on the Reverb at Steam’s recommended resolution, but each time we tested the Vive Pro the results showed that a handful of frames didn't reach their intended destination.
Despite achieving an unconstrained framerate of 245.6 fps, the average frame rate delivered to the Vive Pro was a mere 88.53 fps. The Reverb headset sustained 90 fps at all times, but it had lower unconstrained performance at 240.09 fps.
With the resolution cranked up, our system failed to sustain 90 fps on both headsets. The Vive Pro scored 89.25 fps, which is just shy of the mark. The Reverb suffered with 86.91 fps, which was a noticeable deficit while playing this fast-paced game.
Arizona Sunshine was the most demanding title in our test suite. I wouldn't say the image quality played a role in the gameplay quality with Arizona Sunshine, but the visuals were noticeably better in the Reverb headset compared to the Vive Pro.
Our test system still managed to deliver more than 90 fps to the headset, but the two headsets’ results were much closer this time. We didn't observe any dropped frames in either headset at their recommended rendering resolutions.
The Reverb averaged a flat 90 fps, with the unconstrained frame rate pegged at 127.43 fps. The Vive Pro averaged 89.98 fps with unconstrained performance good for 141.4 fps.
When we raised the bar with a higher resolution, our test system began to struggle. With the Vive Pro, it did just fine with an average frame rate of 89.56 fps. There wasn’t much headway with the unconstrained performance tagged at 104.82 fps.
The Reverb set to 2749 x 2693 proved to be too much for our RTX 2080-equipped test system. We observed an abysmal 76.33 fps average, and even the unconstrained frame rate couldn’t feed the 90 Hz display.
At this frame rate aiming was difficult, and the game started to make me feel dizzy. It is not worth pushing the resolution to the point where you lose performance.
HP's Reverb is easily the best VR headset in Microsoft's Windows MR lineup. It's a comfortable device, the visuals it produces are crisp and clear and like all Windows MR devices, it's relatively simple to set up.
However, a crucial part of any VR experience is the quality of the hand tracking and input method, and this package falls far short of what I would call the bare minimum for tracking fidelity for 2019. The faulty tracking of the controllers—not to mention the poor ergonomic design—makes it difficult to recommend the Reverb headset over competitors like the HTC Vive Pro.
The price tag doesn't help either. The Reverb starts at $599 for the consumer version, and the Pro edition sells for $649 (it includes a 0.6m headset cable for connecting to the optional HP Z VR backpack). That's a lot of money for a headset with sub-par tracking accuracy and inferior controllers--especially considering the recent competition in the VR headset market.
You would need a good reason to favor the higher resolution over better controllers. Unless the highest resolution image is of critical importance to how you plan on using the headset, you might want to consider other cheaper HMDs over this one, which is a real shame because the image clarity in the Reverb is quite exceptional.
Image Credits: Tom's Hardware
MORE: Virtual Reality Basics