How VR Headsets Work: A Sim Racer’s Guide to Lenses, PPD, and What Actually Matters

I’ve been racing in VR for the better part of three years. A Reverb G2 first, then a Quest 3 for about a year, and a Pimax Crystal Light since 2025. Three different pictures, three different sets of compromises, and one realisation: most of the VR spec sheets aren’t telling you the thing that actually matters in your cockpit.

The TL;DR is that lens type and PPD do the work. Refresh rate and FOV are mostly marketing once you cross a basic threshold. So today, we’re going to look at how a VR headset actually works, which numbers on the box are doing the heavy lifting, and which are doing the marketing.

Quick Navigation
Jump directly to what you’re looking for:
What’s inside a VR headset | The render pipeline | The specs that matter | 2026 comparison table | Why I race in a Crystal Light | Where to start in VR sim racing | Common mistakes | Where this is heading

What’s inside a VR headset, mechanically

A VR headset is two small displays held close to your eyes, with lenses to focus them, sensors to track where your head is in the room, and a cable (or wireless link) feeding it from a PC that’s doing the visual rendering. That’s basically it! Everything else is a refinement on those four things.

The two displays render slightly different views of the scene, one from where your left eye would be and one from your right. Your brain reads the parallax and decides the image has depth. The lenses sit between the display and your eye and do two jobs at once: they let you focus on something a few centimetres from your face (you can’t, normally), and they magnify the panel so it fills your field of view. The tracking sensors – either inside-out cameras on the front of the headset or external lighthouse base stations on your rig – tell the system where the headset is in the room so the rendered view follows your head.

The lens layer matters more than people realise. I say lenses, but there are three quite different lens technologies in use right now, and they’re not the same thing at all. Pimax put together a useful technical breakdown that’s worth twenty minutes of your time.

The short version: Fresnel lenses are cheap and light, but suffer from glare and god-rays at the edges. Pancake lenses are small and light too, but they only let through around 15% of the light from the panel because the light has to bounce back and forth between polarisers inside the lens. Aspheric glass lenses are heavier and bigger, but they let through about 99% of the light and give you edge-to-edge clarity instead of a small sweet spot in the centre.

The Crystal Light and Crystal Super use aspheric glass. The Meta Quest 3, the Bigscreen Beyond 2, and the Pimax Dream Air use pancake. The Reverb G2 you might be replacing used Fresnel. The difference shows up in the cockpit the moment you have to scan across the panel to a mirror or a brake board. You don’t turn your head 45 degrees to check the driver’s mirror in a sim rig – you flick your eyes. With a Fresnel headset, the edges of your vision are a smeary mess and only the centre is sharp. With a good pancake lens the sweet spot is bigger but still smaller than the panel. With aspheric glass, the entire panel is usable. The first time I checked my apex at Spa through a Crystal Light after a year on the Quest 3, I noticed it before I noticed anything else.

What’s happening on the PC (and why your GPU is the bottleneck)

The bit that most beginner VR explainers gloss over is the render pipeline. Your sim sees the headset as a display that needs two views per frame, one for each eye, both at the headset’s native resolution. So, if you’re running a Crystal Light, the GPU is rendering 2880 by 2880 pixels twice per frame, 90 times a second. That’s roughly four times the pixel count of a 1440p monitor! Naturally, that’s why VR feels heavy on hardware that runs flat-screen sims comfortably.

Between the sim and the headset sits a runtime, where SteamVR is the default for most headsets. OpenXR is the newer, leaner standard. Pimax Play is Pimax’s own runtime. Meta has its own, too. Whichever you’re using, this layer is where the frame-timing decisions live, and it’s where most of the day-to-day headaches come from. Pimax Play 2.0 cuts about 11ms of latency for me versus running through SteamVR’s compositor. Doesn’t sound like much until you’re trail-braking into Eau Rouge and your visual feedback is a tenth of a second behind your hands (latency is undescribably weird – the less the better).

Let’s look at what’s called foveated rendering. The idea: your eye only sees sharp detail in a tiny circle around wherever you’re looking. Everything else is peripheral vision. So why bother rendering the periphery at full quality? Fixed foveated rendering shaves quality off the corners of the frame regardless. Dynamic foveated rendering (DFR) uses eye-tracking cameras to follow where you’re actually looking and shave quality off the bit you aren’t.

On a headset like the Crystal Super, DFR paired with iRacing and MSFS 2024’s native Quad Views support delivers what one of Pimax’s engineers called “equivalent to a new GPU for free.” Not every sim supports it, and it needs eye tracking in the headset, but where it works, it’s the biggest single VR performance lever you can pull.

Best specs that matter for sim racers

If you’re staring at a Pimax product page and trying to decide which of fourteen numbers to care about, here’s my short version. IMO: Lens type and PPD do the work. Refresh rate and FOV are mostly marketing.

PPD (pixels per degree)

Resolution per eye is what gets put on the box. PPD is what you see. Take the total horizontal pixel count and divide it by the horizontal field-of-view degrees. That’s pixels per degree of your vision. The lower the number, the more obvious the screen-door effect. The higher, the closer you get to looking at the real world.

A cinema analogy, if you’ll indulge me: on the front row, the picture is huge and you can maybe see the pixels. Meanwhile in the back row, the picture is smaller but sharper because the same pixels are packed into a smaller part of your vision. That’s exactly the PPD trade-off in a VR headset. The headset’s panel is the screen. Your field of view determines how close you’re sitting. Higher resolution moves you back. Wider FOV moves you forward. It’s very clever really, as resource allocation can be prioritised to where the VR wearer is looking.

20 PPD is playable but mushy. 30 to 40 is the sweet spot. 50-plus is approaching the limit of what your eye can resolve. 60 is roughly where the human retina gives up. Crystal Super at 57 PPD is close to a retinal display. The practical sim-racing payoff is reading brake markers at distance. On a Quest 3 (25 PPD), a 150-metre brake board at the end of the Kemmel Straight is a pixelated smudge until you’re nearly on it. On a Crystal Light (35 PPD), you can pick it out at the right braking distance. On a Crystal Super, you can read the number on the board.

One methodology note worth flagging, because it confuses readers comparing across vendors. Meta and Somnium quote “peak angular pixel density” measured at the centre of the lens. Naive math (total pixels divided by FOV) gives a lower number. The 25 PPD on the Quest 3 is the optimistic measurement; the real-world average across the lens is closer to 19. Pimax’s numbers are similarly centre-of-lens peaks. The comparison table below is still useful, but it isn’t perfectly apples-to-apples.

Refresh rate – overrated for sim racing

The box says 120Hz. In practice, you will not run iRacing in a Crystal Super at 120Hz on a full grid with rain effects. You will brutally bottleneck even an RTX 5090. What you will run is 90Hz or 72Hz, locked, with frame-pacing that doesn’t dip into reprojection. A rock-solid 90 beats a fluctuating 120 every lap. Buy for the panel’s lowest stable refresh, not its top number.

Field of view – also overrated past about 110 degrees

Wider FOV gives an immersive sensation of speed, but it stretches your panel pixels thinner (lower PPD) and asks your GPU to render scenery you barely look at. Above 110 or 115 degrees, you’re paying serious performance cost for periphery your brain mostly ignores while you’re staring at the apex. The Somnium VR1’s 130 degrees is impressive in the showroom and a GPU drain in the rig. The 105 to 110 range on the Pimax Crystal Light and Crystal Super is, in my experience, the better trade.

Weight – underrated, especially for endurance work

The Crystal Light is 815 grams. The Quest 3 is 515. The Bigscreen Beyond 2 is 107. You don’t really feel the difference at lap one. You feel it at lap forty. For a one-hour sprint, the Crystal Light’s weight very exciting. For a Le Mans 24 stint plan, you’ll be ordering aftermarket pads inside a fortnight. Add £40 to £60 to the real cost of ownership for any of the heavier headsets if endurance is your thing.

Tracking – inside-out is fine for seated sim racing

SLAM (inside-out, cameras on the headset) versus Lighthouse (outside-in, base stations on your rig) is a real debate for body-tracking and motion-capture use cases. For seated sim racing, where the headset only moves a few centimetres in any direction, SLAM is fine. The Crystal Light and Crystal Super have interchangeable face plates that let you switch between them. The Dream Air picks one at order time. The Bigscreen Beyond 2 and Somnium VR1 are Lighthouse-only and want a pair of base stations – add £200 to £300 to the kit cost if you don’t already own them.

The 2026 comparison: three Pimax headsets and three competitors

At the moment there are probably, 6 headsets I’d put on a sim rig in 2026. Three are Pimax (Crystal Light, Dream Air, Crystal Super). The other three: Meta Quest 3 as the mainstream PCVR default, Bigscreen Beyond 2 as the premium PCVR specialist, Somnium VR1 as the high-end enthusiast play.

Headset	Display	Resolution per eye	Max refresh	FOV (horiz)	PPD	Lens	Tracking	Weight	Price (early 2026)
Pimax Crystal Light	QLED + Mini-LED	2880 × 2880	120 Hz	~105°	35	Glass aspheric	Inside-out (SLAM)	815 g	~£799 / $899
Pimax Dream Air	Sony Micro-OLED	3840 × 3552	90 Hz	~110°	~53	Concave pancake	Inside-out (SLAM)	under 170 g	~$1,299
Pimax Crystal Super (57 PPD QLED)	QLED + Mini-LED	3840 × 3840	90 Hz	~106°	57	Glass aspheric	Inside-out (SLAM)	~800 g+	~$1,599 to $1,799
Meta Quest 3	LCD (dual)	2064 × 2208	120 Hz	~110°	25	Pancake	Inside-out	515 g	£479 / $499
Bigscreen Beyond 2	Micro-OLED	2560 × 2560	90 Hz	~108°	32	Custom pancake	SteamVR Lighthouse	107 g	£899 / $899 + base stations
Somnium VR1	QLED + Mini-LED	2880 × 2880	144 Hz	~130°	35	Dual aspheric glass	SteamVR Lighthouse	~750 g	~£1,600 / $1,899 + base stations

Three things jump out of that table. First, the Pimax lineup occupies the highest PPD slots by a clear margin. Second, the only headset under 200 grams is the Bigscreen Beyond 2, and you pay for that lightness with required external base stations.

Third, the Quest 3 is half the price of anything else here, and for a sim racer trying VR for the first time without committing to a £900-plus bill, it’s the obvious starting point – we’ve written about the Quest 3 to Crystal Light upgrade path separately if that’s where you’re heading.

Why I race in a Pimax Crystal Light

Firstly, the conclusion: glass lenses and PPD. Those are my two big things.

I’d been running my Quest 3 for about a year before I tried a Crystal Light. The difference at Eau Rouge was so obvious I’d ordered one inside a fortnight. Sharper braking-zone visibility, no edge degradation, no smudgy outer thirds of the panel when I scanned to a mirror. The Quest 3 was fine – the Crystal Light made me realise what I’d been compensating for.

The honest case against, though, is real. The Crystal Light is lighter but still a little heavy. The Pimax software was a “fiddle factor” for years, and while Pimax Play 2.0 has closed most of that gap, the reputation lingers because the reputation was earned. There’s no passthrough on the Crystal Light, so you’re lifting the headset to find your keyboard between sessions. The Lighthouse face plate Pimax announced for the Crystal Super has slipped multiple times, which matters if you already own SteamVR base stations and want to use them.

And the GPU bill is real. A 4070 Ti can drive a Crystal Light. A 4090 or higher runs it without thinking. A Crystal Super on a full iRacing grid wants more than that. If your rig is mid-range, factor the GPU upgrade into your headset budget before you commit. Triple screens at 1440p are still cheaper, simpler, and easier to share with a family member who isn’t into the hobby. For me, the immersion of VR – and the sheer track-feel you get when the depth perception is right – is worth the trade-offs. For some sim racers, it isn’t.

My Recomendation for VR sim racing

If you’ve already done with your Quest 3 and you know you want to commit into VR, the Pimax Crystal Light is where I’d go. Glass aspheric lenses, 35 PPD, 105 degrees of FOV, a price tag that’s roughly twice the Quest 3 but a clarity uplift that’s genuinely a step change. Pair it with a 4070 Ti minimum, a 4090 if you can stretch.

If money isn’t really the constraint and you want the best clarity available right now, the Pimax Crystal Super (57 PPD variant) is the headset to look at. It comes with the same caveats as the Crystal Light on weight and software, plus a heavier GPU bill, but the spec sheet does back up the price tag. The Crystal Super QLED SuperOpen discount has been bringing the price down to $1,599 periodically, which makes it slightly easier to swallow.

If lightweight is your priority – The Pimax Dream Air looks brilliant on paper. I haven’t spent enough time with one to give an opinion that’s worth more than the spec sheet, but the Dream Air preview covers what I’m watching for.

Common things sim racers get wrong about VR specs

1. Higher refresh rate is always better. It isn’t. A stable, locked frame rate beats a fluctuating one every lap. If your GPU can’t sustain 120Hz at the headset’s native resolution, drop to 90 and lock it. Reprojection artefacts hurt more than smooth 90Hz ever will.
2. Wider FOV is always better. Past about 110 degrees, you’re paying GPU cost for periphery your brain isn’t using. The sweet spot is “wide enough that you don’t see the bezel of the lens” rather than “as wide as the spec sheet allows.”
3. You need a 4090 for VR. You need a 4090 for a Pimax Crystal Super on a full iRacing grid in the rain. You don’t need it for a Quest 3 on most sims. Match the GPU to the headset, not to whatever a sim racing YouTuber is breathlessly recommending this week.

---

Related Posts