Nvidia’s monstrous GeForce RTX 4090 and 4080 revealed: 7 must-know details

More than two excruciatingly long years after the RTX 30-series reveal, a new generation of graphics cards is finally here. Nvidia CEO Jensen Huang unveiled the hotly anticipated GeForce RTX 4090 and not one, but two different RTX 4080 variants during the “Project Beyond” reveal event that kicked off GTC 2022.

These graphics cards look absolutely monstrous, full stop, with the RTX 4090 leaving the RTX 3090 Ti stumbling in its wake. But while the performance of these next-gen GeForce GPUs promises to melt your face, the raw speeds are far from the only interesting aspect of this launch. Here are seven must-know facts from the RTX 4090 and RTX 4080 reveal, from ray tracing advancements to staggeringly high new sticker prices for Nvidia’s GPUs.

(For even more coverage, check out the 5 hidden GeForce RTX 40-series details that Nvidia’s CEO didn’t talk about.)

DLSS 3 and Lovelace’s big ray tracing improvements

Nvidia

Nvidia clearly designed its new “Ada Lovelace” architecture to scream through ray tracing tasks. At its core, the GPU uses new “shader execution reordering” to improve traditional game rendering by a claimed 25 percent, but ray-traced scenes will be two to three times faster. Beyond the shader cores, new technologies introduced in the third generation of dedicated ray tracing hardware found inside Lovelace (such as Opacity Micro-Maps) also aim to drastically speed up ray-tracing performance.

But Nvidia’s “RTX” push doesn’t rely on ray-tracing brute force alone. Huang also revealed DLSS 3, running on a new generation of tensor cores and designed to replace the vaunted second-gen version of Nvidia’s AI-enhanced upscaling feature.

While previous DLSS iterations used machine learning to generate new pixels, allowing GeForce graphics cards to run games at lower resolutions (for higher performance) before upscaling them to fit your monitor’s resolution, DLSS 3 generates entirely new frames to figure out optimal upscaled image quality. It doesn’t even have to render the pixels in those behind-the-scene frames, which (Huang claims) allows DLSS 3 to drastically speed up games even if they’re CPU-bound rather than GPU-bound. Huang showed off Microsoft Flight Simulator running over twice as fast with the technology active, and Cyberpunk 2077 saw an even bigger uplift, going from roughly 22 frames per second with ray tracing on in a demo, then 100fps with DLSS enabled. Whoa.

The proof will be in the independent testing pudding, but the eye-melting promise of real-time ray tracing may finally be achievable with the RTX 40-series.

The GeForce RTX 4090 is fast. Really fast.

Nvidia

Nvidia

Since this is GTC—a developer-focused event—Huang didn’t wade too much into traditional gaming performance, but he teased some tantalizing tidbits. The RTX 4090 is up to two times faster than the RTX 3090 Ti in traditional games like Flight Simulator; three times faster in a new ray-traced version of Portal revealed at the announcement; and a whopping four times faster in Nvidia’s own “Racer X” ray tracing tech demo (which, to be fair, is likely optimized to take advantage of Lovelace’s biggest technological advancements).

The GeForce RTX 4090 is not for gamers

Much like the RTX 3090 before it, the RTX 4090 may technically be a GeForce-branded graphics card, but its not really for gamers. Instead, with its ample 24GB of GDDR6X memory and staggering $1,599 price tag, it once again looks like an absolute powerhouse for prosumers who can put the GPU to work during the day, then enjoy face-melting framerates in games at night.

Again, however, we’ll need to wait for independent reviews to be sure. But most gamers would no doubt be better off looking at the new duo of RTX 4080 options.

The RTX 4080 12GB and 16GB have different GPUs and abilities

Nvidia

Nvidia

Nvidia revealed a pair of GeForce RTX 4080 offerings. One comes with 16GB of GDDR6X memory and starts at a staggering $1,199—the same price as last-generation’s RTX 3080 Ti. Oof. The 12GB RTX 4080, on the other hand, starts at $899.

There’s a massive gulf in pricing between the two models, but Huang didn’t get into technical differences between them, and didn’t even tease gaming performance for the 4080 duo. Usually, identically named GPUs only change in the memory capacity offered, but early RTX 40-series rumors suggested the 12GB RTX 4080 might actually wield a less-potent, totally different GPU than the 16GB model, despite rocking the same name. Lets hope that doesn’t hold true, as it’s a deeply anti-consumer practice, but the wild pricing hints that it might be. Stay tuned.

Nvidia

Update: Nvidia’s website included some specs for the RTX 4080 models, which you can see above. There is indeed a big difference in not just the GPU capabilities and memory capacity, but also memory bandwidth, with the 12GB 4080 using a much narrower memory interface.

RTX 40-series is much more expensive

Now for some bad news. Worries about even-higher price points for the RTX 40-series proved accurate. The RTX 4090 starts at $1,599, a $99 premium over the RTX 3090. The 16GB GeForce RTX 4080 costs $1,199, the same price as the RTX 3080 Ti and an astonishing $400 more than the $700 RTX 3080. Finally, the mysterious 12GB RTX 4080 starts at $899, a $200 increase over its predecessor.

Ick.

The GeForce RTX 4090 launches October 12. Availability for the RTX 4080 duo is unknown.

The RTX 30-series is sticking around

Nvidia

The RTX 40-series may be here, but Huang said that the RTX 30-series will be sticking around at more mainstream price points. It makes sense—the fall of cryptocurrency means that GPU makers are currently sitting on warehouses full of unsold graphics cards. But those new RTX 40-series prices hurt, and if the RTX 4090 and 4080 don’t come out of the gate swinging hard, unsold RTX 30-series GPUs and used graphics cards may very well be a much better value. We’ll see.

AMD’s RDNA3 counterpunch is coming soon

Finally, while Nvidia beat AMD to the punch with the launch of this new GPU generation, Radeon’s counterpunch ain’t far behind. Mere hours before Jensen Huang’s keynote, Radeon chief Scott Herkelman tweeted that AMD’s next-gen RDNA3 Radeon cards will be revealed November 3, while AMD Technology Architect Sam Naffziger penned a blog post touting RDNA3’s vast efficiency improvements—a direct shot at the presumed high power levels that Nvidia’s RTX 40-series will require to run at peak speeds.

It’s been a bleak couple of years for PC gamers, but now cryptocurrency is collapsing, supply chain woes are evening out, and we’ve got a fierce battle for next-gen GPU supremacy brewing. Giddy up—and start saving those pennies. Those prices are high. For even more coverage, check out the 5 hidden GeForce RTX 40-series details that Nvidia’s CEO didn’t talk about.
Gaming, GPUs