The Nvidia GeForce RTX 4090 is finally here, arriving after one of the most contentious graphics card generations in recent memory. While the performance for Ampere-based cards like the RTX 3080 and RTX 3070 was definitely there, the cards themselves noticeably were not. Like many other things, the pandemic led to a shortage in graphics cards that was only exacerbated by cryptocurrency miners.
At the end of the day, what this all boils down to is that many people that were trying to buy an RTX 3080 when it came out simply weren’t able to get one, and the first question on my mind as the RTX 4090 shows one of the biggest generational leaps in PC gaming performance is—will people actually be able to buy this $1,599 graphics card? I hope so, because this thing is an absolute beast.
If the Nvidia GeForce RTX 3080 made 4K gaming mainstream, the RTX 4090 trivializes it, especially if you’re playing your games at 60 fps, like many do. That might make the RTX 4090 sound like overkill, and it kind of is, but there are so many high refresh rate 4K monitors coming out that will really take advantage of the power on offer here, and Nvidia takes it even further with yet another iteration of its DLSS—or Deep Learning Super Sampling—technology.
Starting with the Nvidia GeForce RTX 4090, and later the RTX 4080 and likely the RTX 4070 sometime in the next quarter, Nvidia is introducing its new Ada Lovelace graphics architecture. This isn’t as radical an improvement as last time around, but it does introduce brand-new Tensor and ray tracing cores that radically improve performance in specialized workloads.
It’s through these new and improved Tensor cores that Nvidia is able to introduce optical frame generation, and thus DLSS 3.0. Basically, Nvidia found a way to take a lot of the work away from the processor, and use its deep learning algorithms to generate full frames, which makes performance absolutely skyrocket in games that support it. The downside is that there aren’t a lot of games that support it just yet.
The high price of entry is going to make that little caveat kind of hard to swallow, too. We’re talking an entry price of $1,599.99 here. That’s a lot of money for anyone already, but there are going to be so many versions of this card that will cost so much more. That’s because in the world of graphics cards, Nvidia might make one Founders Edition—what was reviewed here—but the RTX 4090 will also be manufactured by a cavalcade of third-party GPU makers, some of which will radically increase the asking price in exchange for spiffy new coolers or shiny designs.
But beneath every version of the RTX 4090, you’re essentially going to be getting the same thing. Namely a graphics card with 128 Streaming Multiprocessors, which are essentially clusters of GPU cores. Each of those has 128 CUDA cores for a total of 16,384—making this the most complicated graphics processor ever released to a mainstream market. With such a huge amount of hardware, there’s also a really big power requirement, with Nvidia rating it at a 450W TGP, or total graphics power.
If that sounds like a lot, it’s because it is, and Nvidia is well aware of that. It’s recommending at least an 850W power supply with this graphics card, but I’d go even further and recommend a 1,000W unit, just to be on the safe side. Because while the RTX 3090 launched with a power adapter with just two 8-pin connectors, the RTX 4090 has a four-connector adapter. Only three 8-pin connectors are mandatory, but even still, it consumes a ridiculous amount of power.
One side-effect of this power requirement is that the new 12-pin power connector that Nvidia introduced with the RTX 3080 doesn’t supply enough juice for this GPU, which means a brand new connector. It’s pretty similar to the old 12-pin connector, but it’s now a 16-pin connector. Luckily, Nvidia did include an adapter in the box — which also means messy cable management is here to stay.
I did use all four 8-pin power connectors, even if I had to buy a fourth one to make that work, and through all of my testing, the RTX 4090 peaked at around 434 watts, just under Nvidia’s rated 450W TGP. That will go up if you dabble in overclocking, and there’s definitely plenty of room for that with this graphics card. Not only does Nvidia provide an adapter that will provide enough power to push this card into overdrive, but with the stock settings, this GPU only ever reached 67°C, well below the 90°C max temperature for the RTX 4090.
Much of the reason the graphics card operates so cool is due to the physical design of the card. Much like the Founders Edition versions of the RTX 3000 series cards, the RTX 4090 FE has a unique design with a fan on either side of the card, allowing air to pass through the heatsink, then get blown out towards the top of your PC. Not only is this more efficient at cooling, but it helps keep other components like your RAM and CPU at a more reasonable temperature.
But the meaty nature of this graphics card means that it kind of has to have a very imposing design. This thing measures 11.9 inches long and 5.3 inches thick, so you’re going to want to make sure that it will physically fit in your PC. Then, there’s the weight of the thing. I don’t have an accurate scale lying around, but it easily weighs 8 or 9 pounds, which can put a lot of pressure on a PCI-E slot if your motherboard isn’t reinforced. I haven’t noticed any sag in the card in my machine, but your mileage may vary.
Luckily, the giant design and huge power requirements result in some incredible performance, both in gaming and in creative workloads like Blender. On average, the Nvidia GeForce RTX 4090 is around 49% faster than the RTX 3090. And, if I take out the tests where the RTX 4090 was bottlenecked by the CPU — at 4K, mind you — it takes a more staggering 61.85% performance uplift.
Where the performance improvement is most staggering is in Blender, where the huge increase in CUDA core count means that the RTX 4090 is the perfect tool for folks that need to use the 3D renderer, with a 96% performance improvement over last generation.
Gaming performance improvements are a bit less staggering, but we’re still looking at around 50-70% faster framerates when the CPU isn’t bottlenecking the operation. Metro Exodus Enhanced Edition sees ray tracing performance, without DLSS, jump to 95 fps from 60. With DLSS enabled in this game, though, things stay relatively the same, with the new card netting 104 fps to the RTX 3090’s 98 fps. This is all because the RTX 4090 is hitting the top-end of what that game allows, and I noticed that GPU utilization sat at around 60% throughout this benchmark.
The biggest improvement comes in Cyberpunk 2077 (watch Edgerunners, by the way), which sees framerates jump from 58 fps to 89 fps with DLSS on the performance setting. But I got a sneak peek at a build with the new DLSS 3.0 with Frame Generation enabled, which saw performance get a huge bump up to 129 fps with the same settings, without a hit to frame rate or responsiveness. This means that you can play one of the most graphics intensive games on the market at an esports-level framerate.
The Nvidia GeForce RTX 4090 is, without a doubt, one of the most powerful graphics cards ever, full stop. It runs any PC game on the market at 4K with every setting cranked up without breaking a sweat. The only thing that’s really going to stop most people from going out and getting it as soon as it hits the market on October 12 is going to be the $1,599 price tag, but I suspect even then plenty will want to get in on this generation of PC gaming.
Nvidia is also releasing the RTX 4080 (and likely the RTX 4070) over the next few months, and I suspect those are going to be much more reasonable options for most—at least if this generation doesn’t trigger another PC hardware shortage like the RTX 3000 series did. Either way, graphics cards keep getting better, and there is no better way to get in on the cutting edge than with the Nvidia GeForce RTX 4090—you’ll just have to pay for the privilege.