Saying Nvidia's GeForce RTX 4090 GPU is the fastest GeForce card the company's ever released is true, but meaningless -- it would be more notable if it weren't the fastest to date. But it is really, really powerful, and if you're looking to upgrade from an RTX 3080 Ti or lower in search of a kick-ass 4K gaming (or creating) experience, it's the one to beat. Well, at least until we see what AMD's Radeon RX 7000-series GPUs can do in November.
But that's only if you take price out of the equation: At $1,600 (£1,699, AU$2,959), it's out of reach for most folks. And as a triple-slot, 5.4-inch tall, 450-watt card, the Nvidia GeForce RTX 4090 Founders Edition is out of reach for a lot of systems as well.
- Really fast
- Lifts performance of everything, not just games and applications that specifically support the new algorithms
- Relatively quiet
- Big and power-hungry
- Using the four-headed power adapter gets ugly
But I really wouldn't buy one before the discount season starts in mid-November. Yes, the price is high, but it's not gougingly overpriced given its performance; it's just suffering from the business-as-usual, taking-advantage-of-FOMO markup. (Big caveat: I haven't tested the RTX 3090 Ti.) The RTX 3090 launched two years ago at $1,499 and the RTX 3090 Ti in March 2022 for $2,000; prices on the latter have since dropped to $1,550 and below. If you can't or don't want to spend that much money now, don't.
In the absence of performance results for the RTX 4080s and whatever AMD ships, there's no way to make any price-to-performance-based recommendations. Especially since it's not clear if the RTX 4080s will put any more downward pressure on RTX 3080 Ti prices as sellers really need to get rid of stock, or if the 4090's price will drop as buyers balk at the price.
Nvidia GeForce RTX 4090 Founders Edition
|Memory bandwidth (GBps)||1.0|
|Memory clock (GHz)||10.5|
|GPU clock (GHz, base/boost)||2.23/2.52|
|Memory data rate/Interface||21Gbps/384-bit|
|Texture mapping units||512|
|TGP/min PSU (watts)||450W/850W|
|Max thermal (degrees)||194F/90C|
|Connectors||3x DP 1.4a, 1x HDMI 2.1|
|Size||3 slots; 12 x 5.4 in (304 x 137 mm)|
|Ship date||Oct. 11|
Big and bulky
Although much has been made of the RTX 4090's size, the Founders Edition is at least smaller than the RTX 3090 and RTX 3090 Ti -- about an inch shorter. It's just as tall, though (distance from PCI connector side to top), and I wish it were a bit less so. While it fits in my Corsair 4000D Airflow case -- just a midtower, but still pretty wide at 9 inches -- a bit more space would give the quad-hydra power adapter coming off the top of the GPU a more comfortable bend where it hits the glass panel. And you can forget about your system's inner beauty. Time to get some PCIe 5 power cables so I don't have to use the Nvidia adapters.
The card is still big, though: Slightly bigger than an Xbox Series S and definitely fatter than the RTX 3090 Ti FE. Somewhat more mind-boggling, though, is that the Founders Edition only has two fans and it's still a foot long. Third-party cards mostly have three (smaller) fans, so they'll likely be even longer. It's quiet most of the time, too, though when you're pushing it you can hear the fans. I've heard worse, though, and at least they don't whine.
But the FE is heavy, probably because that heft includes a lot of heat-dissipating materials to make up for more fans. While it certainly requires three full slots, even if it didn't, that third slot's screw made me feel a lot more comfortable about relieving some potential stress on the PCI connector.
Nvidia has made a point of saying support for DLSS 3 is limited to the new Ada Lovelace-class GPUs at launch, but post-launch it's likely it will extend to at least some RTX 30-series cards as well. But I don't expect it to deliver nearly as much improvement for the older GPUs simply because Ada has a lot of streamlining, new algorithms and so on that generally lift performance before you even get to DLSS improvements. Similar to how the 30 series made a real jump over the 20 series for many things. (Here's a.)
There's a lot to unpack with respect to the RTX 4090 FE's performance relative to other cards, and I don't have much of the comparative data yet -- I've just switched to a new GPU testbed and haven't completed a lot of my RTX 4090 testing or going back through older cards to retest.
I have tested enough to answer some of my most pressing questions, though. For one, even games that don't yet (or won't) support DLSS 3 get a notable boost in DLSS 2 performance in 4K. With DLSS 3, the CPU tends to run a little hotter and the GPU a little cooler compared to DLSS 2. You can get really good 4K performance without any performance-enhancing operations enabled (i.e. DLSS), and with everything ray-traceable and HDR maxed out at 4K, you can play Cyberpunk 2077 pretty well with DLSS 3 (101 fps). Although there isn't a lot of improvement in minimum latency with DLSS 3 (which isn't surprising), it can significantly lower the maximum latency.
Its ray-tracing performance, both on Port Royal and 3DMark DXR, blew past everything else I've tested by a lot, and the same goes for every test in the SpecViewPerf 2020 pro graphics suite. It seems to run pretty coolly, but I haven't stressed it for an extended period yet.
I'll have my complete set of results ready by the time I've got an RTX 4080 or AMD Radeon RX 7000-series card for review. But I can say that the Nvidia GeForce RTX 4090 Founders Edition is a great card if you've got the room and the power supply for it. However, if you care about the money and aren't desperate, wait a month or two before committing.
Test PC configuration
|Custom PC||Microsoft Windows 11 Pro (22H2); 3.2GHz Intel Core i9-12900K; 32GB DDR5-4800; 2x Corsair MP600 Pro SSD; Corsair HX1200 80 Plus Platinum PSU, MSI MPG Z690 Force Wi-Fi motherboard, Corsair 4000D Airflow midtower case|