Nvidia's GeForce RTX 5090 is now the fastest graphics card in the world. But when it hits store shelves on January 30, 2025, I won't be upgrading from my RTX 4090. And that's because I don't necessarily feel the need to do so. The RTX 4090 is still more than adequate for all the games that I play, even at 4K resolution. But that's not all; I'm not impressed by the RTX 5090 either. I feel like Nvidia is exaggerating RTX 5090's performance with AI-powered technologies like DLSS 4 and multi-frame generation. So, let me list the reasons why I'm keeping my RTX 4090 this year.

4 The RTX 5090 costs 25% more than its predecessor

$1,999 is just the starting price

The GeForce RTX 4090 launched at $1,599 in late 2022, which, back then, was already pretty expensive for a graphics card. However, the RTX 4090 I bought was a liquid-cooled AIO variant, costing me just over $2,000. I felt like I needed some overclocking headroom down the lane, which a liquid-cooled graphics card provides. Triple-fan RTX 4090 variants from board partners like Asus, MSI, and Gigabyte were also upwards of $1,800 at launch.

The RTX 5090, on the other hand, starts at $1,999 for the Founders Edition variant from Nvidia. So, if you want a beefy 5090 with triple fans or liquid cooling, you're looking at an MSRP between $2,200 and $2,500 — probably even more, depending on the stock availability at launch. And that amount is not worth the raw performance gain over my RTX 4090, which I'll get to in a bit.

3 I would have to upgrade my PSU

The RTX 5090 draws up to 125W more power

I have a Corsair RM1000X in my gaming PC, which is the gold standard for high-end gaming PCs, as it can smoothly handle a total system power draw of 800W. However, with the release of the RTX 5090, I'm starting to feel my PSU just might not be enough. The RTX 4090 has a TDP of 450W, meaning it can draw a maximum of 450W from the wall. When you factor in the other components in my PC, like my CPU, RAM, and motherboard, the Corsair RM1000X seems more than adequate for my needs.

Unfortunately, the RTX 5090 draws even more power, with a TDP of 575W. Nvidia recommends at least a 1000W PSU for its RTX 5090 Founders Edition, but if you plan to get a beefier air-cooled or liquid-cooled variant for overclocking, you should get a 1200W PSU just to be safe. Even board partners aren't sure what to recommend. For instance, Asus recommends a 1000W PSU for its Astral GeForce RTX 5090 OC, whereas Palit recommends 1200W for its high-end RTX 5090 GameRock OC.

2 The RTX 4090 will get DLSS 4 support

I'll only miss out on multi-frame generation

DLSS 4 is one of, if not the biggest, features of Nvidia's RTX 50 series graphics cards. However, all RTX graphics cards, from the RTX 20 series to the yesteryear's flagship, the RTX 4090, support DLSS 4. The only feature that's truly exclusive to the RTX 50 series is the DLSS-powered multi-frame generation, which I'll get to in a bit.

The DLSS 4 feature that's not exclusive is Nvidia's Super Resolution and Ray Reconstruction features. DLSS 4 uses an all-new transformer model instead of convolutional neural networks that we're familiar with in DLSS 3.5. Nvidia claims you can expect better detail, less ghosting, minimal artifacts, and higher temporal stability in supported games.

Digital Foundry published a side-by-side comparison of DLSS 3.5 and DLSS 4 on YouTube, and the improvements are quickly noticeable. I'm just glad I don't have to spend money on a newer graphics card just for better visual quality when I have DLSS enabled on graphically demanding AAA games like Cyberpunk 2077 and Black Myth: Wukong.

1 I rarely use DLSS frame generation

Increased latency makes the game feel less responsive

DLSS frame generation, first introduced alongside RTX 40 series GPUs, is a technique that uses AI to insert a "fake" frame by analyzing the last rendered frame and predicting the next one. This essentially improves the final frame rate without needing additional GPU resources. With DLSS 4's multi-frame generation capability, RTX 50-series graphics cards can insert up to three additional frames between actual rendered ones. This will exponentially boost the frame rate in over 75 games that support DLSS frame generation.

One of the biggest downsides of DLSS frame generation is increased latency. When you use DLSS frame generation in Cyberpunk 2077 to make the game run at over 100 FPS, it still doesn't feel as responsive as a game running natively at 100 FPS with DLSS disabled. That's because of the increased latency that frame generation introduces—it takes time for AI to predict and insert additional frames.

With DLSS 4's multi-frame generation, you should technically be able to run Cyberpunk 2077 at over 200 FPS, but what's the point of playing a story-based single-player game at these absurd frame rates? 100 FPS is more than adequate for these games. Where 200+ FPS would really shine is in multiplayer games like Call of Duty: Warzone and Fortnite. However, in competitive titles where every millisecond makes a difference between victory and defeat, the latency hit due to frame generation makes me want to avoid it altogether.

The RTX 4090 is still incredibly powerful

The RTX 5090 indeed offers big performance gains, but the RTX 4090 is still more than capable of 4K gaming. For competitive games that I mostly play, like Valorant, Fortnite, and Call of Duty: Warzone, my graphics card is still overkill. Yes, I do play single-player games occasionally as they come out, but rarely do I come across one that I'm actually interested in. Unless GTA VI or Crysis 4 comes out in the next year or two, I don't feel the need to upgrade from my 4090, especially since my card will get DLSS 4 via a driver update. If you're like me and you don't care much about multi-frame generation, you'll be more than happy with your RTX 4090, too.