Nvidia RTX 5090 mobile GPU hands-on: more efficient and a little faster

Nvidia’s 50-series desktop GPUs have been off to a slightly rocky start, with quality control issues, underwhelming performance gains over last generation, and the occasional melting cable. Its new laptop GPUs, led by the flagship RTX 5090, may be on a similarly modest trajectory, but with one added benefit for laptop gamers: efficiency.

I’ve been testing the RTX 5090 laptop GPU in Razer’s new Blade 16 laptop, which Nvidia is using as a showcase for its top-tier mobile card. My time with it has been more limited than I’d hoped, since my first review unit exhibited some strange graphical anomalies and was prone to blue-screen crashes during basic productivity tasks. Nvidia sent me a replacement, but I’ve had less than two days with it as of press time. From what I’ve seen so far, it is indeed a little faster than the mobile 4090, but like the desktop 50-series cards, its biggest improvements are in DLSS and Multi Frame Generation, rather than raw performance.

The new mobile RTX 5090 has 24GB of GDDR7 VRAM instead of the 16GB of GDDR6 found on the 4090. It’s also got more powerful Tensor cores, slightly more CUDA cores, and an extra hardware video encoder for better livestreaming and video exports, but the same 175W TDP.

The 5090’s specs translate to small gains over the 4090 in synthetic benchmarks, such as a 13 percent higher Geekbench GPU score and a 14 percent higher score in 3DMark’s standard Time Spy test at 2560 x 1440 resolution.

To test the 5090’s uplift in actual games, I ran an array of benchmarks using Cyberpunk 2077 and Black Myth: Wukong — at the native 2560 x 1600 of both Blade 16s and at 4K on an Alienware AW3225QF monitor. The biggest difference I saw was in Cyberpunk 2077 at 2.5K resolution and Ultra settings without DLSS, frame generation, or ray tracing enabled. There, the 5090 was 24 percent faster than the 4090. That sounds great, but once you enable ray tracing and DLSS 4 (something most people would likely do on either of these laptops), it narrows to a delta of just 5 percent. When you enable 2x frame generation, the 5090 is 14 percent faster than the 4090, and going all out with 4x Multi Frame Gen more than doubles what the 4090 can do at 2x frame gen. If you’re playing at 4K, the 5090 still wins, but the gap narrows to within single digits.

Black Myth: Wukong, on the other hand, is a little less forgiving. In this graphically demanding Souls-adjacent RPG, the 5090 is only a few percent faster than the 4090. The highest average the 5090 could muster was 111fps at 2.5K with DLSS and standard frame generation, with the 4090 wasn’t far behind at 104. If Wukong supported Multi Frame Generation, the 5090 could no doubt separate itself further, but that would likely add a couple extra milliseconds of latency. Most people will never notice or care about that, but some players may feel any latency added to their pinpoint dodge-roll timing is a nonstarter.

Like the 50-series desktop cards, the 5090 laptop GPU looks its best in games that support DLSS 4 Multi Frame Generation. Over 100 games now support Nvidia’s highlight feature, and you can also use override settings in the Nvidia app to force it in unsupported games.

Frame generation, which both Nvidia and AMD offer, uses AI models to generate extra frames and insert them between traditionally rendered frames. This gives the appearance of smoother gameplay but doesn’t improve input latency — and actually increases it slightly. AMD and previous Nvidia cards can do 2x frame generation; the 50 series can do up to 4x. There’s an ongoing debate in the world of PC gaming about these “fake frames” — some folks are perfectly content with AI upscaling and generated frames, and others view it as a band-aid solution with too many compromises to image quality and game feel. I think it depends on your personal experience, what kinds of games you like to play and what you’re used to playing them on, and even whether you have an eye (or even a care) for this stuff. It can be great for the right games, and it may even make more sense on a 16-inch laptop screen than a jumbo 4K monitor right in front of your face.

I forced Multi Frame Generation on Assassin’s Creed Shadows, and I was pleasantly surprised with its performance and buttery smoothness on ultra settings. Shadows is maybe a perfect example of a single-player game where frame generation is a nonissue. It’s got dodge rolling, but it’s not the kind of game that really requires frame-perfect timing. It felt A-okay to me, and boy did feudal Japan look lovely on a 2.5K OLED at an average of 167fps.

The obvious games to avoid with frame generation are competitive shooters. While I felt fine playing Marvel Rivals on the 5090 with 4x frame generation reaching nearly 200fps, I was playing pretty casually with a hero that doesn’t require precision accuracy. For something like Valorant or Counter-Strike 2, I’d advise against it. But to be fair, those ultra-competitive games typically have a low performance barrier for entry and practically run on a potato.

One area where the 5090 does seem significantly better so far is power efficiency. In our benchmarks, the new GPU had a lower average wattage, by 20 to 29 percent, than the 4090. That should help improve battery life when playing games away from a power cord.

Lower wattages can also mean lower temperatures, as I was able to play games with the plugged-in laptop on my actual lap and not toast my legs. But this also hinges on an individual laptop’s thermal design and how well it cools — the Blade 16 so far seems to do it well, but it’s also a totally different design than the 2024 model with the 4090 in it. Of course, you’re not going to see the same level of performance when running graphically intensive games on battery power. I’ll have to test this more to see how much of a difference it makes in real-world usage.

The RTX 5090 is Nvidia’s most powerful and expensive laptop GPU, and you can expect to pay $4,000 or more for a laptop that runs it. The Blade 16 review configuration costs $4,499.99. It shows some promise for games that support DLSS 4 and Multi Frame Generation, and maybe for gaming that’s not always tethered to a wall. But if I were shopping right now, I wouldn’t ignore a heavy discount on a laptop with an RTX 4090. (The 4090 desktop supply, on the other hand, has dried up.)

We’ll continue to test the RTX 5090, as well as the rest of the 50-series mobile GPUs as they appear in new gaming laptop models over the next few weeks and months. I’m looking forward to testing the RTX 5080, in particular, and seeing if that’s the sweet spot for laptop gamers.

Photography by Antonio G. Di Benedetto / The Verge