What’s wrong with Nvidia’s new Geforce RTX 50 series

If you look at line number 8 in the table above, you will see some outrageous consumption figures. And the power of the RTX 5090 is a cosmic 575 W. Home blenders need this kind of power to crush ice and make frappes. And here, along the way, you can think about how much you will have to pay for electricity.

For several years now I have been paying attention to the fact that Nvidia is increasing productivity by increasing consumption. It is clear that the new video cards have as many as 92 billion transistors, but still! Last year there was a lot of news about Microsoft, Amazon and other high-tech companies building nuclear power plants to power their servers. There is a suspicion that Nvidia has outplayed Fallout and is sure that soon everyone will have a personal mini-reactor.

By the way, I don’t know about you, but I can’t even wrap my head around that number of transistors. After all, for example, if you take 1 cubic meter of sand (the approximate weight of dry sand of this volume is 1500 kg), then it will contain approximately 2 billion grains of sand. And there are 92 billion transistors on the video card.

Laptops will receive their own version of video cards, which will be available from March. Nvidia boasts that Max-Q technology has become more efficient and now laptops last 40% longer and their performance has doubled. Unfortunately, Nvidia did not indicate the actual number of hours of battery life. Or they could say that now a laptop with a running game lasts 2 hours, but with Max-Q technology it will last 3 hours. But they didn’t say, because the discrete video card still only works fully when the laptop is connected to a power outlet.

Nvidia marketers do the same sneaky trick every time. First, Jensen Huang introduces the new graphics cards and explains how powerful they are. And then he says that there will also be versions for laptops. And the viewer gets the illusory feeling that the new generation of video cards for laptops is just as powerful and innovative. Why, they even have almost identical names. The difference is in the word laptop in the name.

Although, to be fair, video cards for laptops should not have names similar to desktop versions. The performance of laptop video cards is approximately 2 times lower. For example, the mobile RTX 5090 has 2 times fewer CUDA cores, and the power consumption is almost 4 times lower. Although, of course, it would be interesting to look at a laptop with a 575 W card (*sarcasm).

By the way, Nvidia did not indicate performance in the comparison table, so that, God forbid, users would not start comparing.

However, most users will obviously have laptops with RTX 5070 and RTX 5070 Ti mobile cards. Laptop manufacturers are already showing versions with a mobile RTX 5090 for $4,000. And this is such a protective price.

Interestingly, Nvidia claims that the desktop RTX 5070 will be 2 times faster than the 4070. But the company does not make similar statements about mobile video cards. If I read the press release correctly, only the RTX 5080 mobile video card can boast a twofold increase in performance. This may be due to the fact that it has 16 GB of video memory (the RTX 5070 and 5070 Ti for laptops have the same amount of video memory).

Given the abundance of information about desktop video cards in general, very little has been said about versions for laptops. In fact, Nvidia devoted several lines of text to mobile graphics in a press release, which mainly list technical characteristics (which are already in the table).

Correct me if I’m wrong, but it seems to me that the current increase in the power of video cards is due to two things: artificial intelligence (or more precisely, the desire to give the video card the ability to work more confidently with generative neural networks) and the ill-fated ray tracing, that is, ray tracing.

Ray tracing is definitely a cool thing, it makes the gameplay more realistic. It’s cool when a house or a character passing by is reflected in the water. And the glare on the cars and weapons is beautiful. But as a gamer (mediocre and occasional), I can say that you practically don’t pay attention to all these reflections. Usually you notice them not during the game, but on the contrary, when you are distracted. For example, a friend called, you are talking to him, and your gaze wanders across the screen.

At the same time, ray tracing, in my opinion, is still not ready for public use. Here you can remember Cyberpunk 2077, with ray tracing of which, if you did not use DLSS, even the RTX 4090 could not cope.

I can understand why ray tracing is useful when working with graphics (everyone loves beautiful Pixar cartoons), but in games it’s more like a suitcase without a handle: it’s hard to carry and it’s a shame to throw it away. I assume that gamers would be satisfied with some intermediate version of ray tracing, which creates the overall illusion of reflections.

But increasing the power so that the user can use the video card for generative networks is an open question. On the one hand, hardware manufacturers predict that soon every household will have a personal generative neural network, and are even releasing devices labeled AI-ready.

On the other hand, so far neural networks are largely the domain of professionals who use AI for work. Accordingly, separate lines of devices could be released for such users. Nvidia, by the way, has such devices. For example, the Quadro series or RTX 6000. And when it comes to video cards for professional needs, you don’t have to restrain yourself. Even if the power there is even a whole kilowatt, no one will be upset if the card helps you earn money.

And finally, I can’t help but kick RTX graphics for laptops for the fact that the junior RTX 5070 video card received only 8 GB of video memory. And this is not just close, in 2025 this is not enough to play comfortably. Recently the Indiana Jones game came out, and 8 GB of memory is not enough. Anyway, I finally started playing God of War Ragnarok, and on ultra graphics settings, 8 GB is also not enough. You need 12 GB minimum, or better yet 16 GB, like the RTX 5080.

Buy a cast iron bridge cheaper

New video cards are expensive. Especially the flagship desktop RTX 5090 for $2 thousand. However, to be fair, I will say that when I saw the prices, I breathed a sigh of relief. Having read all sorts of leaks, I expected that they could ask for from 2.5 to 3 thousand dollars. Although even 2 thousand (220 thousand rubles) is already indecently expensive.

Laptop prices also cause nervous eye twitching. For example, ASUS introduced the ROG Zephyrus G16 with an RTX 5090 video card for laptops, the price of which will be 4 thousand dollars, that is, something like half a million (maybe the whole country should temporarily strengthen the ruble to 50 rubles per dollar, so that everyone can purchase equipment and others?).

Basic versions with RTX 5070 cards (those with 8GB of video memory and which you don’t need to buy) will start at $2,000. By the way, ASUS continues to hold on to Intel for some reason. Instead of AMD, the processor will be Intel Core Ultra 9 285H. And ASUS still doesn’t want to advertise battery life. However, there is progress in this area. In previous models, this issue was generally avoided, but now they wrote vaguely All day power, that is, “the charge will last for the whole day.”

Conclusion

As a mobile user, I am once again disappointed that I was taken for a ride. Mobile video cards are still just a faded shadow of their desktop relatives. At the same time, their prices are steep.

Desktop versions are good, although you need to be mentally prepared for the fact that your electricity bill will increase by one and a half times. Although, probably, if you buy video cards for 220 thousand rubles, then an additional 2–4 thousand for light will not make a difference. The downside is that the new cards themselves are large: 30 by 14 cm and a thickness of 5 cm.

Nvidia’s solutions look especially outdated compared to arm laptops from Apple, which are capable of delivering similar performance (with a stretch), but do not require a power outlet (although the laptop, of course, will last only 1.5–2 hours under maximum load).

Honestly, I really want Apple to bet on gaming in 2025 and punch Nvidia in the nose. Perhaps this would make the company move. In the meantime, it might be worth looking into AMD video cards.