We're in a little lull for tech releases over the holiday period, but that just means that it's nearly CES and the firehose of new releases it brings. In 2024, CES brought us AI-powered computers, super-speedy OLED monitors, more PC gaming handhelds, and a ton of other computing-adjacent hardware. But it didn't bring new graphics cards, as the major GPU manufacturers weren't quite ready for their next generation to be revealed.

That changes at CES 2025. Intel is getting a head start by releasing Battlemage early, and both of the other GPU makers are scheduled to release new models in January. That includes Nvidia's RTX 5000 series, but I'm finding it difficult to get excited about this release for various reasons. It's not just that I already have a capable GPU because that never stopped me from upgrading before now.

It's more that the lack of competition almost guarantees Nvidia the win here, and that means they can charge whatever they want. I want to see genuine competition in the PC gaming space again, whether that's between graphics card manufacturers or CPU makers, and that's just not likely this year.

6 Price

We know it'll be expensive but by how much?

For years now, Nvidia has been untouched at the enthusiast level, with the competition either falling short or deciding not to compete. That's led to a situation where Team Green can charge whatever it wants for the flagship models, and it has been doing just that. The GeForce RTX 4090 started with a $1,599 MSRP for the Founder's Edition cards, but AIB models started around the $2,000 mark, and supply chain issues due to AI, cryptomining, and the pandemic meant they were regularly higher than MSRP.

So, we've got a situation where the second-tier GPU, the RTX 4080 Super, seems like a relative bargain at $1,000 to $1,150. This model was only released in early 2024, and the card it replaced was $200 more expensive. With AMD's Radeon 8000 series rumored to target the midrange, Nvidia is completely unopposed at the high and enthusiast levels to charge whatever they want.

I can't see a situation where the GeForce RTX 5090 isn't $2,000 or more, not with the rumored 32GB of GDDR7 memory. It wasn't that long ago that Nvidia released the GeForce RTX 3090 Ti at $2,000 because of its 24GB of VRAM, and that line of reasoning will be in full effect in January. Add to the impending promised tariffs by the incoming US administration, and the stage is set for expensive graphics cards from everyone. I'd love to be wrong here, but that price tag is going to mean the GeForce RTX 5090 won't be in the hands of many gamers. Instead, it'll be in AI training labs where the budget doesn't matter.

5 Flagship gets all the good specs

Only the RTX 5090 gets plenty of VRAM, bandwidth, and cores

The Nvidia GeForce RTX 5000 series is likely to be powerful, as the company historically manages to get a decent uplift every generation. But how much of a power boost will depend on which model you buy, and the flagship RTX 5090 is going to get the best of everything. That will leave the rest of the stack to fight over scraps, but even the RTX 5080 is likely to be pretty expensive.

Let's talk numbers. The RTX 5090 is likely to have 21,760 CUDA cores, and 170 RT cores, which is slightly cut down from the full Blackwell GPU die that enterprise graphics cards will be able to access. It's also rumored to have 32 GB of GDDR7 VRAM, with a 512-bit memory interface leading to a rough bandwidth of 1.7 TB/s.

These are impressive numbers, indeed, but the RTX 5080 is supposed to have roughly half that power, with 10,752 CUDA cores, 84 RT cores, 16 GB of GDDR7 VRAM, and a total memory bandwidth of 1,024 GB/s. This is still impressive, as it's 6 GB/s more than the RTX 4090, but the lower core count will hurt performance overall.

The final specifications will determine the performance, but we can speculate based on rumored specifications. A forum user at TechPowerUp did some napkin math, and the RTX 5080 might only be 30% more powerful than the RTX 4080 it's replacing, while the RTX 5090 might be 70% more powerful than the RTX 4090. Basically, every tier except the flagship jumps up a performance level while the flagship gets a significant uplift.

4 Even more power-hungry

I don't want my GPU to raise my power bill

My gaming and working PC currently has an Nvidia GeForce RTX 4080 Super inside, with a 320W TDP. I rarely see that power draw when working or gaming, but benchmark runs and heavier workloads get near that figure. That's about the most power and heat I want to deal with from my PC, especially as I plan to slimline to an SFF case in the future.

The Nvidia RTX 5000 series is going to substantially boost the power requirements, if the early leaks are correct. The RTX 5090 could need as much as 600W, which is 150W more than the RTX 4090 needed. The successor to my card, the RTX 5080, is rumored to need 400W, and all the way down the line, each RTX 5000 series GPU wants more power than its successor. That might mean needing a new PSU for some of you, and it's entirely possible the 16-pin 12VHPWR cable will be used on every card.

That will mean bigger heatsinks to cool them down, which will take up more space in your case. The jet of warm air from my current GPU is already hot enough, but it's going to increase in the next generation. Nice enough while it's winter, but not fun for summer or those who live in warmer climates.

3 Limited VRAM on 5080

Games want more VRAM for textures than ever before

AAA gaming loves lots of VRAM to stuff high-resolution textures into the graphics memory for fast retrieval. With Ada Lovelace, the RTX 4080 nearly ended up with a 12GB option, which would have been hamstrung for performance. That card was still released as the RTX 4070 Ti, which had okay performance but a weird price and positioning that made the RTX 4070 and RTX 4080 both better value.

For Blackwell, only the RTX 5090 gets a good boost to VRAM amounts, going up to 32GB from 24GB on the current flagship. The RTX 5080 gets 16GB, as does the RTX 5070 Ti, with the RTX 5070 getting 12GB. There's even a rumored RTX 5060 Ti with 16GB of VRAM, so it makes no sense that the RTX 5080 would also be similarly equipped.

This is all part of Nvidia's usual playbook. Create a super powerful chip that goes into enterprise graphics cards that consumers never see, then sell cut-down versions of that monolithic core to gamers and other home users. It's part smart business, part by-product of the large monolithic die designs that they've used up til now, which are hard to get large yields of in perfect condition. AMD moved to chiplets for just this reason, and Nvidia is getting close to needing to go the same way with its designs.

2 Low-end model still on old tech

The 5050 is stuck on GDDR6

Not every GeForce RTX 5000 series graphics card is getting all the new technologies. The RTX 5050 is rumored to still have GDDR6 memory, the only GPU in the range not to get an upgrade to GDDR7. That might mean it's a rebrand of an existing GPU, but we'll know that when it gets unveiled. Even if it's the newer graphics core, it's disappointing to see that the card aimed at budget gamers isn't getting all the new tech.

1 Stock will be low at launch

When was the last time it wasn't?

With every new generation of Nvidia graphics cards, there are stock issues at launch. The situation is resolved fairly quickly in some years, but with impending tariffs, trade wars, and other considerations, 2025 will not be one of those years. Expect the flagship GeForce RTX 5090 to be in short supply anyway, as the better binned GPU cores will be going to enterprise customers in Blackwell AI GPUs, which suits Nvidia because they can charge what they want and still say the launch was a success as it sold out.

For the first time in years, I'm not hyped up for the next generation of Nvidia graphics cards

As a PC gaming enthusiast, I should be excited that a new generation of Nvidia GPUs is coming out. I should be, but I'm not. Even if the rumored performance is there, I can't see this being a generation that brings value to consumers, between generational price increases and the looming specter of high tariffs making them even more expensive for US-based gamers. That might make it so we feel the sting the rest of the world feels with GPU pricing, but I'm sure those prices will be increased accordingly as well, making it an expensive time to be a PC gamer wanting to upgrade.