The state of GPUs is about to drastically change

Get ready — the list of the best graphics cards is going to look a lot different in the next couple of months. For the first time, Nvidia, AMD, and Intel are set to launch new generations within weeks of each other. Whatever you know about the three major players is about to change. Not only are we getting new generations but there are also shifts in strategy between Nvidia and AMD, tariffs to contend with, and next-gen AI features like FSR 4 in the pipeline.

Over the next few months, everything we currently know about the current slate of GPUs will change — that much I can say for sure. I’m here to not only catch you up to speed on the past 12 months of leaks, rumors, reports, and confirmations, but also distill all of that information to get a better grasp on the GPU market of 2025. A lot is changing — some good and some bad — but one thing is undeniable: We’re standing on the edge of an exciting new era for PC gaming.

The easy one: Nvidia

I have continued using weasel words like “could” and “may” to describe Nvidia’s plans for its RTX 50-series GPUs because the company hasn’t even confirmed they exist. But if there are any GPUs we know the most about, it’s the next generation from the market leader. There is a mountain of leaks, even more rumors, and brands like Corsair confirming the next generation is coming soon, and Nvidia’s CEO Jensen Huang is set to take the stage at CES 2025.

Get your weekly teardown of the tech behind PC gaming

Nvidia usually gives some special address at CES each year. However, it has never launched an entirely new generation of graphics cards at CES. Nvidia usually launches new generations at its annual — and sometimes semiannual — GPU Technology Conference, or GTC. The spring GTC event, which is held annually, usually focuses on Nvidia’s robotics, AI, and enterprise projects, while the occasional fall GTC is reserved for new gaming graphics cards. That said, earlier this year, we saw Nvidia launch new gaming GPUs like the RTX 4080 Super at CES, not a GTC event. GTC has instead pushed more toward AI, perhaps encouraging Nvidia to move its gaming launch to a broader tech event like CES.

That’s speculation on my part, but the fact that Nvidia is hosting the main CES keynote, and that it usually has a much smaller presence at the show, is not speculation. After all, Nvidia has already deployed its Blackwell architecture in the data center, which is the architecture it will use in its next-gen GPUs, and Nvidia has reportedly ceased production on several RTX 40-series models, presumably to make way for next-generation graphics cards. That’s something even Nvidia hinted at during its latest earnings call.

So, no, Nvidia hasn’t confirmed RTX 50-series GPUs are launching in January. But I’d be shocked if we didn’t see them come early next year.

That’s just what I’ve seen about when these cards are coming out, and there’s even more to dive into on the specs front. It looks like Nvidia is leaning into the dynamic we saw with the RTX 4090 and RTX 4080 this generation. The RTX 5080 will apparently have less than half of the cores of the RTX 5090, placing an even wider gap between the two GPUs than what we saw this generation. It makes sense. For as powerful as the RTX 4090 is in games, the reality is that an RTX 4080 can do just about everything on the gaming front. The RTX 4090 is that Titan-class replacement, branching out to professional workloads instead of just focusing on gaming alone, and it looks like the RTX 5090 will carry that forward.

Something that should show up across the lineup is new GDDR7 memory. We’ve known about GDDR7 for over year, and in recent months, we’ve seen broader production of the modules get underway. There are some leaks pointing to Nvidia using next-gen memory, but those really aren’t what’s important. Nvidia has been on the cutting edge of graphics memory for the last several generations. That fact, combined with the rollout of mass production — I can’t imagine Nvidia won’t use GDDR7.

I’m not going to speculate on performance here because, frankly, it’s up in the air. Nvidia is uncontested at the top-end — something that will be even more true in the next generation, at least from what AMD has said — so it’s really a matter of far Nvidia wants to push things (and how much money customers are willing to spend).

GPU pricing is a different story. It’s no secret that Nvidia has aggressively pushed the price of GPUs up in the last two generations, and I don’t see that changing with RTX 50-series options. Not only is there speculation that Nvidia will push prices up but we’re also dealing with potential tariffs on China and Mexico — two major importing countries of PC hardware — that could push prices up across the board.

The wildcard: AMD

AMD is in an interesting spot right now. The company has already confirmed that its RDNA 4 graphics cards — the RX 8000 series — will arrive in “early 2025.” That’s a quote from the company’s CEO Dr. Lisa Su, and it’s a time frame that AMD has reiterated for several months. That doesn’t confirm that the cards will show up at CES 2025 alongside Nvidia, but that’s the time frame I’m keeping an eye on.

The release date is the least interesting part, though. AMD’s position is best summed up by a quote shared by the company’s own Jack Hugh earlier this year. The executive says that AMD doesn’t plan to play “king of the hill” with Nvidia. Rumors have been swirling for more than a year that AMD won’t target the extreme high end with RDNA 4 graphics cards, and Hugh’s quote all but confirm that those rumors had merit.

It makes sense for AMD to make this shift. The company has been chasing Nvidia’s crown for more than a decade. Sure, there’s been plenty of conversation between if AMD or Nvidia is better lower down the stack, particularly with cards like the RX 7800 XT, but at the flagship level? Nvidia has dominated for a long time. Team Green has essentially squashed AMD’s flagship ambitions the past two generations. My guess — and this is just a guess — is that AMD doesn’t see a winning battle there, and it probably lost some money chasing the crown.

Some context is important here to understand why. With RDNA 2 graphics cards, AMD finally had a flagship on-par with Nvidia in the form of the RX 6900 XT. Short of ray tracing power, it could stand toe-to-toe with Nvidia’s RTX 3090 for a much lower price. Then, Nvidia released the RTX 3090 Ti. This generation, Nvidia didn’t even have to release a Ti model of the RTX 4090. It just made the flagship super powerful, super expensive, and relegated AMD’s RX 7900 XTX to a battle with the RTX 4080.

As I’ve written about previously here on ReSpec, the flagship conversation sets the tone for a new generation of graphics cards. These halo products, generally, don’t sell in high numbers, at least compared to more reasonably priced options like the RTX 4060 or RX 7600. What halo products do is establish some wider narrative about a particular generation and where the players stand. Absent a flagship contender, AMD can shift that conversation lower down the stack as it has done in the past with popular GPUs like the RX 5700 XT and RX 580.

AMD can likely make up some ground for those seeking higher performance with FSR. Like in GPUs, AMD has been chasing Nvidia’s DLSS with its own upscaling and frame generation tech. AMD says the next version of FSR will be AI-based, and I suspect that next-gen RDNA 4 GPUs will be packing some sort of hardware to sweeten the AI deal. There’s some discussion that we’ll need to have around AMD’s change in approach to FSR and limiting software features to particular hardware, but that’s a column entry for a different day.

The shoe-in: Intel

I’ve dealt in a lot of rumors and speculation up this point, but now let’s move to something concrete: Intel. The company has dropped a lot of official details about its upcoming Battlemage GPUs in the many years that it’s been in the works. No, the launch date isn’t confirmed, but a smattering of leaks have hit over the past week, suggesting we’ll learn about these graphics cards in a matter of days, not weeks. Some reports even indicate we’ll get an announcement as soon as next week, but Intel hasn’t confirmed that timeline.

Intel released its Battlemage tech earlier this year — the Xe2 architecture that Battlemage GPUs will use launched in Lunar Lake laptop CPUs. Just this past week, the company posted a series of videos detailing the architectures inside of Lunar Lake, where Intel’s Tom Petersen confirmed that “Battlemage is next,” suggesting a reveal is right around the corner. This all joins months of speculation that Intel will launch its GPUs ahead of AMD and Nvidia, likely to carve out a media cycle away from the competition. For as impressive as GPUs like the Arc A770 and A750 were, Intel has such a low market share right now that it doesn’t even register against AMD and Nvidia. It needs some space away from the big players to stand out.

As if that wasn’t enough, ASRock spoiled Intel’s party a bit and listed two graphics cards prematurely on Amazon this past week. Two models for the Arc B580 went up. The listings are no longer live, and there weren’t any pricing details, but a few other key pieces of information were revealed. First, the lower-end model — from ASRock’s Challenger series — sported a single 8-pin power connector, suggesting the GPU won’t draw more than 150W. They also revealed the card will come with 12GB of VRAM and a PCIe 5.0 x8 interface.

These details suggest the B580 is simply one of the models Intel plans on launching, and likely the lower-end model. The A770 had dual 8-pin power connectors and 16GB of VRAM. Meanwhile, the Arc A580 came with only 8GB of VRAM. The PCIe 5.0 x8 interface is interesting. It’s not an issue if you have a PCIe 5.0 motherboard, but it could represent bandwidth issues if used in a PCIe 4.0 motherboard slot. That’s something I’ll need to dig into once the card is here, however.

As for higher-end models, Intel has shown that it’s targeting AMD, not Nvidia, to swipe away some market share. Aggressively focused on value, it looks like the flagship card from the range will target the performance of the RTX 4070. It’s hard to say that for sure, though. I’ve seen dozens of leaks over the past six months concerning Battlemage GPUs, all with wildly different performance. That suggests we’ve seen a lot of early engineering tests, which aren’t representative of final performance.

Still, there’s a good chance Intel isn’t holding out for flagship performance, or even performance that’s remotely close to what Nvidia could achieve with the RTX 5090. The strategy for Intel, it seems, is to undercut last-gen options while AMD and Nvidia are focused higher up the stack. That’s a strategy Intel put to good use with its first generation of desktop GPUs.

A trio of contenders

Between leaks, rumors, and official confirmations, one thing is clear — PC gamers are about to get a ton of new hardware to choose from. AMD and Nvidia usually launch within a similar time frame, but now Intel is in the mix. Combined with AMD’s shifting strategy, tariffs that could push prices upward, and Intel’s ambitions in the desktop GPU market, it’s an exciting time in the world of PC gaming — for better and worse.

Looking out toward the next generation, I suspect we’ll see a rejection of insanely expensive flagships like the RTX 4090. There was already plenty of pushback against Nvidia’s pricing strategy in the previous generation, so much so that we got aggressive price cuts with the RTX 4080 Super as well as a rebranding with the RTX 4070 Ti at a lower price. If we see aggressive tariffs, that’ll push the prices of these high-end GPUs even further, and push them out of budget for most PC gamers.

The pricing situation is only compounded by Intel and AMD’s midrange strategies for the upcoming generation. It’s lonely at the top, and it looks like AMD, especially, is abandoning Nvidia at the top in the upcoming generation. Most of the conversation concerning value already happens lower down the stack, but I suspect there will be an even greater focus on midrange GPUs this generation given the competition from Intel and AMD.

On the other hand, there’s some good opportunity for last-gen GPUs in the upcoming generation. This generation was marred by remnants of the GPU shortage, so last-gen options like the RTX 3070 didn’t see big drops in price — prices just dropped back down to list price. With GPUs like the RTX 4060 already seeing sales below list price, hopefully we’ll see some good deals on last-gen options. They could really be the winner of the next generation depending on where pricing lands.