6 thumb rules PC newbies should know about the GPU market in 2025

If you’re new to PCs or haven’t been following the industry for the last few years, the GPU landscape of 2025 can be pretty confusing. $700 doesn’t get you a high-end GPU anymore, native performance is somehow less important than upscaling and frame generation, and you can’t find anything at the advertised price. Even the nomenclature you might be used to doesn’t mean what you might think it should. Oh, and Intel and AMD are big players in the GPU wars now.

If you’re building a new PC after years, and tossing up the various GPU contenders on the market right now, it’s important to remember a few things before parting with your money. We might have more competition than ever, but the market is also in worse shape than ever. As they say, “it was the best of times, it was the worst of times”.

6

Always compare native performance

Frame generation? More like hype generation

If you’re using a GTX 10 series or even an RTX 30 series GPU, you’ve probably never come across the term “frame generation”, especially if you aren’t clued into the recent developments in this space. With the RTX 40 series in 2022, Nvidia introduced the technology to boost framerates by inserting an AI-generated frame for every real frame generated in-engine. While this made the FPS number go higher, the added latency meant frame generation wasn’t a silver bullet.

The perceived experience of playing a game at 100 FPS generated natively (or with the help of upscaling) differs vastly from one with 60 FPS boosted to 100 FPS with AI-powered frames. However, during the recent RTX 50 series launch, Nvidia conveniently omitted any such caveats when boasting sky-high FPS numbers, thanks to its Multi Frame Generation tech, which promised up to 3 “fake” frames for every real one. AMD and Intel adopted frame generation as well on FSR and XeSS, respectively.

That is why it’s essential to compare apples to apples when upgrading to a newer GPU. You should ideally compare the raw performance of your existing card with, say, an RTX 50 or RX 9000 series GPU — no DLSS or FSR enabled. This tells you the most transparent picture to help your purchase decision. Fortunately, many tech publications make these benchmarks available after every GPU launch.

Related

4 reasons why Nvidia’s DLSS is ruining PC gaming

DLSS works like magic, but it’s actually making your games worse

5

VRAM is more important than ever

If only Nvidia agreed

The amount of VRAM on your GPU might not have been a huge concern back in the day, but it is now. Gamers were content buying the GTX 1060 3GB instead of the 6GB variant back in 2016, but today, everyone should pay close attention to how much VRAM manufacturers are offering on the latest GPUs, especially the cards intended for 4K gaming. Nvidia became notorious for skimping on VRAM back in the days of the RTX 30 series, and has been doing the same in 2025.

Without sufficient VRAM, your gaming experience can run into stutters, incomplete textures, crashes, and even some titles refusing to launch. AMD and Intel have fared much better in this regard, with the former providing 16GB of VRAM on both the RX 9070 and RX 9070 XT. Things are hilariously bad with the RTX 5070 and its 12GB VRAM, as it struggles even in today’s titles. Imagine what will happen down the line, as textures increase in size and baked-in ray tracing demands even larger framebuffers.

Nvidia’s RTX 3070 faced the same issue, thanks to its measly 8GB VRAM, and even today, except for the RTX 5090, Nvidia’s VRAM allocation on all its latest GPUs is less than ideal. While VRAM is far from the only consideration when buying a new GPU, it ensures your powerful graphics card can fully stretch its wings, and last longer.

ASRock Radeon RX 9070 Steel Legend

The Radeon RX 9070 is AMD’s most affordable 4K gaming GPU based on the RDNA 4 architecture, offering advanced ray tracing and FSR capabilities, and a decent 16GB of VRAM.

Related

3 reasons why it’s time for NVIDIA to start ponying up more VRAM on their mid-range GPUs

NVIDIA has been skimping on VRAM for far too long.

4

GPU classes no longer signify anything

Shrinkflation in full force

You’ll be forgiven for assuming that an RTX 5080 will provide the same value as that of, say, an RTX 3080 or RTX 2080. An 80-class or 70-class GPU no longer signifies the kind of performance it used to in the past. Nvidia’s RTX 5080, in all respects, should actually be called an RTX 5070, based on its CUDA cores and memory bandwidth via-a-vis the RTX 5090. The degree to which lower-tier GPUs are being cut down relative to the flagship has increased by a lot.

You’re paying much more for a GPU that offers lesser value than your existing GPU, despite both of them belonging to the same tier or class. You can see this happening on the AMD side as well when comparing the RX 6800 XT and RX 7800 XT to the RX 6900 XT and RX 7900 XT, respectively. The percentage of cores on the lower-tier variant went down from 90% to around 70% of the higher-tier card when moving from the RX 6000 series to the RX 7000 series.

And before you think that the architectural improvements made up for the core deficit, let me inform you that the raw performance was in line with the core count reduction. This was seen on both Nvidia and AMD GPUs, and has been a worrying trend for the industry as a whole. So, once again, before you buy a GPU based solely on its name, check performance benchmarks from multiple outlets, and decide based on actual data.

Related

5 reasons my RTX 3080 to RTX 5080 upgrade dream is ruined

I hoped the RTX 5080 would replace my RTX 3080, but the dream is dead.

3

MSRP is meaningless now

Only a “suggestion”

Much of the discussion about AMD beating Nvidia on price to performance falls flat when you can’t buy any of these GPUs at MSRP (or even around it). These “suggested” prices have stopped being a thing for a while now, with GPUs marked up so much that an RX 9070 or RTX 5070 will cost you the same as an RTX 4080 Super did at launch. The reasons that we hear for this trend are limited supply, huge demand from both gamers and AI businesses, and even the latest tariffs.

Whether retailers have become scalpers themselves is up for debate, but the result is that gamers can no longer treat MSRP as any sort of metric when making buying decisions. Things like cost-per-frame don’t mean much of anything when you can’t find anything at the advertised prices. Whether prices will return to some sense of normalcy isn’t clear, and the days of overpriced GPUs might just become the new normal.

Related

3 reasons why right now is the worst time to upgrade your GPU

Waiting just a little while longer will pay dividends for your future GPU upgrade

2

Nvidia still leads in ray tracing

But AMD has closed the gap

Despite AMD closing the gap with Nvidia in ray tracing performance, Team Green still stands on top as far as the most demanding RT titles are concerned. Games like Black Myth: Wukong, Alan Wake 2, and Indiana Jones and the Great Circle are still not playable on the RX 9070 XT with high settings enabled, let alone the RX 9070. Although AMD has shown big strides compared to its RX 7000 GPUs, it might take another generation for AMD to catch up with Nvidia.

If you really want to enjoy the latest and greatest ray-traced titles at high settings, Nvidia seems to be the only option. Granted, those titles can still be counted on one hand, but for some gamers, this can be a deal-breaker when considering Team Red’s offerings. On the other hand, if you don’t care about ray tracing, the RX 9000 GPUs are fantastic options (if you can find one at or around MSRP).

Related

5 things to note when swapping GPUs from NVIDIA to AMD

There are important factors to consider when swapping to the Red Team.

1

AMD & Intel are true alternatives to Nvidia

The market is at a crossroads

Despite AMD lagging in ray tracing performance, it’s finally giving Nvidia some tough competition in the mid-range segment. We all wanted AMD to strike the right balance between performance and price, and they seem to have done it with the RX 9000 series. After a long time, AMD has delivered GPUs that not only seriously undercut Nvidia offerings, but also compete head-to-head in rasterization performance and software features.

AMD implemented dedicated ray tracing hardware on their RDNA 4 GPUs, and drastically improved FSR 4 when compared to FSR 3. We’ve gotten to the point where AMD has comfortably established itself as the best mid-range option. And if you’re looking at a budget GPU in 2025, Intel surprised everyone with its fantastic Arc B580, the value champion when it comes to 1440p gaming. It is decent at ray tracing, features much-improved XeSS performance, and is priced at $250 (street prices are around $350, though).

For someone out of the loop for around 5 years, the current Intel-AMD-Nvidia standings in the GPU market will seem bizarre. It’s certainly good for the consumer, as more competition benefits everyone. Now if we could just get prices back to normal, it would be the best time to be a PC gamer.

Related

6 reasons to pick Intel or AMD for your next GPU instead of Nvidia’s RTX 50 series

Nvidia’s RTX 50 series might be hot right now, but there are enough reasons to consider an AMD and Intel GPU

The best of times, the worst of times

In many ways, the current GPU market is the best and worst it has ever been — simultaneously. We have more competition than ever, consumers are willing to switch loyalties, and things look promising as far as some upcoming GPUs are concerned. At the same time, paper launches are the new normal, MSRPs don’t mean anything, and generational improvements have begun to stagnate. Whether we’re going through a rough patch or entering a new status quo, only time will tell.