(Image credit: Future)

It's 1.55 pm, March 5, 2025. I'm pacing the confines of the home office/wooden shack at the bottom of my garden I've been inhabiting fairly solidly for the past two months as I benchmark two whole new generations of GPUs from both Nvidia and AMD, and it's five minutes before the review embargo drops for AMD's Radeon RX 9070 XT graphics card. I'm on the phone with someone I've known at AMD for years. Again. And at least they're apologising for the fact the company has been unable to give me any kind of reliable pricing about the different versions of its GPU I've been testing.

There have been no reference cards for AMD's new RDNA 4 architecture, and so I've got a selection of different options—all with different price tags—in front of me to base my effectively definitive review of what is actually the best AMD GPU in many a long year. And potentially none of them are at the ultra-aggressive $599 price point that should make this impressive card an outright winner.

And then eight minutes after the embargo lifts, XFX messages to say, "change of plan, sorry" and that now the premium XFX Mercury RX 9070 XT is going to be a reference-priced card on launch the following day. Though actually, that never comes to pass in any way that's visible online, and the card actually ends up retailing for over $900 for the next few months instead of the $599 I'm promised.

What a shit way to spend a birthday. A birthday where I'm not even technically working.

Keep up to date with the most important stories and the best deals, as picked by the PC Gamer team.

I'm starting with AMD here only because the RX 9000-series launch was the culmination of a frustrating period of late nights and long days of graphics card benchmarking, but the early-year Nvidia launches of the RTX Blackwell generation were no better.

Sure, we had Founders Editions for most of them, giving Nvidia at least the opportunity to pay a little more than lip service to its suggested retail prices. But as reviewers, we still had to struggle with late and sometimes flaky drivers and ultra-short review times; then as consumers, we had to deal with black screen issues, constant worries about that damned cable, low supply, and absolutely brutal price gouging from both manufacturers and retailers combined.

(Image credit: Nvidia)

Not to mention concerns that much of the vaunted generation-on-generation performance increases were largely down to the introduction of Multi-Frame Generation (MFG) in a few key games and not widespread actual rendering performance increases from the silicon itself.

The relative chip sizes got smaller, arguably less complex within a GPU tier (RTX 5070 vs RTX 4070, I'm looking at you) and meant you were left with the feeling that Nvidia was huffing its own excretions when it came to the notion that graphics cards should be judged purely on performance over silicon.

AMD should have been able to completely run rampant, what with its mix of lower price and practically equivalent performance compared with the competing RTX 5070 Ti—itself arguably the best RTX 50-series GPU. But AMD had a weird start to the year, where the RX 9070 XT and RX 9070 cards were announced to the press prior to the new year but were notably absent from its big CES press conference.

Image

1

of

3

(Image credit: Future)

I was quickly ushered behind the scenes out in Vegas after the press conference, where the notably downbeat David McAfee and Frank Azor tried to talk up the RDNA 4 architecture despite the red team presumably having been spooked by Nvidia's announcing its RTX Blackwell cards later that day. And, if you remembner, that was an announcement that presented the $549 RTX 5070 as set to deliver RTX 4090 gaming performance.

As the necessary inclusion of MFG in those performance calculations became clear, however, AMD solidified its position. Though the demand and lack of supply inevitably made the RX 9000-series launch a blood bath. With no way for AMD, and to a slightly lesser extent Nvidia to control prices, and the knowledge that voracious GPU buyers would pay through the nose for cards—as they did during the pandemic years—manufacturers and retailers both sniffed an opportunity to gouge gamers on pricing, and so the bloodletting began.

(Image credit: AMD)

It was frankly offensive the way that each was determined to extract their own pound of flesh from the supply/demand iniquity. Though it was obviously understandable, as both had watched scalpers make bank during the previous COVID and mining-related GPU supply crises, and now wanted to make sure they could fill their respective coffers from gamers' desire for new silicon.

We can talk about dodgy drivers, misleading performance promises, and GPU tiering we don't agree with, but the pricing side was arguably the biggest issue around this calamitous generation of graphics cards, and that, at least, wasn't either AMD or Nvidia's fault. Just blame capitalism, I guess.

The RX 9070 XT and the RTX 5070 Ti, nominally $599 and $749 cards respectively, both ended up being priced at $900+ for at least the first few months of their lives, and only towards the end of the year have prices and stock begun to stabilise and settle down to around the original launch prices of the cards.

Image

1

of

2

(Image credit: Future)

It was the worst collective generational GPU launch I can remember in my 20 years as a PC gaming tech journalist, and I was there for the nightmarish AMD R600 launch and carried a super-heated GTX 480 all the way back from Paris. I, for one, am very glad it's passed. At least as 2025 fades into memory, we can sleep safe in the knowledge that GPU prices have come down. However temporarily.

Welcome to another pricing crisis as the RAMpocalypse surges and suggests GDDR pricing for graphics cards is set to rise alongside rumours that Nvidia has stopped bundling VRAM with its chips to card manufacturers. After all, anything with memory is going to be getting ever more costly for the foreseeable future, thanks to damned 'AI' demand.

So yeah, sleep safe, and don't have nightmares.

Dave has been gaming since the days of Zaxxon and Lady Bug on the Colecovision, and code books for the Commodore Vic 20 (Death Race 2000!). He built his first gaming PC at the tender age of 16, and finally finished bug-fixing the Cyrix-based system around a year later. When he dropped it out of the window. He first started writing for Official PlayStation Magazine and Xbox World many decades ago, then moved onto PC Format full-time, then PC Gamer, TechRadar, and T3 among others. Now he's back, writing about the nightmarish graphics card market, CPUs with more cores than sense, gaming laptops hotter than the sun, and SSDs more capacious than a Cybertruck.

You must confirm your public display name before commenting

Please logout and then login again, you will then be prompted to enter your display name.