Skip to main content
  1. Home
  2. Computing
  3. News

AMD’s new GPU offers RTX 5070 Ti performance for $150 less

Add as a preferred source on Google
Several AMD RX 9000 series graphics cards.
AMD

After a long wait, the future of AMD’s best graphics cards is finally here. The company just announced the RX 9000 series, comprised of two new GPUs: The RX 9070 XT and the RX 9070. Both are set to launch on March 6 at $599 and $549, respectively. If AMD lives up to the expectations performance-wise, these GPUs might be an absolute mainstream hit.

AMD’s focus throughout the presentation was the fact that gamers buy mainstream GPUs more often than behemoths along the lines of the RTX 5090. To that end, AMD cited its own research, saying that 85% of gamers buy GPUs that cost less than $700. At the same time, higher resolutions are slowly becoming mainstream, with more and more people buying 1440p and 4K monitors. The needs of gamers have gone up, now including solid performance in ray tracing, but keeping it affordable is important too.

Recommended Videos

At $599 and $549, both the RX 9070 XT and the non-XT counterpart will fall close to the RTX 5070 in terms of pricing. However, the RX 9070 XT is expected to perform on a similar level to the RTX 5070 Ti, which spells great news for AMD. The RTX 5070 Ti launched at $750, but it’s largely sold out and hardly available at MSRP right now.

The specs for AMD's RX 9070 XT and RX 9070.
AMD

The specs of RDNA 4 have been leaked months ago, and it turns out that leakers got a lot of it right. The flagship RX 9070 XT serves up 64 compute units (CUs), 64 hardware ray tracing (HW RT) accelerators, and 128 AI accelerators. The boost clock nearly hits 3GHz, maxing out at 2.97GHz, which means that overclocked GPUs will definitely go past that threshold. The card comes with 16GB of VRAM across a 256-bit bus, and this memory configuration also applies to the RX 9070 non-XT. The flagship has a total board power (TBP) of 304 watts.

The RX 9070 cuts back on every spec, scaling down to 56 CUs and a maximum clock speed of 2.52GHz. On the other hand, the power consumption is also significantly lower, now at 220 watts.

Both GPUs will use GDDR6 memory. AMD chose to stick to GDDR6, clocked at 20Gbps, which it found to offer the best balance of performance and cost. Nvidia’s RTX 50-series already made the switch to GDDR7, having used GDDR6X in the previous generation, but AMD offsets for the loss of bandwidth by adding plenty of VRAM to its GPUs.

Benchmark for the RX 9070.
Image used with permission by copyright holder

AMD revealed some benchmarks for both GPUs, and they seem pretty great. However, when comparing to Nvidia, it used an RTX 3090 as opposed to something from a current generation. AMD explained that this was because people who bought the RTX 30-series will be looking to upgrade now, but it would’ve been nice to see some scores for that.

Fortunately, AMD revealed that it expects the RX 9070 XT to be on par with Nvidia’s $750 RTX 5070 Ti or the last-gen RTX 4080. AMD promises up to 21% better performance-per-dollar than the RTX 5070 Ti, which is a great deal if the two cards really end up going head-to-head.

Benchmark for the RX 9070 XT.
AMD

Regardless, the RX 9070 is said to offer up to 21% faster performance than the RX 7900 GRE at 4K, which is a sizeable uplift. Meanwhile, the RX 9070 XT will be up to 42% faster at 4K, while also showing solid ray tracing improvements.

The new AMD RDNA 4 architecture and compute unit.
AMD

With the new compute unit (CU) in RDNA 4, AMD was able to boost efficiency, clock speeds, and register allocation. The new graphics cards will come equipped with third-gen ray tracing accelerators, which will offer 2x ray tracing throughput per CU. Historically, AMD trailed behind Nvidia in ray tracing, but it’s certainly making improvements in that regard.

The new GPUs will also come with second-gen AI accelerators, which AMD promises will result in up to 8x faster AI performance when compared to RDNA 3. The die size is at about 350mm2, and the transistor count at around 44 billion. AMD also revealed that the GPU was built on a 4nm process. The RX 9000 series makes the switch to PCIe 5.0, and it also offers DisplayPort 2.1a and HDMI 2.1b support.

The company will not be making its own versions of the cards, also known as MBA (Made By AMD), but it says it’ll work closely with its partners to ensure availability, and more importantly, availability at MSRP.

Alongside the RX 9000 series, AMD also announced FSR 4, which will remain an RDNA 4-exclusive for the time being. AMD has previously said that it would like to open up FSR 4 to other cards, but it’s unclear when or if that might end up happening.

Monica J. White
Monica is a computing writer at Digital Trends, focusing on PC hardware. Since joining the team in 2021, Monica has written…
The US government just hit the brakes on Anthropic’s most powerful AI models
Anthropic disables Fable 5 and Mythos 5 after US government directive
Laptop running Claude Fable

Anthropic’s troubles with the US government do not seem to be easing. The company has now been ordered to suspend access to Fable 5 and Mythos 5 for all foreign nationals, including foreign national Anthropic employees working inside the United States.

Anthropic said it received the directive on June 12 and is disabling the two models for all customers to comply. Other Anthropic models are not affected. The government has not publicly explained the full national security concern, but Anthropic says it understands the order is linked to a reported method for bypassing, or jailbreaking, Fable 5’s safeguards.

Read more
Nvidia’s RTX Spark made me hate content creation a little less
Adobe's AI-powered demos showed me a future where masking, rotoscoping, and scene detection might finally stop being a chore.
NVIDIA RTX Spark for Creators Image Generation Demo Computex 2026

Every video editor has a list of tasks they'd happily outsource to someone else. Exporting isn't one of them anymore because modern laptops are already plenty fast. The real-time sinks are the boring bits: manually masking subjects, finding scene cuts in long recordings, rotoscoping frame by frame, or wrestling with tedious edits that require more patience than creativity.

That's exactly why NVIDIA's RTX Spark demo at Computex 2026 caught me by surprise. I walked into the booth expecting another presentation full of AI buzzwords and benchmark charts. Instead, I walked out thinking that for the first time in years, hardware might actually be changing the editing experience itself, rather than simply making renders finish a little sooner.

Read more
I tried Acer’s new 5K MiniLED Gaming monitor, and OLED kept popping into my head
After seeing it in action at Computex, I finally understand where MiniLED shines and where OLED still wins.
MiniLED vs OLED Hands On Computex 2026

If Computex 2026 taught me one thing, it's that monitor makers are no longer interested in building one-trick ponies. They want displays that can wear multiple hats, seamlessly switching between work and play without making users choose. Acer's new Nitro XV345CKR P is perhaps the best example of that philosophy, and after spending time with it on the show floor, I walked away impressed by its ambition while also questioning whether MiniLED is really the future for gaming monitors.

I've always had a slightly complicated relationship with MiniLED. On a massive living room TV, it works wonders because you're sitting several feet away, and the local dimming zones blend beautifully. Put the same technology on a monitor that's sitting barely two feet from your face, however, and suddenly you're no longer admiring the display, you're inspecting the physics behind it.

Read more