For AMDs upcoming GPU launch there had been rumors for months but this past week AMD confirmed that they were launching both the Radeon RX 9070 and Radeon RX 9070 XT. Both cards being their first RDNA 4 architecture cards. Today the performance embargo for both cards lifts and we can check out how they perform ahead of them hitting stores tomorrow. So let's dive in and find out how they compare to yesterday's announced RTX 5070 and more!
Product Name: Acer RX 9070 Nitro OC
Review Sample Provided by: AMD
Written by: Wes Compton
Amazon Affiliate Link: HERE
What is the RX 9070 and RX 9070 XT?
The most interesting thing about AMDs new RDNA 4 architecture was the change in focus from the higher end to targeting the larger upper mid-range market with a targeted price of around $700. With that where some architectures are only focused on how to get the best possible performance, AMD wanted to find the best performance per dollar. RDNA 4 compute unit features an enhanced memory subsystem, improved scaler units, dynamic register allocation, increased efficiency per CU, and clock speeds that are a lot higher than RDNA 3. With all of that, they were able to improve performance per compute unit and offer (according to them) performance similar to their previous generation RX 7900 Series but with a lower compute unit count which helps keep the costs down.
They upgraded their Raytracing Accelerators to their third-generation design with a second ray intersection engine and optimized some of the overhead needed to execute raytracing calculations. This doubled the capabilities when compared to RDNA 3. They also have their 2nd generation AI Accelerators as well to improve performance in AI-accelerated games, content creation, and generative AI. They did this by adding new math pipelines for AI calculations, adding support for new data types like FP8, and support for inference optimization techniques like structured sparsity.
Exclusive to the new RDNA 4 cards, they also introduced FSR 4. FSR 4 improves image quality over FSR 3.1 with an ML-based algorithm designed to improve temporal stability, better preserve detail, and reduce ghosting. By utilizing features that were already part of the FidelityFX API when game developers integrated FSR 3.1, FSR 4 will be available on over 30 games at launch on the 9070 and 90070 XT. It’s always good when the game developers don’t have to push updates out to support!
I touched on it before that AMDs focus was on increasing value but one of the slides in their presentation caught my eye where they highlighted that they are aiming for 4k gaming at a 1440p price. They have the RX 9070 series targeted to sit in between their RX 7800 XT and XTX and the RX 7900 GRE for launch pricing with both 4K and 1440p gaming being possible.
Both the RX 9070 and RX 9070 XT have a lot fewer compute units than any of the 7900 Series cards but how do they compare to each other? The regular RX 9070 has 56 CUs to the XT’s 64. The same can be seen with the raytracing accelerators and for the AI accelerators, those numbers are just doubled up. The RX 9070 has a boost clock of 2.52 GHz and the XT is significantly higher at an impressive 2.97 GHz. You can see how much the clock speed changes things with the AI performance included in the chart, the XT has 14% more AI Accelerators but it does 33% more performance due to the clock speeds as well. Both cards have 16GB of VRAM which is needed for 4k gaming and both cards run on PCIe 5.0 at x16. They have DisplayPort 2.1a and HDMI 2.1b connections. Then for power with the lower clock speed, the RX 9070 has a board power of 220 watts whereas the XT is 304 watts.
Before getting into testing I also ran GPUz to double-check that our clock speeds match up with the specifications. The Acer RX 9070 Nitro OC is an overclocked RX 9070 and has a clock speed of 2700 MHz. GPUz confirmed that. I tested using the 24.30.31.03 beta driver that AMD provided to press ahead of the launch.
