AMD Radeon RX 9070 XT Review: A 1440p powerhouse with solid gaming chops, but does it hold up against NVIDIA in AI and efficiency?
The AMD Radeon RX 9070 series edition cards launch today. AMD sent the ASUS Prime RX 9070 and RX 9070 XT variants of the cards to the lab for testing. These GPUs, built on the company’s cutting-edge RDNA 4 architecture and TSMC’s advanced 4nm process technology, target the sweet spot of price-to-performance for gamers seeking high-refresh 1440p gaming and entry-level 4K experiences. The 9070 series introduces impressive technological advancements while maintaining reasonable power consumption figures relative to their performance class.
Positioned in the upper mid-range tier, these GPUs blend gaming performance with AI acceleration, catering to gamers and creators. With 16GB of VRAM, enhanced ray tracing capabilities, and AMD’s latest AI-powered technologies like FidelityFX Super Resolution 4 and HYPR-RX, the RX 9070 series aims to deliver high performance while maintaining power efficiency.
The RX 9070 XT is the higher-end model, offering increased clock speeds and more graphics cores for demanding gaming workloads. ASUS’s PRIME design brings robust cooling and power delivery, ensuring consistent performance under heavy loads. Both GPUs are future-ready, with PCIe gen 5, DisplayPort 2.1 support, and AI accelerators for next-gen applications, making them an ideal choice for immersive gaming experiences.
AMD Radeon RX 9070 & RX 9070 XT Specifications
| Radeon RX 9070 XT vs. Radeon RX 9070 Specifications | ||
| Radeon RX 9070 XT | Radeon RX 9070 | |
| GPU Name | Navi 48 | Navi 48 |
| GPU Variant | Navi 48 XT | Navi 48 XL |
| Architecture | RDNA 4.0 | RDNA 4.0 |
| Foundry | TSMC | TSMC |
| Process Size | 4 nm | 4 nm |
| Transistors | 53,900 million | 53,900 million |
| Density | 151.0M / mm² | 151.0M / mm² |
| Die Size | 357 mm² | 357 mm² |
| Base Clock | 2400 MHz | 2070 MHz |
| Boost Clock | 2970 MHz | 2520 MHz |
| Memory Size | 16 GB | 16 GB |
| Memory Type | GDDR6 | GDDR6 |
| Shading Units | 4096 | 3584 |
| AI Accelerators | 128 | 112 |
| TMUs | 256 | 256 |
| ROPs | 128 | 128 |
| Compute Units | 64 | 56 |
| RT Cores | 64 | 56 |
| AMD Infinity Cache | 64 MB (3rd Gen) | 64 MB (3rd Gen) |
| Memory Bus | 256 bit | 256 bit |
| Bandwidth | 640 GB/s | 640 GB/s |
| Peak Pixel Fill-Rate | 190.1 GPixel/s | 161.3 GPixel/s |
| Peak Texture Fill-Rate | 730.3 GTexel/s | 564.5 GTexel/s |
| Peak Single Precision Throughput | 48.7 TFLOPS | 36.1 TFLOPS |
| Peak Half Precision Throughput | 97.3 TFLOPS | 72.3 TFLOPS |
| Peak INT8 AI TOPS Throughput | 779 TOPS | 578 TOPS |
| Peak INT4 AI TOPS Throughput | 1557 TOPS | 1156 TOPS |
| TDP | 304 W | 220 W |
| Suggested PSU | 750 W | 650 W |
| Outputs | 1x HDMI 2.1b, 3x DisplayPort 2.1a | 1x HDMI 2.1b, 3x DisplayPort 2.1a |
| PCIe Interface | PCIe 5.0 x 16 | PCIe 5.0 x 16 |
| Power Connectors | 3x 8-pin | 2x 8-pin |
| Launch Price | TBD | TBD |
What’s New with RDNA 4
AMD’s RDNA 4 architecture represents a significant evolution from previous generations, bringing substantial improvements across multiple domains. At the heart of these new GPUs lies the redesigned RDNA4 Compute Units, which deliver approximately 40% more performance per compute unit than RDNA 3. This efficiency gain comes from improved memory subsystems and more optimized data flow pathways throughout the GPU.
Ray tracing performance sees a dramatic uplift with AMD’s third-generation RT accelerators. Adding a second ray intersection engine effectively doubles the throughput for ray-box and ray-triangle testing operations, critical bottlenecks in previous generations. New dedicated ray transformation hardware further accelerates these workloads, while support for Oriented Bounding Boxes enables more efficient ray traversal with lower memory overhead.
Perhaps the most significant advancement comes in AI acceleration capabilities. The RX 9070 series features second-generation AI accelerators that deliver approximately 1,100 AI TOPS for the 9070 and over 1,500 AI TOPS for the 9070 XT when using INT4 calculations with sparsity. These dedicated AI units feature expanded math pipelines and support emerging AI data formats.
The new FSR4 technology is AMD’s answer to NVIDIA’s DLSS upscaling, leveraging the enhanced AI accelerators to deliver what AMD claims can be better-than-native image quality. Using machine learning-based upscaling algorithms specifically optimized for RDNA 4 hardware, FSR4 can increase performance while maintaining excellent image fidelity. AMD has designed the system to be backward compatible with the FSR 3.1 API, allowing developers to implement it easily into existing titles.
Build and Design ASUS Prime RX 9070 and RX 9070 XT
The build and design of the new ASUS Prime Radeon 9070 and 9070 XT cards closely resemble the ASUS Prime RTX 5070 Ti we previously reviewed. They differ on the power side — neither card requires a special 12-pin power cable, with the 9070 using standard 8-pin power with 2x 8-pin connectors and the 9070 XT using 3x 8-pin connectors. Both cards feature the 2.5 PCIe slot design height requirement.
For cooling, ASUS has implemented a triple Axial-tech fan design with dual ball bearings on the Radeon 9070 and 9070 XT to improve durability and maintain consistent airflow. The fans are paired with a large heatsink and heat pipes across the entire card, while a vented backplate and side cutouts help maximize heat dissipation. All three fans remain idle when GPU temperatures fall below 50°C, enabling silent operation during light workloads or less demanding gaming sessions. The fans automatically resume operation once temperatures surpass 55°C.
Looking at the bottom slot side of the cards, both models share an identical design for cooling, featuring the same heatsink, PCB layout, and shroud with venting cutouts to aid airflow.
On the end of both cards is an identical mounting point for an optional card stiffener to prevent card sag while installed in a case and provide added stability during transport.
On the business end of both cards, we have display connectivity options, each featuring 1x HDMI 2.1b and 3x DisplayPort 2.1a ports. ASUS has also incorporated a 304 stainless-steel PCIe bracket design to ensure longevity and durability.
Other notable features of both cards include a switchable BIOS, controlled via a switch on the card, which toggles between “P” (performance) mode and “Q” (quiet) mode. Both units also incorporate a phase-change thermal pad designed to improve cooling performance and enhance the longevity of the thermal material in contact with the chip. Lastly, ASUS provides GPU Tweak III software for fine-tuning the card’s performance to meet specific requirements.
Engage with StorageReview
Newsletter | YouTube | Podcast iTunes/Spotify | Instagram | Twitter | TikTok | RSS Feed
