Introduction

ASRock Radeon RX 9070 XT Taichi is the company’s top-of-the-line custom design rendition of the exciting new performance segment GPU from AMD that’s breathing life back into the gaming graphics market. The Taichi line of graphics cards represents the very best in product design, cooling performance, and overclocker-friendly features from ASRock. These cards are designed to visually match with ASRock’s latest Taichi, Taichi Lite, and Taichi OCF series motherboards based on the Intel Z890 and AMD X870E chipsets, although they should be the star attraction in any PC build. The Radeon RX 9070 XT is designed for 4K Ultra HD gaming at high settings, and 1440p gaming at maxed out settings; both with ray tracing enabled. AMD claims to have made leaps in ray tracing performance, as well as performance per CU, allowing the RX 9070 XT to be a lean GPU focused on PC gaming.

The Radeon RX 9070 XT is powered by AMD’s latest RDNA 4 graphics architecture that is claimed to have made substantial gains in performance per compute unit (CU), as well as ray tracing and AI acceleration; which together make the RX 9070 series contemporary. The ray tracing performance gains lower the cost of having it enabled in games, given that RT is no longer a novelty. AI acceleration performance gains allow AMD to bring AI ML closer to gaming. A case in point is FSR 4, a major update to AMD’s performance enhancement suite, which uses an ML-based upscaler that offers significant image quality upgrades at every performance tier over previous versions of FSR.

The Radeon RX 9070 XT is based on the 4 nm Navi 48 silicon, which it maxes out. Unlike NVIDIA, which stuck to the exact same process node for the Blackwell generation as the Ada generation, AMD gave the Navi 48 two key process-level upgrades. Firstly, the company switched to the 4 nm TSMC N4P node, which offers clock speed and efficiency upgrades over the previous TSMC N5 node, and secondly, Navi 48 is a monolithic silicon unlike Navi 31, which is a chiplet-based GPU with a 5 nm GCD and 6 nm MCDs. The entire Navi 48 chip is made on 4 nm, including the memory controllers and Infinity Cache, and then there are power management and IPC improvements from the RDNA 4 architecture.

The Navi 48 silicon features 64 compute units (CU), all of which are enabled on the RX 9070 XT. This works out to 4,096 stream processors, 128 AI accelerators, 64 RT accelerators, 256 TMUs. The chip also has an impressive 128 ROPs, which is an increase over the 96 that the Navi 32 comes with. The card has 16 GB of memory across a 256-bit wide memory interface. AMD stuck to the older GDDR6 standard and is using 20 Gbps memory speeds, which results in 640 GB/s of memory bandwidth. AMD is counting on new architecture-level features, such as out-of-order memory management, and the faster 64 MB Infinity Cache to keep the RX 9070 series competitive with the GeForce RTX 5070 series.

The ASRock Radeon RX 9070 XT Taichi comes with the latest generation of the company’s Taichi 3X cooling solution that uses a massive aluminium fin-stack heatsink with six heat pipes, and a solid copper baseplate to pull heat from the GPU and memory. It also implements the new 12V-2×6 power connector standard, although you do get an adapter that converts from three standard 8-pin connectors. It’s the only other AMD board partner apart from Sapphire to implement this. ASRock has given the RX 9070 XT Taichi a factory overclock of 2570 MHz Game clock, compared to 2400 MHz reference. There are also other overclocker-friendly features such as dual-BIOS. ASRock is pricing the RX 9070 XT Taichi at $730, a $130 premium over the $600 baseline price by AMD.