AMD at the 2025 International CES announced the Radeon RX 9070 XT and Radeon RX 9070 desktop performance-segment graphics cards. These will be the face of AMD's next generation of gaming graphics products, and will be powered by the new RDNA 4 graphics architecture. AMD hopes to launch both cards within Q1 2025. AMD changed the nomenclature of its gaming GPUs mainly because it has made a tactical retreat from the enthusiast graphics segment, its fastest products will compete in the performance segment. From the way AMD arranged the Radeon RX 9070 series and 9060 series product stack against the backdrop of the Radeon RX 7000 series, the GeForce RTX 4000 series, and the anticipated GeForce RTX 5000 series, the RX 9070 XT will offer performance roughly similar to the Radeon RX 7900 XT in raster, with the RX 9070 being slightly faster than the RX 7800 XT. The RX 9060 XT will beat the RX 7700 XT, while the RX 9060 beats the RX 7600 XT.

With RDNA 4, AMD claims generational SIMD performance increase on the RDNA 4 compute units. The 2nd Gen AI accelerators will boast of generational performance increase, and AMD will debut a locally-accelerated generative AI application down the line, called the AMD Adrenalin AI, which can generate images, summarize documents, and perform some linguistic/grammar tasks (rewriting), and serve as a chatbot for answering AMD-related queries. This is basically AMD's answer to NVIDIA Chat RTX. AMD's 3rd Gen Ray accelerator is expected to reduce the performance cost of ray tracing, by putting more of the ray tracing workload through dedicated hardware, offloading the SIMD engine. Lastly, AMD is expected to significantly upgrade the media acceleration and display I/O of its GPUs.

AMD is expected to have a rather lean lineup of next-generation gaming GPUs powered by the RDNA 4 graphics architecture. The series is expected to debut at AMD's 2025 International CES keynote address, with product launches of the series-leading Radeon RX 9070 XT and RX 9070 performance-segment GPUs later this month. The RX 9070 should be available by late-January, although add-in board partners from China expect availability to ramp in February 2025. The series will see expansion with more announcements in March.

The RDNA 4 generation is driven mainly by two chips—the larger "Navi 48," and the smaller "Navi 44." The "Navi 48" will power the RX 9070 series, which are performance-segment and designed to compete with the NVIDIA GeForce RTX 5070 series; but cut-down variants of the chip are also expected to power certain upper mid-range SKUs that go up against the RTX 5060 series. The "Navi 44" chip is expected to power certain high performance/price SKUs in the mid-range, which AMD will use to target price-points well under the $300-mark. This segment is expected to heat up as NVIDIA has current-generation RTX 4060 series, Intel just made a stab with the Arc B580, and is expected to launch a faster Arc B700-series SKU based on a maxed-out "BMG-G21" silicon.

The "Navi 48" silicon powering AMD's next-generation Radeon RX 9070 series could feature AV1 hardware-accelerated encoding with support for AV1 B-Frames. In video compression, a B-frame is an intermediate frame that lacks image information, but has motion-vector and other data from the previous and next image frames (or I-frames), which helps the decoder reconstruct the image component of the frame based on temporal frame data. This is compute-intensive, but greatly reduces file-size or bitrate of the stream, as almost every other frame lacks image information. Support for AV1 B-Frame hardware-accelerated encode was sniffed out by HXL in a recent commit to one of the SDKs AMD maintains in a public repository through its GPUOpen initiative.

AMD's Radeon RX 9000 series generation powered by the RDNA 4 graphics architecture will be based almost entirely on two chips, the "Navi 48" and "Navi 44," with the latter powering mainstream and mid-range SKUs; while the former powers performance-segment ones. There is no enthusiast-segment chip this time around. The "Navi 48" is expected to feature a more advanced video encode/decode hardware than the one RDNA 3.5 comes with; and AV1 is likely to get the bulk of development as the royalty-free codec gains popularity with online video streaming services. It remains to be seen if next-generation architectures like RDNA 4 or NVIDIA's "Blackwell" support acceleration for VVC.

A Chinese tech forum ChipHell user who goes by zcjzcj11111 sprung up a fascinating take on what the next-generation AMD "Navi 48" GPU could be, and put their imagination on a render. Apparently, the "Navi 48," which powers AMD's series-topping performance-segment graphics card, is a dual chiplet-based design, similar to the company's latest Instinct MI300 series AI GPUs. This won't be a disaggregated GPU such as the "Navi 31" and "Navi 32," but rather a scale-out multi-chip module of two GPU dies that can otherwise run on their own in single-die packages. You want to call this a multi-GPU-on-a-stick? Go ahead, but there are a couple of changes.

On AMD's Instinct AI GPUs, the chiplets have full cache coherence with each other, and can address memory controlled by each other. This cache coherence makes the chiplets work like one giant chip. In a multi-GPU-on-a-stick, there would be no cache coherence, the two dies would be mapped by the host machine as two separate devices, and then you'd be at the mercy of implicit or explicit multi-GPU technologies for performance to scale. This isn't what's happening on AI GPUs—despite multiple chiplets, the GPU is seen by the host as a single PCI device with all its cache and memory visible to software as a contiguously addressable block.

GPU chip packages of the "Navi 4x" generation of GPUs could be generationally smaller than their predecessors, according to leaked package dimensions of the "Navi 44" chip put out by Olrak29_. With its next-generation Radeon RX gaming GPUs based on the RDNA 4 graphics architecture, AMD has decided to focus on gaining market-share in the performance and mainstream segments, ceding the enthusiast segment to NVIDIA. As part of its effort, the company is making RDNA 4 efficient at every level—architecture, process, and package.

At the architecture level, RDNA 4 is expected to improve performance, particularly the performance cost of ray tracing, through a more specialized ray tracing hardware stack. At the process level, AMD is expected to switch to a more efficient foundry node, with some reports suggesting the TSMC 4 nm, such as the N4P or N4X. For a mid-range GPU like the "Navi 44," which succeeds the "Navi 23" and "Navi 33," these mean a rather big leap from the 7 nm or 6 nm DUV nodes. The leak suggests a smaller package, measuring 29 mm x 29 mm. In comparison, the "Navi 23" package measures 35 mm x 35 mm. The smaller package could make these GPUs friendlier with gaming notebooks, where mainboard PCB real-estate is at a premium.

AMD's next generation RDNA 4 graphics architecture will see the company focus on the performance segment of the market. The company is rumored to not be making a successor to the enthusiast-segment "Navi 21" and "Navi 31" chips based on RDNA 4, and will instead focus on improving performance and efficiency in the most high-volume segments, just like the original RDNA-powered generation, the Radeon RX 5000 series. There are two chips in the new RDNA 4 generation that have hit the rumor mill, the "Navi 48" and the "Navi 44." The "Navi 48" is the faster of the two, powering the top SKUs in this generation, while the "Navi 44" is expected to be the mid-tier chip.

According to Kepler_L2, a reliable source with GPU leaks, and VideoCardz, which connected the tweet to the RDNA 4 generation, the top "Navi 48" silicon is expected to feature a 256-bit wide GDDR6 memory interface—so there's no upgrade to GDDR7. The top SKU based on this chip, the "Navi 48 XTX," will feature a memory speed of 20 Gbps, for 640 GB/s of memory bandwidth. The next-best SKU, codenamed "Navi 48 XT," will feature a slightly lower 18 Gbps memory speed at the same bus-width, for 576 GB/s of memory bandwidth. The "Navi 44" chip has a respectable 192-bit wide memory bus, and its top SKU will feature a 19 Gbps speed, for 456 GB/s of bandwidth on tap.