Despite the next generation of graphics cards drawing near, we haven’t received formal announcements from the trio of GPU manufacturers. Like Intel and Nvidia, AMD, too, has kept its cards close to its chest. Leaving my terrible pun aside, there have been a couple of rumors circulating about the RX 8000 family. If you’re interested in Team Red’s upcoming GPUs, here’s a list of all the leaks we have on the successor to the RDNA 3 series.

AMD Radeon RX 8000 series: What we’re hoping to see, and how much we're willing to pay

Going by past trends, AMD releases new GPUs in a 2-year cycle, and the company has a tendency to unveil low-end models and XT variants anywhere between eight and twelve months after releasing the first set of its next-gen cards. Considering the first two Radeon RX 7000 graphics cards were released in December 2022 and the RX 7600 XT and RX 7900 GRE started making the rounds this year, one can speculate that AMD intends to launch the RX 8000 series in early 2025.

These estimations also overlap with the claims made by the prominent leaker Kepler (@Kepler_L2). According to them, the first (Navi 48) of the two sets of next-gen GPUs from Team Red will debut at CES 2025, while the second one (Navi 44) could hit the shelves in Q2 2024.

As leaked by Moore’s Law Is Dead, the flagship next-gen GPU from AMD will range between $499 and $599. If that sounds a little on the low side for a flagship card, that’s because Team Red plans to change its stance with the Radeon RX 8000 family.

AMD Radeon RX 8000 series: What we could see

Four SKUs, split into two series

Last year, the code for AMD’s Linux LLVM patches revealed the existence of two RX 8000 GPUs, the GFX1200 and the GFX1201. The first one is expected to belong to the weaker Navi 44 series, while the latter is believed to be a part of the Navi 48 lineup. Fast-forward to September 2024, and Benchlife claimed that both sub-series would feature two SKUs, bringing the total number of GPUs to four.

Recently, leaked Geekbench scores revealed the key details for the Navi 48 SKU. Assuming it’s the flagship card, we’re looking at a 2,101MHz clock speed and 16GB VRAM. Both specs are somewhat low for the top-of-the-line GPU, though the most surprising aspect is the presence of merely 28 Compute Units in the Geekbench statistics. That said, it’s possible that the number corresponds to Work Group Processors (WGPs) instead of CUs. Since each WGP is equivalent to 56 Compute Units, we’re looking at 56 CUs this generation, which is nearly half the number found in the current-gen flagship, RX 7900 XTX.

According to his sources, Moore’s Law Is Dead estimated the power draw will vary between 210W to 280W, noting the flagship will compete with the RTX 4080 in rasterization performance and the RTX 4070 Super in ray-tracing workloads. The memory details have been in line with the Geekbench leaks: 16GB GDDR6 operating on a 256-bit bus. The boost clock and CU count have been rather inconsistent, with MLID’s leaks pointing at 64 Compute Units and 2.9GHz-3.2GHz frequency.

Mostly mid-range GPUs in this generation

Rumors about AMD’s exit from the high-end GPU market have been floating around for a while. In a recent interview with Tom’s Hardware, Jack Huynh, the company’s senior vice president, all but confirmed that Team Red won’t compete with Nvidia for enthusiast-grade GPUs.

While it’s still possible for AMD to change gears, it seems like the RX 8000 series will mostly cater to the mid-range audience. This lines up with the leaks about the (supposedly) high-end RX 8900 XTX getting canned by AMD. Assuming the flagship RX 8000 card costs $499-$599, it lies somewhere between the prices of the Radeon RX 7800 XT and the RX 7900 GRE, both of which are mid-range cards designed for 1440p gaming.

4nm monolithic die

We’ve seen AMD GPUs with monolithic designs like the RX 7600 for a while now, though AMD’s new Multi-Chip Module sent rumors about the company switching up the architecture for the RDNA 4 series. While there were rumors about the canceled high-end model featuring an MCM design, Moore’s Law Is Dead reported that new GPUs will feature the old monolithic design. The manufacturing process, however, is expected to switch to 4nm.

Another fun aspect of AMD’s future GPUs is that the company plans to take a unified approach to consumer and data center cards. Rather than splitting the cards into the RDNA and CDNA ecosystem, Team Red intends to go with a unified UDNA architecture. That said, Jack Huynh never mentioned RDNA 4 in the interview with Tom's Hardware. Given how a sudden shift in the architecture wouldn’t make a lot of sense this late in the game, it’s safe to assume that UDNA probably won’t be featured in the Radeon RX 8000 graphics cards.

Are you excited about AMD’s upcoming cards?

Although AMD’s focus on mid-range graphics cards sounds reassuring for those gaming on a budget, I’ll admit that the transition makes me a bit fearful about the future of the GPU industry. Sure, the low power requirements of the RX 8000 flagship sound pretty tantalizing. However, my biggest issue is that Nvidia will be the sole manufacturer of premium GPUs. As much as I don’t want to sound like a fearmonger, there’s no denying that things are looking grim for high-end gamers.

Given Team Green’s previous track record, we’d already expected borderline insane prices for the upcoming RTX 5000 series. But with AMD pulling out of the $1000 GPU market, Nvidia has no reason to set reasonable prices for Blackwell GPUs. And if the rumors about the RTX 5090’s 600W TGP and 32GB GDDR7 VRAM are anything to go by, Nvidia’s top-of-the-line Blackwell card might have the most ridiculous price tag out of all consumer-grade GPUs released in the last decade.