Today, the infamous leaker Moore’s Law Is Dead (or MILD for short) released a video going over everything he knows about Navi 33, one of the GPUs from the RDNA 3 lineup. For context, at launch AMD is expected to debut three AMD GPUs, Navi 31, Navi 32, and Navi 33. This is similar to 2020’s rollout where AMD launched Navi 21, 22, and 23, that were later used in more and more graphics cards; cut-down versions were utilized to diversify the portfolio.
Navi 33 breakdown
Regardless, Navi 33 will be AMD new entry-level/budget GPU class competing with the likes of NVIDIA’s RTX 4060 and other GPUs in its class, possibly even one from Intel. Navi 33, like the rest of the RDNA 3 lineup, will be fabricated using TSMC‘s 6nm process node and feature a monolithic die. Navi 31 and 32, on the other hand, are set to be the industry’s first GPUs utilizing a MCM (Multi-Chip-Module) design. MILD’s sources say that the die size for Navi 33 will be between 360mm – 460mm² and that it will carry 128MB of Infinity Cache, though the possibility of 256MB of Infinity Cache is still on the table. Now, the crazy part about this GPU is that it’s said to offer performance on par with the Radeon RX 6900XT. That’s right, the top-end flagship RDNA 2 GPU is about to be bested by a budget offering very soon. MILD claims that Navi 33 will offer better rasterization (standard rendering) than the RX 6900XT at 1080. It will be on-par with the 6900XT at 1440p, but trail behind at 4K resolution. As you can tell, the more GPU-reliant the image will become, the less chances there is of the Navi 33 beating the 6900XT. Still, better or similar performance at up to 1440p resolution for a card of its class is no joke. Moreover, raytracing performance is also said to get a massive bump and improve significantly for the Navi 33. So, how is AMD doing all this? How can they offer performance like that yet still retain Navi 33’s lower-end to lower-midrange price bracket? Well, they’ll do this by cutting corners everywhere possible. For starters, Navi 33 is said to come with only 8GB of GDDR6 memory running across a 128-bit bus with speeds expected to be around 18Gbps. Sources see 20Gbps and 22-24Gbps as high-end speeds going forward. Moreover, we’re looking at 8 PCIe 5.0 lanes and an expected TDP of around 180 – 230W. Despite that power limit, MILD’s sources say that they’re heard various AMD engineers claim that RDNA 3 will “decimate NVIDIA in efficiency across the entire lineup”. That’s certainly quite the statement and one that I’m eagerly awaiting to come to fruition as it can give AMD a real leg-up against NVIDIA in this highly-competitive market. The next-gen GPU war between AMD and NVIDIA is shaping up to be likely the closest one ever as both companies are shelling out every resource they have in their belt to make the best GPUs possible. With Intel, a third player being added to the space, the market is now more competitive than ever and with GPUs becoming more and more in demand, the field is open for those who wish to conquer and it only remains to be seen who emerges victorious this next generation.
