Tweet From AMD’s Gaming Marketing Director Might Give Us A New Hint At RDNA 3 GPUs Performance For Radeon RX 7000 Cards
Sasa didn’t mention any product within the tweet but I think we can make a few guesses. First of all, it doesn’t seem like this is related to AMD’s Ryzen 7000 CPUs because we have already got a load of information regarding the Zen 4 CPUs. The only other gamer-focused product that AMD has coming out is their Radeon RX 7000 series based on the RDNA 3 GPU architecture which were briefly teased at the “Together We Advance_PCs”.
— Sasa Marinkovic (@SasaMarinkovic) September 14, 2022
AMD Focusing On DIY & PC-Friendly Graphics Card Designs & Cooling?
So the tweet itself which reads “bigger doesn’t necessarily mean better” could be pointing out to the recent NVIDIA GeForce RTX 4090 leaks which have shown the card to feature insanely huge & quad-slot cooling solutions. The NVIDIA GeForce RTX 4090 graphics card is expected to be unveiled in a few days at GTC 22 however, AMD has already started to take shots at their competitor’s next-gen parts which will be coming out a little earlier than their own next-gen solutions. The NVIDIA GeForce RTX 4090 custom graphics cards that have leaked out are so far reported to feature a single 16-pin connector (12VPWHR) for up to 600 Watts of power and given their higher TGP, they will require some hefty cooling solutions like the one we have seen leak out recently. It looks like AMD might not go that route and deliver solutions that feature a more user & PC-friendly dual-slot design. Earlier rumors have pointed out a slightly bigger yet still 2.5-slot cooling solution with two 8-pin connectors to boot. So AMD’s Radeon RX 7000 series may come out with a card that not only consumes lower power but also doesn’t require large cooling solutions.
Or Is This About The GPU Die Sizes?
AMD also pointed this out a while back, stating that while their cards are a bit more power-hungry than RDNA 2, they will be much more efficient than the competition. Do remember that the red team is targetting a 50% perf/watt uplift over RDNA 2 with its new 5nm RDNA 3 GPUs. The other thing that Sasa could have pointed out could be the die size. AMD’s RDNA 3 powered Radeon RX 7000 series graphics cards are going to be the first gaming products to feature an MCM solution. As such, the GPU could move several operations off-die in chiplets such as the Infinity Cache, leading to a smaller GPU die size. The GPU will be designated as a GCD or Graphics Complex Die while the Infinity Cache and Memory interface will be moved over to multiple MCDs or Memory-Chiplet Die.
— Kepler (@Kepler_L2) September 14, 2022 This would in return lead to a smaller GPU footprint and as per the latest rumor by Kepler_L2, the Navi 31 will measure 533mm2 (308mm2 GCD + 37.5mm2 per MCD) while the NVIDIA AD102 GPU will measure 628mm2 in a single monolithic package. This is a reduction of 40% in die size using the same TSMC 5nm process node. Plus, it will be easy to compare the transistor density and transistor efficiency of both chips since they are both using the same TSMC node instead of the last gen where RDNA 2 and Ampere chips were based on different process technologies from different companies (Samsung 8nm / TSMC 7nm). At 308mm2, the Navi 31 GPU will be more densely packaged than the Navi 21 GPU which measures 520mm2 and houses 26.8 Billion transistors (51.5/mm2).
Regardless, We Are Excited For AMD’s RDNA 3 GPUs, Bring It On!
AMD did manage to close up a huge performance gap with its RDNA 2 GPU-based Radeon RX 6000 in the previous generation and even managed to outperform its competitors with stellar efficiency & performance. The company is planning to do it once again and this time, even better with its first chiplet GPU architecture, the RDNA 3, designed by David Wong and his team for the next-gen RX 7000 series graphics cards. We can’t wait to see the latest GPUs in action later this year.
