The key points of this speculation include that AMD is focusing on “area, area, area” which means smaller chips are still powerful enough, Enough to meet Team Red’s performance targets for RDNA 3 and also focus on performance per watt (efficiency). Of course, the company has already boasted that the latter is a 50% increase over RDNA 2 (the original RDNA architecture achieved the same jump compared to RDNA 2).
Put your own seasoning on it. Compared to all this, Angstronomics tells us that the Navi 31 – the flagship model of the RX 7000 graphics card – will, as rumored, come with a Graphics Chiplet Die (GCD) and six Memory Chiplet Dies (MCD) with 12,288 cores (called ALU). The size of the main GCD is apparently 308mm², another big change from the previously mentioned guess is the size of the Infinity Cache, which is apparently 96MB for the top model (the reduced Navi 31, probably the non-XT version, will use 80MB) .
The rumor mill mentions a larger size than that, like 192MB; in fact, the existing Navi 21 uses 128MB of Infinity Cache, So it will actually be smaller than that. Angstronomics reckons AMD tried twice the stack cache for the Navi 31, but that didn’t provide enough extra performance to be worth doing. So some of the specs surrounding these rumors are slightly disappointing, and the second point that raises a little concern is the purported Navi 32 configuration, Angstronomics claims This configuration will run 7,680 cores instead of 8,192 as previously thought. The Infinity Cache is apparently set up to be up to 64MB, maybe 128MB for a 3D stacked model (but again, it seems like AMD might have considered this idea before dropping it because it didn’t make sense in terms of cost for the extra performance gained).
As for the mid-range Navi 33, if Angstronomics makes any money, it will be much smaller, it was rumored that the chip might be close to 400mm2, but it actually weighs about 200mm2. It should run with an infinite cache of 32MB.
Regarding the cooling system of the flagship RDNA 3 graphics card, AMD is believed to have kept it with the existing flagship, but it will be a bit higher. According to the leaker, the GPU will use a pair of 8-pin PCIe power connectors.
Analysis: Worried about it all? Well, maybe it’s Nvidia that should be worried…
As mentioned before, some people have been worried about this so-called Elements of the specification, and some seem to have been slightly reduced by a few bits compared to the previous gossip – or in the case of Infinity Cache a lot. However, there’s really no need to worry, in fact, it could be that Nvidia should be doing it, and here’s why.
Interestingly, Moore’s Law is dead
(opens in new tab)
(MLID) is a well-known YouTube leaker who quickly jumped on the topic, and after speaking with multiple sources, supports pretty much everything Angstronomics has to say here. That said, there is one key difference, that for the Navi 32 GPU, MLID still hears that it will be 8,192 cores instead of the 7,680 mentioned in this new leak. Of course, MLID still admits that his sources could be wrong, so it could go in any direction (and 7,680 cores is certainly doable).
The source of the MLID does broadly support the specs mentioned here and the smaller die size, and less Infinity Cache load – but that’s the key aspects. MLID noted that the performance estimates he’s heard haven’t really changed, and he still expects AMD to keep the rumoured performance leaps we’ve heard in the past — enough to compete with Nvidia’s RTX 4000 series — even if the specs have been tweaked along the lines above. .
Even if the performance is indeed slightly weaker than Nvidia, the theory of MLID shouldn’t make much difference. And we have to keep in mind that given the mentioned smaller die size, it will be cheaper for AMD to manufacture these RDNA 3 GPUs, so they are priced to attack the Nvidia RTX 4000 models. That’s probably the most exciting thing about next-gen graphics cards, and Nvidia could be in trouble if AMD gets aggressive with pricing — especially given that its more power-hungry GPUs (at least the rumor mill) could mean other Complications such as power upgrades.
Not to mention the difficulty Nvidia has had in deciding on the release schedule for the next-gen Lovelace, which will likely give AMD the first RDNA 3 graphics card, among all these other factors On the basis of the use of their own sales advantages. Of course, we shouldn’t be carried away by any of these rumors, but we can imagine that the latest leak has to worry Nvidia in several ways.
Oh, as for Intel, Angstronomics mentions that the Navi 33 surpasses the flagship Arc Alchemist GPU while being more power efficient and costs less than half the cost, which is clearly sinking Team Blue’s desktop GPU ( already struggling) torpedoes. Even without a next-gen competitor’s offering, things are bad).
Darren is a freelancer and writes news and features for TechRadar (and occasionally T3) on a wide range of computing topics, including CPUs, GPUs, various other hardware, VPNs, antivirus software, and more Wait. He has spent most of the past 3 years writing tech articles and writing books in his spare time (his debut I Know What You Made Your Last Supper – published by Hachette UK in 2013).
