The truth is that AMD advanced it at the time and it was really hard to believe because it was something more informal, like a supposed advancement that would be possible in the future and that it could happen or not . But at FDA 2022 last night, the company laid the groundwork for what the new RX 7000 GPUs will be and as such, some of the rumors have been confirmed and some of the details are surprising. So will be the AMD RDNA 3 graphics card.
Confirmations here and there, rumors that are true and others that are not, but in the end the river bed narrows. What was revealed yesterday by AMD is really little, very brief, but it leaves big impressions and invites to have illusions about a possible surprise from NVIDIA.
AMD RDNA 3 with RX 7000, small big improvements
We will start by saying that the main thing is the announced increase of 50% in performance per watt. It’s the same jump from RDNA to RDNA 2 and now it’s happening again in RDNA 3 which is very impressive considering the node used is TSMC’s 5nm. So how did AMD do it? Well, there are quite a few big changes, which are small individually, but together add up to big improvements.
Starting with the fact that AMD will use chiplets in this architecture, but apparently not as expected. It’s all about changes in Calculation units or CUsince they have been redesigned in the new architecture with an “alleged” premise: there will be less of them, which is an interesting concept change.
It looks like there will be more WGPs, but instead they will have two UCs with double the SIMD32
Five chiplets at its maximum configuration
And it is that AMD is going to surprise with an unusual and totally revolutionary configuration, since now we have to talk about GCD or Graphics calculation matrices (GPU cores as such) and MCD or Multi-Cache Dies (the annexes where they release the Infinity Cache, and supposedly the L2).
What we do not know because it has not been specified as such, is in which lithographic process each one arrives. It is assumed that TMSC will manufacture the GCD at 5nm, while the MCDs could come from Samsung or the Taiwanese.
Going back to the architecture, each GCD is equipped with 4 Shader Engines and each of them will consist of 2 Shader Arrays, where at the same time each of them has 4 WGPs, a total of 32 SMID32 ALUs, a total number of 8192 Shaders on the mid-range chip.
The oldest of them goes to 48 WGP with 12 SA and 6 SE to go 12,288 coreswhile Navi 33 will have 5120 in all. It is being done this way as it is estimated that AMD might include Navi 32 instead of Navi 31 in the market first, leaving NVIDIA at the high end and focusing on support to capture more sales.
The dates are unknown at the moment, as are the prices, but it is said that they could arrive between October and November of this year.
