The first batch of AMD’s RDNA 3 GPUs are about to be internally tested for initial performance estimates and any possible design flaws. This process usually is known as the bring-up and follows tape-out after the first batch of dies arrives at the chipmaker’s labs. In this particular case, we are probably looking at Navi 33 which will be the only monolithic die from the RDNA 3 family.
According to Greymon55, it will be the first Navi 3x GPU to be announced, sometime towards the end of the year. Fabbed on TSMC’s 6nm node, Navi 33 will power the Radeon RX 7700 XT with performance exceeding that of the Radeon RX 6900 XT (Navi 21) at roughly half the price.
The Radeon RX 7900 XT (and RX 7800 XT) at the top of the stack will reportedly consist of two GCDs or Graphics Compute Dies. These will be connected using an active bridge interconnect, 3D stacked on top of the two along with 256MB of L3/Infinity Cache. The RX 7900 XT will pack up to a total of 15,360 shaders across 60 Work Group Processors (WGPs), paired with 16GB of GDDR6 memory across a 256-but bus.
The RX 7800 XT, on the other hand, is expected to feature 10,240 shaders across two GCDs. The memory and bus configuration should largely be the same as the 7900 XT. The GPU core will run at around 2.5GHz, the same as Navi 21 which is likely a result of the densely packed logic. Overall, the Navi 31 die powering the Radeon RX 7900 XT should be 2.2-2.5x faster than the RX 6900 XT, with the RX 7800 XT (Navi 32) beating the RX 6800 XT by 50-80%.
Previous coverage on the Radeon RX 7000 GPUs:
AMD’s RDNA 3 graphics architecture is expected to get a major overhaul at the front-end, with redesigned Work Group Processors in place of Compute Units, or Dual Compute Units. With RDNA 1 and 2, the WGPs were the basic units for workload scheduling (from CUs on GCN/Vega), but it looks like that is going to change again with Navi 3x. Dual Compute Units are being discarded in favor of wider Work Group Processors, packing as many as 256 stream processors across eight 32-wide SIMDs. This means that the wave32 format of scheduling will be retained, but the number of overall active waves will be increased.