AMD brings chiplets to Radeon Mobile at last with the new RX 7900M

pcgamer.com:

Ten months after the first RDNA 3-powered laptop GPUs appeared, AMD has finally got around to using its premier Navi 31 processor in a mobile configuration. The new Radeon RX 7900M is being touted as being 7% faster on average than Nvidia's RTX 4080 Laptop, so fans of high-end gaming on the go might end up going all Team Red.

In terms of specs, just think Radeon RX 7900 GRE, but a little bit slower. Like all of the Radeon RX 7900-series models that use the RDNA 3-based Navi 31, this new GPU is a medium-sized lump of silicon, containing all of the graphics circuitry (called the GCD), surrounded by a number of tiny chiplets that house the L3 cache and memory interfaces (MCDs).

Something else that's common to them all is that the MCD is pretty much identical. In fact, the only differences are the number of active Compute Units (CUs) and the clock speeds used. In the case of the RX 7900M, you're getting 72 CUs with a game clock of 1,825MHz. For reference, the RX 7900 GRE sports 80 CUs and has a game clock of 1,880MHz.

The boost clock for the RX 7900M is 'up to 2,090MHz' because the value is really set by the laptop vendor. The maximum clock speed is limited by how much power the GPU is allowed to consume and in the case of this one, the specifications list a total graphics power of 'up to 180W'. So if the laptop has a lower limit than this, then the boost clock will be less than AMD's value.

Something else that it shares with the GRE card is the number of MCDs. This determines how much VRAM the GPU will have access to, along with the total width of the memory bus and the amount of L3 Infinity Cache. For the RX 7900M, it's 16GB of GDDR6, all running at 16Gbps (576GB/s of bandwidth) and 76MB of cache.

For me, the best thing