AMD’s “Big Navi” reportedly has 2x as many CUs as Radeon’s RX 5700 XT

AMD's

AMD’s “Big Navi” reportedly has 2x as many CUs as Radeon’s RX 5700 XT

Two independent leakers have released information about AMD’s RDNA 2 “Big Navi” flagship, stating that it will have 80 total Compute Units (CUs), delivering a two-times increase in CU count over today’s RX 5700 XT.

Both @KOMACHI_ENSAKA and @_rogame this information on Twitter, refuting prior rumours that AMD’s “Big Navi” silicon would feature 72 total CUs. 

Given AMD’s promise of a 50% increase in performance-per-watt with their RDNA 2 architecture and the size of chips like Nvidia’s Ampere A100, a graphics chip that’s 2x larger than today’s RDNA-powered RX 5700 XT seems very feasible. That said, such a chip would be harder to manufacture than today’s RX 5000 series products, increasing production costs for Radeon.  

When compared to today’s RDNA-powered PC graphics cards, the core clock speeds of AMD’s RDNA 2-powered console silicon (Microsoft’s Xbox Series X and Sony’s PlayStation 5) are much higher than what many analysts expected for a new console. Within the console market, power efficiency, and system cooling are essential, making PC level clock speeds out of the question for most system designs. With this in mind, Radeon’s RDNA 2 discrete (PC) graphics solutions should be expected to clock much higher than their RDNA-powered counterparts. This means that a greater than 2x performance gain could be possible with AMD’s RDNA 2 flagship. 

Looking at the peak clock speeds of Sony’s RDNA 2 powered PlayStation 5 console, we know that AMD’s next-generation GPUs will be able to deliver peak clock speeds that meet or exceed 2.23 GHz. Sony’s PlayStation 5 can clock its RDNA 2 GPU to a maximum of 2.23GHz, a speed which should be achievable on practically all PS5 SoCs to maintain high silicon yields. With this in mind, similar or greater clock speeds should be expected on PC, thanks to the PC market’s ability to utilise more aggressive boost clock speed technologies. 

Compare the PlayStation 5’s GPU clock speeds to AMD’s RDNA-based RX 5700 XT’s maximum (reference) boost clock speed of 1905 MHz. The PlayStation 5’s peak clock speeds are 17% higher than what we have seen with AMD’s current RDNA flagship, which means that the same, or similar, boost should be possible on AMD’s RDNA 2 flagship. 

Assuming that no other limitations come into play, a 100% increase in CU count and a 15% increase in GPU clock speeds should allow AMD’s RDNA 2 flagship to deliver a 130% boost in performance over today’s RX 5700 XT. That’s more than enough to surpass Nvidia’s RTX 2080 Ti. Beyond that, AMD could deliver further performance gains through architectural enhancements. These estimates are a best-case-scenario for AMD, as scaling a graphics card to this chip size could result in sub-par performance scaling. This was the case with AMD’s RX Vega series GPUs, and history could repeat itself with RDNA 2. 

Rumour has it that AMD's  

If these rumours are legitimate, AMD’s RDNA 2 flagship is going to be a powerful product. That said, it is unknown how powerful Nvidia’s Ampere gaming products will be, and that will ultimately dictate the pricing of AMD’s upcoming RDNA 2 products. 

You can join the discussion on AMD’s “Big Navi” RDNA 2 flagship containing 80 compute units on the OC3D Forums.