GamingGPUsNews

AMD’s Big Navi (Navi 21) To Feature up to 80 CUs|5,120 Cores; XSX Packs 52 CUs|3,328 Cores

As per reputed Twitter source _rogame, AMD’s Navi 21 GPU will feature up to a maximum of 80 compute units or 5,120 stream processors (cores). That’s twice as much compared to the present Navi 10 flagship, the Radeon RX 5700 XT. Keep in mind that this is the completely unlocked Navi 21 die and the actual Big Navi GPU might come with some compute units disabled.

Looking at the chart above, you can see that the RDNA 2 design is largely unchanged from RDNA 1. The number of shaders or cores per compute unit is unchanged at 64 and the number of SIMD units per CU is also the same (2 per CU). Both designs feature 2 shader arrays per shader engine, although Navi 21 features twice as many shader engines compared to Navi 10:

Navi 21 is basically Navi 10×2 plus all the necessary uarch improvements and additions

The registers and wave configurations are also identical and so are the cache and buffer depths. What’s different is the render backend count: Navi 21 reduces the RB to half as much as Navi 10 – 4 per shader array while the latter had eight. Furthermore, the maximum number of waves per SIMD (grouping of 32 cores, two SIMDs make up one CU) has been reduced to 16 from 20 in Navi 10.

Keep in mind that while the maximum number of waves a WGP (work-group processor or Dual Compute Unit) can fetch is much higher, it only works on two waves at a time (one per SIMD).

For comparison purposes, here are the specs of the Arden APU powering MS’ next-gen Xbox Series X console:

  • The XSX will contain a total of 56 Compute Units with four disabled: That means 52 CUs or 3,328 cores. Navi 21 has 80 CUs or 5,120 cores.
  • There will be two Shader Engines with each SE packing two Shader Arrays each. Navi 21 has four Shader Engines. However, while Navi 21 has 10 CUs per SE, Arden has 14. In other words, Big Navi has 5 WGPs per shader engine while Arden has 7.
  • Arden’s (XSX) has slightly wider Shader Engines compared to Navi 21 but the total number of SEs is half as much. Generally, the GPU utilization or efficiency decreases as you increase the number of CUs in an SE. This means that the utilization on Arden is likely better compared to Big Navi and RDNA 2, most likely due to an enhanced GCP (Graphics Command Processor).
  • You can read more about AMD’s Navi Graphics Architecture here:
  • AMD Navi vs Vega: Differences Between RDNA and GCN Architecture

(Source)

Areej

Computer Engineering dropout (3 years), writer, journalist, and amateur poet. I started Techquila while in college to address my hardware passion. Although largely successful, it suffered from many internal weaknesses. Left and now working on Hardware Times, a site purely dedicated to. Processor architectures and in-depth benchmarks. That's what we do here at Hardware Times!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Back to top button
Close