Comments by "Valen Ron" (@valenrn8657) on "Can AMD beat NVIDIA?" video.

  1. Big Navi with 80 CU and 20.5 TFLOPS is about RTX 3070 level. IF Big NAVI has 1TB/s memory bandwidth from HBM2e, then it would be above RTX 3070 and RTX 3080 level. Against GA-102 (RTX 3090 and RTX 3080), AMD needs to scale higher than double NAVI 10 aka Big NAVI.
    1
  2. AMD is aware of hardware RT and RTX IO (DirectStorage) which are included with XSX.
    1
  3.  @outsidefactor  The problem with RDNA 2 is scaling. GA-102 (RTX 3090 and RTX 3080) may need three scaled NAVI 10. CDNA has 128 CU, hence AMD needs to combine CDNA scaling with RDNA 2 features. Big NAVI with 80 CU at 2Ghz yields 20.4 TFLOPS which is RTX 3070 level. .
    1
  4.  @denverbasshead  RTX IO(aka DirectStorage API) also exists with Xbox Series X
    1
  5.  @outsidefactor  Big Navi was based on rumors. XSX GPU has two Shader Engines with a total of 26 DCU (56 CU) 5700 XT has two Shader Engines with a total of 20 DCU (40 CU) Based on history, R9-290X has four Shader Engines which is 2X scale from Radeon HD 7870/R9-270X's two Shader Engines design. From XSX GPU's two Shader Engines design, Big NAVI's four Shader Engines could scale to 112 CU which could yield 30 TFLOPS at 2.1Ghz RTX 3080 (GA-102) can be targeted with XSX GPU 2X scale. Both RTX 3080 and RTX 3090 uses GA-102 chip design.
    1