Ryuu96 said:
Pemalite said:
Okay. 4096 shaders is Vega 64/Fury X level (64CU) CU counts. 4096 * 2 * 1400 = 11.46 Teraflops.
22GB is 11x memory chips @ 1GB. Probably looking at around 500-600GB/s of bandwidth with those clocks (Can't be arsed doing the calculations), possibly using a crossbar memory controller layout like the Xbox One X, depends on the ROP layout too.
CPU is an 8-core Ryzen without Hyper-threading. L3 cache size is really really really Good @ 32MB.... (AMD EPYC/Threadripper sizes) 1GB of L4 might be an off-chip ESRAM/EDRAM-like implementation.
All of that can change of course.
|
I understood none of that.
Does that all sound good to you?
|
Essentially it's what was expected given AMD's current Graphics capabilities.
To break it down to make it a little bit easier for various folk...
We are looking at roughly double the Graphics performance of the Xbox One X for the GPU.
Roughly 8-10x increase on the CPU, potentially more depending on clocks.
Which means, Anaconda might actually be inferior to the Xbox One X in regards to GPU capability, but superior in terms of CPU capability.