| Trumpstyle said: A single CCX means 4 core, 8 threads, it's just too little performance and die space saving is maybe 30-40mm2. |
Well. No it doesn't.
A single CCX is 44mm2. - That doesn't include other logic, additional caches, memory controllers and so on.
AMD's 8-core Jaguar by comparison is 24.8mm2.
Zen 2 may also increase CCX core counts to 6 or 8.
And as someone who has a Ryzen 2700u Notebook... A quad core Ryzen is actually pretty capable anyway, certainly beats the crap out of an 8-core Jaguar.
I standby the single CCX comment, because it makes the best price/performance sense... These are consoles remember, not high-end PC's, so we need to keep expectations in check.
At the end of the day... Every mm2 that is dedicated to the CPU, is mm2 that is not dedicated to the GPU. - The component that draws all the pretty pictures, ends up being used for advertising, trailers and showing off a platform.
| Trumpstyle said: It's just no point to go so cheap on the CPU. I expect 2 CCX with maybe 2 cores deactive for better yields, but I'm betting 8 cores for Sony as they need it for easy backwards compatibility. |
You don't need 8 cores for backwards compatibility. - I suggest you read up on how threading and scheduling works.
| Trumpstyle said:
|
The thing with rumors... Is that they are just that. Rumors. They can be discarded or ignored until something solid comes along.
And just because someone is right about a single prediction, doesn't mean all their prior predictions or future predictions will be correct either... You need to judge each and every claim on an individual basis.
Navi with 40CU being equivalent to Vega 56 is nothing special at the end of the day. - Maybe the performance boost is because AMD finally got Draw Stream Rasterization to actually work? And possibly made Primitive Shaders a proper thing that isn't simply relegated to an API? Either-way, it's AMD playing catch up as nVidia has them soundly beat.
| Trumpstyle said: It's very easy to make big improvement to GPU cores, take a look at Mali-g72 and Mali-g76 they manage to double the gpu performance per core. |
Just don't go expecting a multiples increase in performance when AMD isn't building a new chip from the ground up.
People hyped Polaris... People hyped Vega... People hyped Fury... And at the end of the day, they were all let downs in retrospect.
As for ARM's G76 and G72... There is reason why they managed to achieve what they did. - They made some extremely fundamental changes to the uArch. - Each GPU core is twice as wide so it can do twice as much work. - Graphics are a highly parallel task... So you can either increase your GPU cores width to do twice the work, or double the amount of cores. - End of the day, you potentially achieve the same outcome. (G72 was a 4-wide SIMD, G76 is 8-wide.)
However... AMD's GPU's are already wide cores, rather than ARM's typically narrow cores, so it doesn't make sense for AMD to take a similar approach anyway.
--::{PC Gaming Master Race}::--
