| DonFerrari said: Understood. But let's say that if you increase the number of CUs it also makes senso to increase bandwidth to feed them right? Or would you say design would chose more CUs just to cut down the clockrate and consumption but keep the same general processing capacity? |
Depends on the workload and how memory bandwidth starved the hardware was to begin with.
FuryX has more bandwidth than Vega 64 for example (512GB/s vs 483GB/s) but Vega 64 beats it every day of the week.
Sometimes a significant cut-down in bandwidth has a negligible performance hit.
With that in mind, more bandwidth is generally better.
Microsoft/Sony/AMD will have the data available to work out their bandwidth needs for their hardware anyway, so they will choose whatever offers the best price/performance.
| KingofTrolls said: To a degree I guess. We are not talking about Nintendo, also Xbox One shows that power matters. I think the next gen will be a 180 turn in console philosophy, but we shall see. |
The Xbox One X isn't even using remotely high-end hardware.
--::{PC Gaming Master Race}::--
