Captain_Yuri said:
Yea but it feels like they are going to be sacking features to achieve that. Nvidia GPUs don't use more power or have a bigger die for no reason, they have Tensor Cores and RT cores take up a good chunk of the area that still use power even if you aren't using the features. It's one of the reasons why 6000 series was very power efficient compared to Ampere. RDNA 2 focused a lot on standard cores but basically left out dedicated Ai accelerators and more performant RT cores. Also using G6/TSMC instead of G6X/Samsung helped greatly too but now it's TSMC vs TSMC. So it could be another case of competitive Raster but falls behind in a lot of other areas which would be lame but we will see this Thursday. |
It's not that they are sacking features, it's just that they haven't added things like the Tensor Cores to do part of those features, at least so far. And while DLSS2 and now 3 will still be king, the truth is that for many FRS 2.1 looks almost the same, and it's extremely difficult to notice the difference while in motion.
After all, they don't need to copy all their festures to stay competitive.
There's also another thing to consider with RDNA3, and that is that AMD has, allegedly, managed to reduce the amount of redundant logic in the chip while moving of some others to the extra MCM dies. That, if true, will help AMD with its efficiency, reduce the wasted energy and increasing the useful logic in the chip.
But well, hopefully Thursday we'll know more and have a better idea of where the cards may end in all fronts.
Please excuse my bad English.
Currently gaming on a PC with an i5-4670k@stock (for now), 16Gb RAM 1600 MHz and a GTX 1070
Steam / Live / NNID : jonxiquet Add me if you want, but I'm a single player gamer.