Carzy Zarx’s PC Gaming Emporium - Catch Up on All the Latest PC Gaming Related News

[Jizz_Beard_thePirate]

JEMC said:

Jizz_Beard_thePirate said:

Honestly Intel should stick to sub $250 gpu class. They still have plenty of driver issues and cpu overhead isn't doing them any favors. Being at sub $250 with 12/10GB of vram where neither Nvidia or Radeon will bother trying to compete means they could have the class all to themselves even at a low profit margin. At the very least, they can gain a customer base that Green and Red team no longer cater to.

True. But if they only do entry level cards (if $200 can be called that), the perception most people will get is that they only make good enough products, and that can harm them in the future if/when they ever try to put a fight in other segments.

It's like people arguing that AMD should always do higher end cards, even if they don't sell, because not competing in that segment damages their image.

The problem with Radeon competing in the high end is they can never seem to consistently do it. They have these one hit wonders and then spend the rest of the successive generations going down in gpu segments.

RDNA 2: Top "202" competitor

RDNA 3: Top "203" competitor

RDNA 4: Mid "203" competitor

Personally instead of Radeon investing in designs that don't seem to work properly like MCM with RDNA 3 and HBM with GCN, they need to figure out what works and scale it up. Otherwise they are just largely wasting R&D into failed projects such as HBM which is no longer used in consumer gpus and by the looks of it, neither will MCM anytime soon or maybe ever if Ai is used to generate frames. Where as with Nvidia, their R&D into Ray Tracing, Tensor Cores and Ai are paying off in multitudes.

I think if they could have scaled RDNA 2 to say 120CUs with 5nm for RDNA 3, probably would have achieved similar if not better results than going MCM. Then invested R&D into their own Tensor cores, Ray Tracing and Ai, RDNA 3 likely would have not only competed against 4090 in Raster instead of 4080 but also in features. But that's just in theory of course where it's easier said than done.