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

[EricHiggin]

haxxiy said:

EricHiggin said:

If Navi is solely GCN based, then it's going to have an uphill battle against Pascal even at 7nm, cheap or not. Odds are good though that Navi will be another step passed Vega's GCN/NCU, which was already a step passed Polaris' GCN. Sales of Ryzen have been strong since launch and much of that is probably being spent on the Radeon division. Ryzen pretty much has everything it needs in place, it just needs those pieces put together over time, with tweaks along the way, so AMD can focus more so on Radeon now. If Vega did suffer because of Navi, then this fits into place nicely.

I don't think Keller had much to do with Radeon, as his job was to produce both an x86 and an ARM CPU. He may have worked with Radeon somewhat if they had planned for GPU to be part of infinity fabric back then, but it seems unlikely at this point since AMD has said they weren't sure if it could be done and that they are looking into it. They could be bluffing though. AMD also picked up a Manager and Engineer with plenty of experience and history with Nvidia and ATI to lead Radeon forward.

AMD is using very smart marketing. Not only are they pushing the 7nm 'benefit', but they apparently are going to align Navi with Ryzen so both will be 3000 series which is a fantastic marketing decision. Whoever decided to change the model numbers to that is brilliant, because it will also give casual consumers the idea that the Radeon series of cards will always be ahead or better than the Nvidia counterparts, since you will have Nvidia RTX 2000 vs Radeon RX 3000. If your buying an AMD 3000 CPU, you might as well buy an AMD 3000 GPU. While Intel's 10nm should be on par if not exceeding 7nm, AMD doesn't have to worry about that for a year or two in the consumer market.

To be frank, AMD's graphics division, ever since the beginning of the decade, reminds me of Ferrari in F1. Both are sort of fast, but simply can't keep up with development. Their newer stuff ends up being misguided sidesteps or downgrades, and then they take the "L" home and are back to zero, rinse and repeat the next year.

What I'm expecting, while sincerely hoping to be wrong, is some 7 nm designs who might sort of match Turing's power consumption and performance, which are solidly beaten in turn by Nvidia's next generation GPUs just a few months later. AMD counters with their rebranded and overclocked, more power hungry chips, AMD fans claim it's a victory since "no casual gamer cares about power consumption anyway" and cue Nvidia walking away with 95% of dGPU marketshare again.

Can't argue too much with that, other than the fact that Ferrari is still a Ferrari. People who own Corvettes, typically don't care they don't have a Ferrari or McLaren, etc, since they could never afford one anyway. A Corvette will do everything and more that the majority would expect it to do, so who really needs anything more, other than the elite minority who can afford to professionally race those cars?

You very well could be right, but I think all of us are wondering if some of that Zen magic will rub off on Navi. With RTX launching fairly recently, I think it would be quite unlikely to see the next gen GPU from Nvidia for another 2-3 years. That doesn't mean they won't just get Pascal or Turing on the 7nm process sometime in 2019, so they can drop the price considerably for the 3000 models if they feel they need to better compete. Like you said, Navi could be in the ballpark for a short duration before Nvidia counters and resets the clock. Is Nvidia that worried though and would they bother re-launching on a new node instead of mostly focusing on the future of GPU design?