So AMD 480 is a dud, not nearly the amazing deal after all

[Pemalite]

vivster said:

Pemalite said:

It's actually NOT a bad card.
It's just priced badly.

If it was priced at $150 for the 4Gb and $200 for the 8Gb rather than starting both at essentially $50 higher than that... Then this would be an entirely different discussion.
Can't forget the early adopter tax on new hardware either which is driving the price up higher.

For future titles the RX 480 will fare better, it's shown to excel in Direct X 12 whilst still providing adequate performance in Direct X 9/10/11.

I think people were thinking this was some kind of 1440P/1600P/4k gaming monster based on TFLOP's alone, which was never AMD's intention... And based on that, had unreal expectations for where the product was going to be situated in AMD's lineup.

If you wanted great performance, Vega was where it was always at. Polaris is mid-range. Navi should shake things up as well.

I think that's cutting a bit too much slack for AMD. They definitely tried to sell it as a high range card for a mid range price. They banked in on the ignorance of the mainstream buyer.

I disagree. I saw the writing on the wall a long time before that, people certainly blew it's expectations up.

It was meant for a good $200 1080P/VR experience and AMD peddled that since it's announcement, people expected more, it was almost a plague on this forum when people expect a 5Tflop GPU to be a 4k gaming chip based on flops alone.

Slimebeast said:

Lafiel said:

I heard rumors the Vega is produced at TSMC in 16nm, like the 10XX Nvidia chips, so maybe it can clock at 1.6GHz+ like those chips

Wow. So does that mean that it's a common opinion now that the TSMC 16nm tech is superior to the 14nm from GlobalFoundries/Samsung?

Because performance from a node can't mature, right? Only yields can mature, but not the tech itself, or? Or can it be improved and tweaked a little bit?

Global Foundries 14nm is *slightly* superior to TSMC 16nm, it has a smaller gate pitch and sram feature size, Global Foundries is just struggling to get acceptable power characteristics at this stage, that will be resolved in time.

And yes, performance from a node can mature, leakage can be reduced which results in less heat for instance.

28nm saw constant improvements. (And still is even today)