I agree. I doubt we will see a 2060 that isn't a rebrand if at all. RTX are simply quadro cards and to increase sales they just as well can also release them to consumers. After all the 2080 and 2080 ti are cut-down so they send the defective dies that else go to waste to consumer cards.
As for AMD Polaris refresh, there are 10gbps GDDR5 dies available so a 25% increase over RX580. Don't now about power usage however. Also doubt this is true. Not really much money in it over selling RX580s.
Yeah, I just don't see the need for it. It'd probably cost in line with the 1070, while offering similar performance. So why waste the engineering resources (which Nvidia often tells us, remember when they touted how Pascal cost billions to develop?) when you can achieve that already? Put those towards 7nm.
I don't know that there is 10Gbps GDDR5, that's all GDDR5X as far as I know, which only one company was making and they stopped development (possibly production) for GDDR6. I'd guess GDDR5X wasn't and wouldn't be produced in numbers high enough, so its either they stick with what they have or they move to GDDR6. I disagree. OEM deals would make it worthwhile.
I think you are being very optimistic here. RX 5xx series increased clocks, but also increased power use. It was nothing more than a much more mature chip a year later than the 4xx parts.
Turing, while large adds some large special purpose units like RT and Tensor cores. They don't have to be on all the time. Nvidia is just taking advantage of abundant cheap transistors which at the same time deliver minimal power reduction. A company that's struggling like AMD has an even harder time when things get difficult.
If its really Polaris, and its RX 680, it'll be limited to 2304 SPs and have little clock improvements for 5-10% gain. It's a refresh of a refresh. A GTX 1060 refresh will take care of it. But they'll likely do better, since 2060 is planned.
Did you think I was inferring that the Polaris refresh would compete with the 1070? That's not what I was meaning. I said that Nvidia doesn't need anything in the 2060 market because they can just drop Pascal prices to counter any Polaris refresh. And they wouldn't have to put any resources towards that, plus it'd let the channel move that large stock buildup much easier. A $300 1070 would likely stomp the Polaris refresh in perf/$, so even though it'd be more expensive it'd be worth it for people to spend the extra, especially if AMD tries to sell Polaris refresh at $250 or above. And at $300 a 1070 would almost certainly still make Nvidia money and it's still solid in performance so it'd be popular and keep Nvidia mindshare. And the lower Pascal stuff could be priced to beat the Polaris stuff by simply adjusting prices. Chances are that a 2060 that matches the performance of the 1070 isn't going to be much cheaper to produce than the 1070, and pretty much definitely wouldn't be if it has the RTX stuff at all.
Nvidia's best option would be to just roll with Pascal in the lower priced stuff ($100-400) until they transition to 7nm. It lets the RTX features develop so they can tout real tangible things for them when they launch in the more mainstream/affordable cards. They can start the 7nm with smaller chips to let the process mature (plus that's where AMD is going to compete). And instead of wasting resources on more 12nm chips, they just put that towards moving to 7nm.
Those transistors aren't as cheap as you seem to think or else we wouldn't have seen prices go up like they did. I don't think Nvidia could just move the Pascal from 16nm to their specialized 12nm (whereas for AMD moving from 14nm to 12nm is apparently very easy, because the GF 12nm basically is just their 14nm matured). I'm not sure its worth it, when Nvidia has little to no need to change what they already have there. They can just adjust prices if there is an actual need until they can roll out 7nm products where AMD wouldn't have an advantage that could equalize Nvidia's engineering and software advantage.
Is it? I see people speculating (many of them the same ones that speculated for a year about Volta cards that never came). I'm not saying there's no chance its coming, just that I don't think it makes a lot of sense for Nvidia right now.