He's right.
nV has 1070 and 1080 right now in their lineup. Maxwell parts are being EOL'd as new Pascal parts fill out the remaining slots in the lineup.
No, discounted 970 isn't competition for the RX 480 although both have similar performance, 980Ti cards are in fire sale because no one would buy that instead of a 1070 or 1080. Once 1060 hits 970/980 are going to be replaced, if rumors are right this will happen in the next week or two.
One has to be a bit insane to buy 28nm hardware this late in the game, so early in the start of 14/16nm class hardware that clearly show tangible benefits over the old hardware. GCN is nearly an industry standard at this point thanks to both present, future consoles and current APIs like DX12/Vulkan, Pascal is more GCN like and seems to handle compute loads better than Maxwell, if one cares for a card's longevity a 970 plus its weird memory segmentation just isn't a viable buy today.
As of now AMD seems to be selling out whatever RX 480 stock they can produce, just like nV and the 1070 and 1080. Different market segments being serviced by different products in different categories. Vega 10/11 fill out nV's 1070/1080 slot on AMD's part, GP106/1060 fills out RX470/480 slot in nV's lineup, all coming in the following weeks and months.
RX460/P11 I don't know, maybe there's a cut down GP106/GP108/GTX1050 around or lower. GP102 as the next Titan/1080Ti, I don't know if Vega can stand up to that.
--------------------------------------
As for the RX480 reference PCB, this reminds me of the R9 290/x reference launch for different reasons. R9 290/x reference PCB is built like a tank and has a
nuclear power station for a VRM as most AMD reference PCBs do, yet the cooling solution was crap for a 300w TBP design. Decently cooled cards arrived months later. Irreversible damage done. Product image tarnished so bad a refresh with decent cooling in the 300 series couldn't completely live down.
The RX480 has a decent cooling solution for the price, providing similar temperatures and noise levels to what nV gets with their reference blower that now carries a $100 founder's edition tax/ripoff, yet we have overvolted dies everywhere with
lots of reports of undervolting WHILE overclocking and keeping temperatures and power in check after that. Forums, AMD's subreddit, it's everywhere. Yields can't be that bad if there are so many reports like that.
Poor factory voltage adjustment that leads to:
- Noisier than necessary product
- Slower than it could be because that extra voltage leads to higher temperatures that keep the GPU off the maximum boost clock at all times.
- Board power peaking higher than 150w
- Weird PCB power design that although having passed PCIe certification doesn't play nice with crappy hardware, and is certainly not helped by the higher than necessary voltage. Actually seems more suited to the RX470.
Image tarnished, not as bad as the R9 290/x, but it could have been better. Now if we have custom boards and cooling coming in the next week or two, we're still on time for some of that tarnish to go away.
Sapphire Nitro model seems decent... a Sapphire rep said it has a more traditional power balancing between the PCIe slot and connectors, has an 8pin power connector and also has decent open air cooling that helps get clocks at maximum boost and
lowers power consumption thanks to lower temperatures, this could be translated to higher clocks and performance just like it happened to the 300 series.
There are reports that some of the fancy power management tech shown in the Polaris slides isn't enabled. I suppose there could be some truth to that, supposedly Polaris 10 calibrates voltage at startup and therefore if there's so much variability in each die produced voltages should be all across the board in all the samples... it doesn't seem to be that way, 1.15v as load voltage seems to be repeated here and there. Supposedly there's a driver fix coming July 5, maybe this just lowers voltage across the board to a more reasonable point for stock clocks, or enables the remaining power management stuff. Who knows. It's all strange.
In spite of that, this time GF could have failed AMD in lots of ways and maybe AMD couldn't resort to other means of maximizing yields on an early process other than overvolting across the board, it's difficult to see where the blame is to be put here. If this were TSMC silicon (one less variable that differs, seeing how nV's products behave on their process) then it would be clear there's something at fault on AMD's design. Maybe P10 silicon starts behaving better in the following batches as the process is fine tuned, maybe P10 is just a pipe cleaner for the real deal (Vega and Zen), who knows.
As a counter example you have Phenom IIs and FXes out there on different processes and architectures that can be undervolted 0.1-0.3v depending on how high stock clocks are pushed on each part while retaining complete stability, so I don't know if this is AMD being AMD as usual, or if there's actually someone somewhere at fault for this.
If I didn't have a 290 Tri-X / 390, sure, I'd buy a RX 480. I'd just wait for custom cards and boards. 1080p60 at such a low price is a great deal. This card is certainly much more affordable where I live (Argentina) compared to more traditional stuff like 970/390
TL;DR should we have custom boards and cooling in the following week or two, all of this can be more or less forgotten and we can get to see P10 as it should have been from the beginning.