one thing i am noticing is alot of people trash talking AMD for re-branding, whilst claiming Nvidia have never pulled such shenanigans
yeah i dont know why people are flipping there sh!t. This happens all the time both both sides. this is nothing new. New flagship cards are new tech most gens. New mid-range cards are re-branded flagship cards from last gen half the time. Its always been like this.
Think people after all this time expected the 390x to add actually be the answer to nvidias price gouging instead it seems like Fiji will be same cost $999 or whatever it comes in at is a lot more than most will pay no matter how it performs. Even if it matches the Titan in performance and price there will be a lot of disappointed fans who wanted AMD to undercut them hard. Think people expected AMD to be more price competitive and offer alternatives to the 980ti at a price point similar. The expectation at the minute is the 390x will be struggling to beat the 980 let alone perform close to the 980ti. Furi and Titan are just so far out of most people's price range for a gpu. It also makes the 295 very good value whilst still in stock if you can stomach duel gpu. The likely hood is it will still be faster than the highest single gpu card at a 3rd of the price. Rebrands are nothing new but 1 new gpu to market after all this time and still no confirmed release date for its true next gen card is a worrying place to be.
I've not looked into it much as it doesn't bother me, but isn't the reason people are a little miffed because of the amount of rebranding being done, isn't the only truly new chip the Radeon Fury? The rest of the "new" cards are using either the same fab or architecture as the last two generation of cards, all that seems to have been done are a few minor tweaks. ICBA to look into it but has Nvidia re-rebranded the same fab or architecture for three years on the trot in the past?
It's not that nVidia don't do it, it's just that AMD seem to be doing a lot of it just now... and what the GPU industry needs is new stuff from AMD, not yet another rebrand.
actually the GTX 280M was also the G92 chip - so that's 4 rebrands bare in mind 1 thing - Nvidia has nothing using the GM chips and 9 series branding for under £140 and mid range btw a tear down for you: http://www.eteknix.com/radeon-r9-390x-teardown-reveals-r9-290x/
AMD has pretty much done all they want to or can do with 28nm and now appear to be focusing on 20nm. The problem is, 20nm is new, so the yields are going to be lower and the process more expensive. I suspect that AMD may have done a minor respin of the 28nm stuff (otherwise what's the point of the NDA?), but the time spent on a major respin would probably be better spent preparing to move the mid range cards to 20nm when the yield/cost makes sense. In any event, from a marketing standpoint, we @ bit-tech know it's BS, but we're not the target market of the rebrand efforts (OEM). Maybe as protest bit-tech should enforce a new policy in sigs such as: Tahiti 1792:112:32 3GB @925/1250mhz ?
The big 3 fabs sorta went past it already could have something to do with it. TSMC is on 16nm Samsung is on 14nm and global foundries is on 16nm. All 3 are trying to get Apples business for chip development.
28 to 16 should offer more than 28-20 could also be part of AMDs reasoning. Or it could simply be price that they are getting or availability. Too many unknowns in that one enough neither Nvidia or AMD are big enough to be able to demand much of anything from the big 3 foundries ( ignoring Intel here who will not sell AMD or anyone else fab space) Once Qualcomm and Apple / Samsung have had there demands fulfilled the bits that left go up for grabs seems to be the general consensus. You could buy AMD and Nvidia for what Samsung or Apple spends alone in chip purchasing on a quarterly basis let alone yearly.
A SOC isn't a GPU, I can't find more details ATM but I would assume the 20nm part is the ARM CPU and the Maxwell GPU is using 28nm.
http://wccftech.com/nvidia-tegra-x1...res-maxwell-core-architecture-256-cuda-cores/ the chart lower down says Maxwell on X1 is on 20nm
That's referring to the ARM CPU part of the SOC (afaik) the GPU part uses the Maxwell 2 GPU fabricated on 28nm, then the two cores along with the other gubbins are brought together in the SOC. I'm fairly certain a SOC isn't fabricated, it's made by bringing together lots of different components and building all those into a single package.