Well the RX 680 would be a refresh of a refresh, as such 15% ain't that bad.
However, really wish Navi would come soon. Probably struggling to get the PC version running as expected. After all, PC games need a monolithic GPU, as they consider a setup like Epyc or Threadripper a Multi-GPU setup. So they possibly had to reconsider late into development and create multiple mask sets instead of just one
I didn't post the new about the Intel CEO because it wasn't "hardware related".
As for the AMD new mobile chips, it doesn't really affects us much as few people here game on a laptop, at least as far as I know. I'm more interested in the new 4-core Ryzen desktop chips.
And regarding the GPUs, maybe AMD should reconsider the way they develop GPUs and go back to a single, scalable architecture, and forget to develop two archs for the mainstream and high-end market. It saddens me to say this, but they seem to lack the resources to do it.
I wonder if PUBG will ever get a retail release on PC, if it does I may end up getting it to try it out.
A physical launch won't do much for them, specially with stores having less and less space for PC games.
I am of two minds about Quad-Cores these days.
In a laptop they can get away with higher clocks... Which is important for single threaded performance... But on the other hand... It's a Quad-Core. In 2018. Yuck.
Give me a Hex and I will think about it, give me an Octo and call me keen, maybe next year when Intel start pushing 10nm.
AMD is doing that. The problem was more that they couldn't afford development of both a complete lineup of chips and their masks. As a result AMD had to develop only partly releases, with the gaps getting bigger and bigger over time. They had to reduce in size between 2008-2017, so AMD couldn't develop both a new CPU and a new GPU in parallel at full steam. So after they developed Bulldozer and GCN, they had to choose which one to develop first. Considering GCN was doing well originally but Bulldozer was pretty much the definition of a development fail, it was clear the CPU would come first. Hence why the later GCN developments were mostly pretty small. Vega was the last one where they had all possible ressources bound by the CPU development
It will still have an effect on Navi, GPU's have extremely long development cycles, so something that happened a few years ago can have an effect today.
Navi is still a derivative of GCN, it's another iterative refinement just like Fury, Vega, Polaris was and so on.
Graphics Core Next is getting old and tired at this point.
Where all bets are off is with AMD's next-gen GPU architecture, which is hopefully what comes after Navi.
Every release after the first gen (Radeon HD 7xxx) had just some new GPUs, and the rest were rebrands. Second Gen was first Bonaire, followed a couple months later with Hawaii. Third gen was Tonga, followed a year later with Fiji. Fourth Gen, Polaris, had 2 chips at first(RX480/470 and RX 460), followed by a very slow one (RX 550) a year later. Vega only has the high-end chip for now apart from the variants in the APUs.
Graphics Core Next is actually an extremely modular design, so what AMD could/did do is make an update to one part of the chip like the geometry engines whilst leaving the rest of the chip alone... And dropping that as a new high-end offering.
So it made sense they would take things in that direction while the company was hemorrhaging cash.
AMD also tried to replicate what they did with RV770 and that was to adopt the latest DRAM technology to get a leg up over the competition... Unfortunately for AMD's sake, nVidia was making their GPU's more efficient by implementing Delta Colour Compression and Tiled Rasterization which allowed for better culling which saves on bandwidth anyway.
AMD "Tried" to catch up with Vega by implemented it's Primitive Shaders and Draw Stream Binning Rasterization, except they either made it so Developers had to opt-in to use the feature or never bothered to implement it in drivers, meaning it was a waste of time from a gamers perspective, maybe with Navi they might fix that shit.
Polaris had those small chips mostly to get rid of their 1./2. gen entry level chips (Oland, Cape Verde and Bonaire), which went through several generations and rebrandings, especially as mobile chips. Especially Oland (which got similarly developed to get rid of the old Terascale chips in the lineup) really had to go, as it couldn't keep up anymore with the APUs, even when bottlenecked, and even Cape verde was getting too slow for them. Btw, when you put a RX460 against Bonaire and a RX 550 against Cape Verde, you can see that AMD was improving GCN quite a lot - just not enough to keep up with NVidia.
They were reserved for OEM's. OEM's need "higher numbers" every year in order to advertise their new products every cycle, even if it's the exact same chip, sometimes they might have a significant inventory of stored chips as well.
RX 550 is the lowest us mere-mortals can get our hands on through regular channels, meaning the entire 500 series is all GCN 4.0 based in public channels.
No one doubts that GCN hasn't seen great strides in efficiency and performance over the years, but compared to nVidia they are just slower at making those incremental improvements.
Ultimately though... I would like AMD to return to it's small-die strategy that they employed with Terascale, it worked and it was great.
With the resources from the CPU architecture development being freed up since Zen got feature complete in early 2016, Navi certainly did profit from the additional work and will probably either come with a full lineup or with radically different chips in PC and server segments (where they can bring the announced scalability without the problems they would have in a PC), and possibly both.
It's not just a freeing up of resources (AMD is hard at work building the next several Zen lineups.) but they can afford to bring in additional resources as they have drastically increased profits.
I'm glad to see AMD doing well enough to actually start hiring again. Hopefully they hire the right people for the right jobs.
And of course Intel will have to face AMD with whatever they had in their roadmap for the time being, but they had to do that no matter the CEO, right? So it doesn't matter who's at the helm as long as he/she doesn't start messing around.
For now, Intel has the 9xxx series, which by the looks of it is just the same processors with a 100/200 MHz increase and the launch of the 8 core part.
Intel actually has a strong Micro-architecture, certainly has an edge over AMD, albeit slight. It's just they aren't offering the core counts to match AMD.
They stagnated and tried to profit as much as they can for as long as they can while AMD fumbled.