2016 was a bit of a mixed bag when it came to hardware releases. We had Nvidia push the boundaries of top-end performance with its powerful 10 series graphics cards, AMD shoring up the mid-range with very cost effective and energy efficient Polaris GPUs, but Intel kind of dropped the ball with its Kaby Lake chips. They might be more efficient when it comes to 4K video, but performance wise, they barely raised the bar at all.
However that creates a very interesting opportunity for AMD and the landscape of both CPUs and GPUs, even at the very top end, is ripe for plunder. AMD has the potential to make a big comeback in all sorts of exciting ways and if history tells us anything, when there's competition in the hardware game, we all win.
So what can we expect over the next 12 months?
The first big release we're waiting on tenterhooks for in 2017, is AMD's upcoming Ryzen CPUs. Originally termed Zen, these powerful processors could be the first AMD chips to legitimately challenge Intel's top-end performance in over a decade. We could be stepping back in time to the Pentium 4 vs AMD 64 era and we couldn't be more excited.
Ryzen CPUs have shown themselves as capable as Intel's i7 processors when matched clock for clock, but in more recent tests they even outstripped certain Intel chips and did so with a lower power draw too. They support AMD's version of hyperthreading, so have cores and threads to spare, automated overclocking, thermal controls that mean you're never using more power than you need and AMD is even talking about being hotly competitive on pricing too.
While that would be par for the course with AMD CPUs, it's been a long, long time since AMD had an advantages in performance, efficiency AND pricing.
While we will of course reserve judgement until we get our hands on these chips, we're quietly confident that AMD has something real and tangible to show us this time and not a big disappointment like previously hyped AMD CPUs have been. We're looking at you Bulldozer.
Expect the first Zen CPUs to show up in the first quarter of the year.
But even if AMD does end up disappointing us with its central processors, it will have a chance to redeem itself, as it is also looking to make a big splash with new graphical hardware. Acting as a sequel to the Polaris 10 and 11 GPUs it debuted in 2016, in 2017 we have what's known as Vega and it is not aiming to be mainstream.
While there will be entry level and mid-tier iterations, Vega is swinging for the fences and looking to challenge the graphical might of Nvidia head on. Set to support High Bandwidth Memory 2 (HBM2), feature more than double the instructions per clock than high end Fury GPUs, higher clock speeds and a much larger memory address space, early estimates suggest Vega could potentially be more powerful than a GTX 1080 and perhaps even a Pascal Titan X.
If so, once again that would be big news for AMD, though it has arguably been more competitive with its graphics chips over the years than it has with its CPUs.
There are questions that still remain about the potential for AMD's Vega chips. Can AMD secure enough of a supply of HBM2? It previously ran into issues of supply with new memory standards when it used HBM on its Fury cards in 2015. However so far AMD is staying confident and has promised improved performance using modern graphical APIs like DirectX 12 and Vulkan, areas where it has previously held an advantage over Nvidia, despite its hardware being typically slower overall.
Early showings of the technology have seen games like Star Wars Battlefront playing at 4K resolution on Ultra settings, with a smooth 60 frames per second or more.
TDP is expected to creep close to, but not exceed 300w, with over 4,000 of AMD's graphics core next compute units on each chip. That will be combined with up to 16GB of HBM2 if rumors are to be believed, which would give Nvidia something to worry about for sure.
Vega is expected to debut for the first time at at pre-Computex event in May.
But that was a lot of discussion about AMD. How will Nvidia and Intel fire back in 2017?
Nvidia is expected to refresh its Pascal line up with a number of new cards, adding more GDDR5X – which is fast, but not a patch on HBM2 – and higher clockers. We're expecting a GTX 1080Ti, or possibly a 2080 if Nvidia chooses that naming convention.
Either way though, they will come fitted using higher-clocked Pascal cores, rather than some brand new architecture. Nvidia's next-generation Volta chips are not expected to show up until 2018, which could put Nvidia on the back foot if AMD really leapfrogs ahead with Vega. However more likely we'll see AMD simply be very competitive with whatever Nvidia has to show, even if the red team does pull ahead dramatically in DX12 and higher-resolution scenarios.
Intel has only recently released Kaby Lake, its 7th generation architecture, so is unlikely to have any real response to AMD if Zen does prove as competitive as we're hoping. For that, we'll need to wait until the end of 2017, where Intel's recently delayed Cannonlake, 8th generation CPUs should come to bear.
Although Cannonlake is designed as a die shrink to 10nm, problems with yield will see high-end processors confined to 14nm versions, known as Coffee Lake. They will feature the same architectural improvements as Cannonlake, but not the benefits of a die shrink.
These chips are simply getting too small to manufacturer practically in large numbers due to physical imperfections with silicon. We may well be edging towards the end of Moore's law.
But laws aside, this creates a great opportunity for AMD. It may not only be able to get a six month lead in on Intel, but will have a strong contender even when Intel does release its counter.
Much of what we're hoping for here is speculative and based around AMD marketing, so do take it with a pinch of salt, but this level of quiet confidence from many sectors so close to the launch of new AMD hardware, has us very excited for 2017.
