While AMD’s Financial Analyst Day is at the start centered on the corporate’s monetary efficiency – it’s proper there within the title – this doesn’t cease the corporate from dropping a nugget or two of technical info alongside the way in which, to assist excite traders on the way forward for the corporate.
One such nugget this 12 months entails AMD’s forthcoming RDNA 2 household of shopper GPUs. The successor to the present RDNA (1) “Navi” household, RDNA 2 has been on AMD’s roadmap since final 12 months. And it’s been beforehand revealed that, amongst different issues, it will likely be the GPU structure utilized in Microsoft’s forthcoming Xbox Series X gaming console. And whereas we’re nonetheless a while off from a full structure reveal from AMD, the corporate is providing just some extra particulars on the structure.
First and foremost, RDNA 2 is when AMD will fill out the remainder of its shopper product stack, with their eye firmly on (lastly) addressing the high-end, excessive efficiency phase of the market. The excessive excessive finish of the market is small in quantity, nevertheless it’s inconceivable to overstate how vital it’s to be seen there – to be seen as competing with the perfect of the perfect from different GPU distributors. While AMD isn’t speaking about particular SKUs or efficiency metrics presently, RDNA 2 will embrace GPUs that tackle this portion of the market, with AMD aiming for the efficiency essential to ship “uncompromising” 4K gaming.
But don’t name it “Big Navi”. RDNA 2 isn’t only a collection of bigger-than-RDNA (1) chips. The GPUs, which would be the codename “Navi 2X” household, additionally incorporate new graphics options that set them aside from earlier merchandise. AMD isn’t being exhaustive right here – and certainly they’re largely already confirming what we all know from the Xbox Series X announcement – however {hardware} ray tracing in addition to variable fee shading are on faucet for RDNA 2. This stands to be vital for AMD at a number of ranges, not the least of which is closing the present function hole with arch-rival NVIDIA.
And AMD didn’t cease there, both. Even to my very own shock, AMD isn’t simply doing RDNA (1) with extra options; RDNA 2 will even ship on perf-per-watt enhancements. All advised, AMD is aiming for a 50% enhance in perf-per-watt over RDNA (1), which is on par with the enhancements that RDNA (1) delivered final 12 months. Again talking at a excessive stage, these effectivity enhancements will come from a number of areas, together with microarchitectural enhancements (AMD even lists improved IPC right here), in addition to optimizations to bodily routing and unspecified logic enhancements to “reduce complexity and switching power.”
Process nodes will even play some type of a task right here. While AMD remains to be going to be on a 7nm course of right here – and they’re distancing themselves from saying that they’re going to be utilizing TSMC’s EUV-based “N7+” node – the corporate has clarified that they are going to be utilizing an enhanced model of 7nm. To what extent these enhancements are we aren’t positive (presumably utilizing TSMC’s N7P?), however AMD received’t be standing nonetheless on course of tech.
This robust deal with perf-per-watt, in flip, will likely be a key element of how AMD can launch itself again into being a totally viable, top-to-bottom competitor with NVIDIA. While AMD is already typically at parity with NVIDIA right here, a part of that benefit comes from an atypical benefit in manufacturing nodes that AMD can’t depend on preserving. NVIDIA isn’t standing nonetheless for 2020, and neither can AMD. Improving energy effectivity for RDNA 2 (and past) will likely be important for convincingly beating NVIDIA.
Overall, AMD has vital ambitions with RDNA 2, and it reveals. The structure would be the cornerstone of a era of consoles, and it will likely be AMD’s first actual shot in…
