TechShot: Big Navi. How big is it?
AMD's Radeon RX 6000 Series is here and it's causing quite the commotion
Big Navi was officially announced by AMD last week, and oh boy was it BIG... though, not quite as literally as my hilarious colleagues would have you believe...
Boasting an array of new improvements at the architectural level, as well as new software and some impressive looking benchmarks, AMD's 6000 series of GPUs looks as though it was built to be bold. In fact, given how strong of a contender these GPUs could be in the 1440p and 4K market, the RX 6000 series could well be to Radeon what Zen 2 was to Ryzen - which was nothing less than market stirring in the CPU scene earlier last year.
Though the launch event was only around half an hour, AMD sure did pack a whole lot of information into their slides, and certainly the first half was very focused on their engineering achievements. Throw in all the new features on offer with the 6000 series and some benchmark slides with a few subtle differences, and you've got a bit of an information overload on your hands. So today, we'll be taking a quick-stop tour of AMD's livestream and picking out the need-to-know from the spools of marketing speak and pages of footnotes, to get a better idea of just how exactly these graphics cards could stack up...
Architecture, Architecture, Architecture - A Graphics Card for Gamers
It should come as no surprise to anyone that a launch event entitled "Where Gaming Begins," would be one focused on gaming. In recent years especially, AMD have been trying to make gamers feel like it's all about them; and this announcement was no different. What was different though, was just how hard that AMD were doubling down on the gaming scene this time, highlighting that the RX 6000 series cards were designed for gamers, first and foremost.
One of the biggest contributors to this claim comes from the new architecture AMD have used in Big Navi, which is likely why they spent so long explaining it. But to keep it brief and to the point, the architectural engineering is quite the marvel, and has a vital hand in improving their GPU performance.
In essence, Big Navi consists of two key elements which greatly improve the efficiency of the GPU's raw processing capabilities. Carrying over a great deal of what they learnt from their initial RDNA architecture seen in the RX 5000 series, AMD have fine-tuned this architecture even further; by shuffling around some of the design, and tweaking how some of the resource management works, they've essentially brought everything closer together, and then added the ability to process Real-time Ray Tracing and Variable Rate Shading. Drawing some inspiration from their Zen 2 and Zen 3 processors no doubt, AMD have also integrated a large cache specifically for the GPU, which they've dubbed the Infinity Cache.
All three cards on show will also feature a whopping 16GB of VRAM, and will also have much higher clock speeds than AMD's previous generation of graphics cards, hitting figures as high as 2250MHz.
What all of this means in layman's terms is that data processed in the GPU gets where it needs to be faster. Which is not only more energy efficient, but also reduces latency between communicating components. This translates into what AMD are claiming is up to 65% more performance over their previous generation of graphics cards. Not too shabby at all...
Fancy New Features - Potent or Pointless?
In addition to an improved architecture, AMD told us that their hardware would now have full support for DirectX12 Ultimate features (i.e. the Real-time Ray Tracing and Variable Rate Shading mentioned above), and gave us a run down on a variety of other additional features they have already implemented or are planning to expand further with their FidelityFX.
One of the notable mentions here was their intention to implement a "Super Resolution feature," which sounds as though it could be a similar technology to NVIDIA's DLSS, so that could be something worth keeping an eye out for. In fact, given the results of some early scores in some leaked 3DMark benchmarks, the 6000 series cards could be demonstrating Ray-Tracing capabilities that out-shine the new RTX 3070 cards - beating the RTX 3070 with DLSS enabled, despite using no Super Resolution feature, and having 20% less powerful Ray Accelerator cores (think similar to NVIDIA's RT cores). So goodness knows how far that potential lead could widen once they've got a DLSS of their own...
Either way, it would seem AMD have really broadened their feature-set with these new GPUs, bringing them ever closer to par with (and potentially even surpassing) NVIDIA. And that's good news for everyone... well, except maybe NVIDIA.
Besides their FidelityFX, most of which probably won't be of concern to many gamers, there was also mention of two brand new features.
The first of these features is "Rage Mode" (because... I don't know, they're "down with the kids" maybe?) which offers one-click overclocking to enthusiast gamers, and those looking to get every last drop of performance for their money. And from early reports and speculations, it really is the last few drops, with many expecting around 1-2% additional performance.
But more interestingly (and far less ridiculously named), AMD also announced their new SAM feature, otherwise known as "Smart Access Memory." In the simplest sense, this feature works by allowing any RX 6000 series GPU to communicate directly with any Ryzen 5000 series processor (so long as both are mounted in a 500 series motherboard), by allowing the unrestricted transfer of specific data types between them, making it more easily accessed, and in greater quantities. What this means in real-world speak, is up to somewhere in the region of 10% more performance pretty much across the board (provided the applications in question take advantage of the feature, which currently, many don't).
This has certainly raised a few eyebrows in tech communities, with a handful of folks citing an infamous, fruit-centric tech giant who offers many brand-specific features, optimisations, and so-on. But the key difference here is that SAM is nothing new; it's just an AMD rebrand and utilisation of something called PCIe Resizable BAR. Windows 10 has supported it for a while and arguably, so long as the relevant hardware and software support it too, then anyone could take advantage. So let's just see how it goes, and maybe in the next generation of GPUs there'll be some kind of new-fangled "NVMemory" touted at launch... and let's not forget Intel with their new GFX next year either; maybe they'll be vaunting some sort of "Intel 11th Gen Intelligent Divergent Graphics Memory" to complement their new "i9 11950KF"...
Something else I felt was an interesting take-away from all of these new features, was the level of focus AMD has seemingly put on the software side of their GPU launch. The reason this is interesting to me, is that it could put a lot of potential buyers at ease. Why? Well, given their history, Radeon Graphics haven't exactly been known for having the most stable of drivers, and the RX 5000 series cards were no exception the last time round. They did eventually sort out a lot of issues, but by the time AMD had managed to iron out these kinks, the damage was already done, and even then, there were still reports of people experiencing the issues they'd had from the get-go.
But with all of these new features added, hardware compatibility with DX12 Ultimate, and apparent collaboration with a variety of game devs, as well as being the driving force behind the new console hardware, if the guys at Radeon have brought all of this together and taken onboard the hard lessons learned from the RX 5000 series, the drivers for the 6000 series cards could well be working as intended. We'll have to wait and see to be sure of course, but AMD certainly had quite a bit of confidence in their new features, which one would hope is founded on the knowledge that this time round they've nailed the drivers.
Preliminary Benchmarks - will the new kings of 1440p and 4K wear red-jewelled crowns?
This, as always, is a tricky one to call just yet. Much like in my coverage of the RTX 3070 the other week, ultimately, we will have to wait for the "unfavourable" third-party benchmarks to arrive, in order to really understand just how well these cards will perform. But from what we've been shown so far, it looks like AMD could well be trading blows with NVIDIA's high-end offerings.
Now, even without the real-world benchmarks, this is still a point of contention among many, and I would agree. At first look, it seems that across the board, the 6800, 6800XT and 6900XT all eek out the 30 series equivalents in both 1440p and 4K, but not in every title tested.
When you look at the individual game results in the provided graphs, you'll notice that AMD are pretty much tied in most titles and at both resolutions. And where there are wins, the performance increases are as minor as the handful of loses. The only exception to this rule, however, is the RX 6800 at 1440p, versus the RTX 2080Ti (which is essentially an RTX 3070, in light of real-world benchmarks). Here, AMD have a clear win, with a lead over every title tested, with some more significant than others.
But it is also here that we see the subtle differences I mentioned right back in the introduction to this piece. Every benchmark slide is slightly different to the next, in that a handful were indicated as having used SAM in their testing, whilst others employed both SAM and Rage Mode, as well as a select few which used neither at all. This makes it a bit trickier to visualise where these cards actually sit in terms of their baseline performance (i.e. without either features used), and as such, where small victories were made, these could well be ties or losses when software advantages aren't being utilised.
That said, for those who'd have been in the market for a Zen 3 processor anyway, it's still going to be of benefit, and for free ultimately. Plus, given that we were told SAM and Rage Mode combined can only offer around 10% additional performance on average, even if you wanted to remain platform agnostic, these cards would still be pretty much on par with NVIDIA and offering a fairly comprehensive feature-set (for gamers at least), which hasn't happened for a very long time.
And of course, what's a few percent here and there if the value proposition is strong?
Pricing and Availability
Not just trading blows with performance, but pricing too, AMD really have come in swinging for this launch. The RX 6800, 6800XT and 6900XT will be retailing at $579, $649 and $999 MSRP respectively, marginally undercutting NVIDIA's 30 series cards. Once again, the RX 6800 will be the exception here though, with that model actually being $80 more expensive than the RTX 3070, though offering more performance too (if AMD's benchmarks are anything to go by).
I've said it before and I'll say it again though, we'll need to wait for the actual benchmarks before we can start weighing up the actual value on offer here, as well as see what these cards actually retail at; though it would seem AMD have taken steps towards avoiding any availability issues by addressing bots, scalpers and launch stock levels themselves, there's still no guarantee that they won't all sell out instantly, nor that they'll be retailing at that MSRP.
But whatever happens, it'll only get even better for the consumer
Even if these cards underperform, or retail for a little more than those MSRPs, either way, this is all a good thing for the consumer. More cards mean more choice, with more price-points to pick from. Plus, if these cards really do give NVIDIA a run for their money, then we may well see even more cards hit the market, or else a war on driver optimisations and updates to offer more performance, and even price undercutting from either side.
So like I've been saying in various posts this year; sit tight, wait out the shortages and new launches, and hopefully in a few months' time, maybe you'll have one hell of a new gaming PC in your hands?
That's all from me today though, so let me know your thoughts in the comments. Excited for Big Navi? Waiting for NVIDIA to throw out a few Super/Ti variants? Building a new system from scratch or just upgrading? We love to hear from you, so hit us up down below.
