Ask the green GeForce team and it will say the Navi-spoiling Nvidia RTX 2070 Super and its RTX 2060 Super graphics card brethren aren’t a reaction to the AMD RX 5700 XT and its own little RX 5700 sibling. It will point to the fact that you can’t just release a range of new graphics cards in a couple of months as evidence that the new Navi RDNA GPUs have not been any sort of catalyst for these bizarrely suffixed GeForce cards.
And Nvidia is absolutely right. But though the Turing RTX refresh feels about due – and will have doubtless been long-planned to arrive around a year after the first cards hit the shelves – the final form, the GPUs picked for its vanguard, and most especially the launch timing, all point to a set of graphics cards trained specifically to counter the new AMD Radeon Navi generation.
That’s the problem with pre-announcing targets to your enemy; it gives them the chance to move the target, leaving you to fire your best shots into the empty ground where they once stood. And if you’ve put everything into your opening salvo that can leave you wide open to a savage counter-attack.
Or just maybe you give them a chance to fire first. There’s a reason Nvidia opted not to peg its review embargo to the actual on-shelf date, and that’s to give it the chance to control the conversation around the July 7 launch of the competition’s new GPU architecture by flanking it with a review announcement and a final release. The early reviews were the initial sucker punch before the performance of the RX 5700 XT and RX 5700 could be revealed, before the GPUs actually go on sale, and before the Nvidia Super cards themselves hit the shops on July 9.
But then Nvidia has the RTX 2080 Super big guns in reserve to lay waste to whatever’s left sticking its head up above the trenches once the RX 5700-series and the RTX 2070 and 2060 Super cards have finished duking it out too.
Sorry. I’ve been listening to a lot of Dan Carlin recently and I think that might have coloured my references somewhat…
Whatever. The Nvidia RTX 2070 Super is on its way and a mighty fine new card it is too. And in many ways it’s pretty much all we wanted from the RTX 2080 when it first launched last year. And that’s finally something you can say about almost the entire Turing stack now that it’s been rendered all Super. And shiny. Far too shiny.
The new RTX 2070 Super is an altogether very familiar graphics card. The GPU technology inside is exactly the same as the 12nm silicon that went into the first Turing GPUs last year. If you want to get down and dirty with the architectural details of the GeForce RTX cards you can check out our full breakdown of the Nvidia Turing GPU design.
But despite that technological familiarity this new card is still rather different to the non-Super version of the RTX 2070, and that’s because it’s using an entirely different GPU. Where the OG RTX 2070 used the full TU106 silicon this Super version comes with the same TU104 GPU the RTX 2080 first launched with.
Which makes it a closer relative to that higher-spec card than its 2070 forebear. To all intents and purposes this is an RTX 2080 Lite… though that would be a lot tougher to market.
|RTX 2070 Super||RTX 2070||RTX 2080||GTX 1080 Ti|
|Memory||8GB GDDR6||8GB GDDR6||8GB GDDR6||11GB GDDR5X|
As such we’re looking at 40 SMs, spread across either five or six different graphics processing clusters, depending on the GPU (where the TU106 was split across three), which gives the RTX 2070 Super a total of 2,560 CUDA cores. With that you get more Tensor cores and RT cores, which all help squeeze just a little bit more ray tracing grunt out of the new third-tier Turing GPU.
It’s also been given a slight clock speed bump too, shifting from a boost clock of 1,620MHz on the RTX 2070 to 1,770MHz with the Super version. Incidentally, that’s also a higher boost clock than the standard RTX 2080 had at launch too. Though, in the heat of gaming the RTX 2070 Super and the OG RTX 2080 end up knocking around the same sort of GPU clock… which is always higher than the rated boost clock anyways.
Nvidia always likes it to look like you’re getting that little bit more from your GPU than you might have expected.
The memory side of things hasn’t changed, however. While the RTX 2080 Super gets a faster roll of the GDDR6 dice at 15.5Gbps, the RTX 2070 Super sticks on 14Gbps with its 8GB GDDR6 memory.
This is exactly what we wanted from the RTX 2070 when it launched last year, and honestly it’s kind of what we had expected. No-one thought Nvidia would release three different RTX cards – the RTX 2080 Ti, RTX 2080, and RTX 2070 – with three different GPUs, but that’s what happened. Even at the Gamescom reveal I assumed the RTX 2070 used a cut-down version of the TU104 GPU until I sat down with Nvidia’s (now Intel’s) Tom Petersen.
But now the RTX 2070 Super is doing the business and really punching above its weight in the way it maybe always should have done. When Nvidia shifted the GPU stacks up, so the RTX 2070 existed in the same pricing category as the GTX 1080 before it, it was disappointing to see it performing largely in same frame rate category too. With the RTX 2070 Super now either nailing the same performance as the mighty GTX 1080 Ti – sometimes even beating it – it’s become the card we always wanted. Yes, the $499 RTX 2070 Super is capable of beating the $699 GTX 1080 Ti… imagine what the new overclocked versions are going to do.
It’s also, therefore, incredibly close to the straight gaming performance of the standard RTX 2080 too. Which would be a genuine concern for the old guard if we didn’t already know that it’s also getting a Super edition of its very own on July 23. With the current performance of the RTX 2070 Super the OG 2080 has effectively been rendered obsolete. Less than a year after it was released.
And what of AMD? Well, things have changed a fair bit since the Super cards first launched, most notably team red slashing the prices of its Navi-based GPUs before they’d even been released. Yes, the Super cards’ performance, and the changes wrought by their own pricing, forced AMD to change what it was hoping to sell its cards for.
While that’s not a performance thing in and of itself, it does have a bearing because it changes the Navi cards’ targeting. With the RX 5700 XT now costing $399 it’s significantly cheaper than the RTX 2070 Super and yet in some instances the Nvidia card is less than 10% quicker.
Though it is definitely quicker. In gaming terms the new GeForce GPU still has the edge, which is lucky considering the new price delta.
It is also, evidently, a far more efficient architecture too, even with Nvidia’s 12nm lithography playing AMD’s 7nm. In order to get as much performance as possible out of the reference RX 5700 XT its clocks have been pushed to the edge and that means it maxes out its temps too. Despite being a quicker, and incidentally much bigger GPU, the RTX 2070 Super consumes just 10W more peak power, and runs nearly 20°C cooler.
The new RTX 2070 Super is a great GPU. Super in fact… though I am still uncomfortable with that suffix. We are talking about a high-priced graphics card here, sure, but the price/performance ratio is where it should be considering it’s upstaging a $699 card from the top of the Pascal tech tree.
It has taken almost a year, but Nvidia’s RTX Turing range has finally coalesced into the form we always wanted it to be in. And the form it could, and maybe should have been in for the launch back in September of 2018. All it’s taken is the promise of potential pressure from AMD’s Navi graphics cards and a year of solid 12nm GPU yields. Though the fact that something as ephemeral as some vague Radeon performance predictions has been the effective catalyst is maybe a little frustrating.
Realistically Nvidia could have stacked the first flush of Turing cards in the same way, offering this level of performance at the same price points, and without the ugly Founders Edition tax too. But that’s what happens when there’s no competition, there was nothing from AMD at the time to encourage Nvidia to do anything differently, and so it just ended up running plays from the classic capitalist business playbook.
But now Turing feels right. The bumped pricing tiers – where the RTX 2070 previously sat in the same category as the GTX 1080, but without the necessary performance increases – now make sense. If you’d had a GTX 1080 before card what did you get from spending the same cash on an original RTX 2070? Maybe a couple of extra frames per second here or there.
Now the RTX 2070 Super is a card that can lay claim to the old gaming crown the revered GTX 1080 Ti once wore it’s a far more desirable product. This is now the card GTX 1080 owners should consider upgrading to, and that’s as it should be from generation-to-generation.
The AMD competition is gaining ground, however, and the RX 5700 XT is a card that’s really not that far off the pace of the $100 more expensive Nvidia card. You’d still want to have the RTX 2070 Super over the Navi GPU if you’re chasing performance, but the sudden price cut has made AMD’s XT a far more tempting prospect.
Nvidia still has dominion over next-gen real-time ray tracing effects, something that will become more and more relevant over the next 12-18 months, and has some AI smarts in its chips to boot, but in straight gaming terms AMD is coming up on the inside with its own super card.
But, however much I think the ‘Super’ suffix sounds like something a Santa Clara grade school came up with, and however much I dislike the mirrored finish of the super-shiny new shroud, with its slightly off-brand green colouring, the RTX 2070 Super is a fantastic gaming graphics card.
Nvidia RTX 2070 Super
The RTX 2070 Super is the GTX 1080 Ti-beater that we should have had at the launch of the Turing generation last year. The new RX 5700 XT competition, with AMD's sudden price cut, runs the new Nvidia card close, but the GeForce GPU still has the edge. You just have to pay for that privilege.