The rumors are true: Nine months after the RTX 2070’s launch and six months after the RTX 2060 arrived, both Nvidia graphics cards are getting retail upgrades. See: the RTX 2060 Super and RTX 2070 Super, which hit stores July 9 for the same price as their previous numbered versions ($399 and $499 MSRP, respectively).
These will be followed on July 23 by the $699 RTX 2080 Super. Nvidia will continue to sell the RTX 2080 Ti ($999) and RTX 2060 (discounted to $349), while the standard 2070 and 2080 graphics cards are already being phased out by Nvidia. (If you’re wondering, AMD also has a big launch during this calendar period: the Radeon RX 5700 series, coming July 7.)
We’ll have to wait and see what to expect from the 2080 Super, whose review hardware hasn’t made it to the shops yet. In the meantime, a last-minute shipment gave us a chance to put the RTX 2070 Super through a few tests ahead of today’s announcement. That helped us confirm some good news: Nvidia finally got an RTX card up to the performance point that was worth a $500 investment (and frankly, the card is now neck and neck with the original RTX 2080).
More power, not necessarily more efficient
For this new Super line, Nvidia went back to the original RTX cards and packaged them with more compute cores (“CUDA cores,” in Nvidia’s parlance), more texture units, more streaming multiprocessors (SMs), and a higher base and ” boost” clock speeds. In the case of the 2060 Super, the VRAM count has jumped to 8GB, catching up with its siblings’ 256-bit memory interface, while jumping to 448GB/s memory bandwidth. The Super version of the 2080 gets a slight increase in memory bandwidth; the 2070 Super gets no memory-related boosts.
The main drawback of the Super line is that it in no way indicates a new or more efficient process. While Nvidia prides itself on the “performance per watt” measure of the Turing line, you still need more wattage for each Super card than the non-Super version. It’s not a ridiculous amount, but be warned if you’re already on the brink with your PSU of choice. (There’s one exception: the RTX 2080 Super advertises the same power consumption as its non-Super variant. We’re excited to see that in action.)
|RTX 2080 Ti||RTX 2080 Super||RTX 2080||RTX 2070 Super||RTX2070||RTX 2060 Super||RTX 2060||GTX 1080 Ti|
|Memory bus width||352-bit||256-bit||256-bit||256-bit||256-bit||256-bit||192-bit||352-bit|
|Memory size||11GBGDDR6||8GBGDDR6||8GBGDDR6||8GBGDDR6||8GBGDDR6||8GBGDDR6||6GBGDDR6||11GB GDDR5X|
A look at the table above is pretty succinct and dry: more power across the RTX board. That’s all we wanted from the expensive RTX line in the first place, especially while waiting for games to actually unleash their RTX-exclusive bits. To date, the number of games that use such tricks via proprietary Nvidia technology (global lighting, ray tracing, deep-learning supersampling) can be counted on your fingers.
What’s more, the only RTX compatible game on the market that I would currently argue is a no-brainer, heck-yes upgrade is 1998’s newly modified version earthquake II† So I wouldn’t have blamed you for waiting for a better price opportunity to dive into the RTX line (according to Nvidia’s public statements, you were far from alone). The Super line may be the perfect opportunity to do this.
Without an extensive RTX test suite at my disposal, I’m limited to some admittedly meager test results (also due to a tight turnaround time between cards arriving in-house and this embargo). But my early tests say enough. I pitted the RTX 2070 Super “Founder’s Edition” against my everyday driver, an RTX 2080 Founder’s Edition that I reviewed 10 months ago.
And I was shocked to discover something Nvidia didn’t mention in a nearly hour-long briefing on the cards: that the RTX 2070 Super is basically right up there with the RTX 2080. Not identical, mind you, but the results are surprisingly tight. for a price difference of $200.
The above benchmarks are standard for game testing at Ars, and they all see most in-game settings jacked up to near-maximum values at 4K resolution to ensure the largest possible GPU-bound testing environment . Grand Theft Auto VThe longest benchmark scene features a three-frame differential in both standard and overclocked versions. The RTX 2080 wins the Hitman 1 showdown in all sizes by about 7%, while Assassin’s Creed Odyssey is a dead-heat until both cards are overclocked.
Incredibly, by using EVGA’s X1 Precision overclocking app (which plays well with Nvidia cards), I managed to get the RTX 2070 Super to defeat his RTX 2080 sibling in at least one game: Quake II RTX, a game whose performance appears to be largely RTX bound. That resulted in a performance edge of about 5% for the 2070 Super with the game’s RTX-exclusive global lighting advantage enabled, in both “medium” and “high” RTX settings. But be warned: that is not based on a rigid, perfectly controlled benchmark test, so don’t assume the 2070 Super has a clear overall lighting edge over the RTX 2080.
$499 is, of course, still a pretty penny to spend on a graphics card, let alone one whose selling point revolves around a few standards of its own. But this looks like a decent 2019 GPU. It compares favorably with the existing RTX 2080, meaning it does great things in 1440p and even 4K resolutions, while providing RTX benefits that would bring PC players closer to truly smooth RTX gameplay scenarios in at least 1080p resolution. have to bring.
For now, consider this an “impression” post as we wait for the RTX 2080 Super to land at Ars HQ later this month. At that point we’ll see what exactly to be new “$200 more” sales pitch will be – and we’ll be putting more games and apps to the test, especially as AMD continues to chase Nvidia.
List image by Nvidia