NVIDIA's new GeForce RTX 2070 Founders Edition is the company's third Turing card release in recent weeks. With a base price of $499 and a Founders Edition price of $599, it is targeted at the performance-segment, at people who are looking for a graphics card that can handle 1440p gaming or deliver high refresh rates in 1080p. Unlike previous generations, the Founders Edition is only available from NVIDIA—board partners will have to come up with their own designs.
When averaged over all our benchmarks at 1440p resolution, we see the RTX 2070 Founders Edition right between the GTX 1080 (+15%) and 10% behind the GTX 1080 Ti. The difference to the RTX 2080 is 20%, and the flagship RTX 2080 Ti is 45% faster. Compared to AMD's fastest, the RX Vega 64, the RTX 2070 is 19% ahead. With those performance numbers, the RTX 2070 is an excellent choice for maximum details 1440p gaming, able to give you at least 60 FPS in all recent titles. High-FPS 1080p gaming is possible, too, but the card won't reach 144 FPS in all titles at that resolution unless you dial down some settings.
The visual appearance of NVIDIA's card is amazing; the GeForce 20 Founders Edition cards are the sexiest graphics cards released to date, and the RTX 2070 Founders Edition is no exception. Compared to the RTX 2080 and RTX 2080 Ti, the card is much more compact, measuring only 20 cm in length. A surprising change is the location of 8-pin PCIe power input: it's located near the backside of the card and no longer on the long side. This makes it easier to install the card in thinner cases, but could complicate things in cramped Mini-ITX setups. For full-size cases, this might actually be a good change because clean cable-routing is much easier.
NVIDIA's dual-fan, dual-slot thermal solution does a reasonably good job of keeping the card cool, at temperatures of up to 75°C, which is alright given the cooler. What I really miss is the highly popular idle-fan-off feature that's almost standard on graphics cards today because it will turn off the fans in idle, desktop work, Internet browsing, and light gaming for the perfect noise-free experience. Fan noise while gaming is definitely improved over previous Founders Edition cards. I would describe it as "quiet", with 34 dBA, even though I see some headroom for custom board design to improve those noise levels even more.
NVIDIA has only made small changes in their Boost 4.0 algorithm compared to what we saw with Pascal. For example, instead of dropping all the way to base clock when the card reaches its temperature target, there is now a grace zone in which temperatures drop slowly towards the base clock, which is reached when a second temperature cut-off point is hit.
Power limits are a different beast, though. All Turing cards are no longer thermal limited, but, rather, limited by power. During normal gaming, they will all sit in their power limit, which means the highest boost clocks are never reached during regular gameplay, which is in stark contrast to Pascal, where custom-designs were almost always running at peak boost clocks. The rated boost clock for the RTX 2070 Founders Edition is 1635 MHz—we measured an average clock frequency of 1862 MHz, so Boost is still working fine, giving you almost 250 MHz higher clocks without any tweaking.
Manual overclocking has once more become more complicated with this generation. Since the cards are always running in the power limiter, you can no longer just dial in stable clocks for the highest boost state to find the maximum overclock. The biggest issue is that you can't just reach that state reliably, so your testing is limited to whatever frequency your test load is running at. Nevertheless, we managed to pull through and achieved a decent overclock on our RTX 2070 that translated into 8% additional performance. Our sample didn't overclock just as well as the bigger Turing cards. Whether that is a one-off or expected for the TU106 GPU we can't say just yet—custom designs from board partners are already underway, so stay tuned.
Power efficiency of the RTX 2070 is similar to Pascal, matching the GTX 1080 and GTX 1070 almost exactly. Compared to the RTX 2080 and RTX 2080 Ti, we do see the RTX 2070 behind a bit in power efficiency, but the differences are small enough to not matter during daily use. What is an issue, though, is the high multi-monitor power consumption (38 W), which is more than 5x as high as with the GTX 1070 (7 W). We hope NVIDIA will continue working on that as many professionals use more than a monitor and this 30 W power draw difference could add up to a not insignificant power bill increase.
NVIDIA RTX has been the great big promise of this generation, as without it, the performance gains over generations aren't as great as we've come to expect from NVIDIA. Gamers are being asked to wait for real-time ray-traced elements in their game that add more visual fidelity than is possible with rasterization. Unfortunately, it's been a month since the RTX 20-series launched, and there is not one real-world game that uses RTX. This is unfortunate for NVIDIA because consumers could be drawn to purchase better-priced Pascal-based alternatives, which would cut them out of the RTX captive audience base. Since it has fewer RT cores than say an RTX 2080 Ti, the RTX 2070 could in theory take a greater hit in performance with RTX enabled than the RTX 2080 Ti, but NVIDIA has confirmed that RTX games will come with more fine-grained settings than just "RTX on" or "off".
The second big novelty of Turing is acceleration for artificial intelligence. At first thought many assumed that it won't do anything for gamers, but NVIDIA devised a clever new anti-aliasing algorithm called DLSS (Deep Learning Super-Sampling) which utilizes Turing's artificial intelligence engine. DLSS is designed to achieve quality similar to temporal anti-aliasing and to solve some of its shortcomings, with a much smaller performance hit. For the RTX 2070 this could mean that 4K 60 FPS gaming becomes a reality—if this materializes, it'll be a game changer.
NVIDIA is currently pricing the RTX 2070 Founders Edition at $599 in their online store, which isn't cheap. This is a $100 increase over the more basic designs that are already out by board partners. However, even at $600, the RTX 2070 offers you more value for the money than the GTX 1080 Ti, plus it offers all the new features, like RTX and DLSS. The GTX 1080 is about 10% more cost effective though, possibly more when buying a used card. When choosing between the RTX 2070 and GTX 1080 it comes down to how much you trust in NVIDIA (and game developers) to commit to RTX and DLSS. Especially DLSS seems to be the biggest chance for the RTX 2070 to turn into a killer card that will give you 4K 60 FPS gaming at a previously unseen price. I do trust that NVIDIA can pull it off, and for me, this 10% increase is more than justified; I'd buy a RTX 2070 over the GTX 1080 any day. Competition does come from the board partners and their MSRP-priced variants. If these turn out good, then the saved $100 could end up turning the tide for them; that is, unless you want the gorgeous looks of the Founders Edition, of course.