Well, folks have been saying for some time that the graphics card market desperately needed some competition. Nvidia’s pricing and naming shenanigans with the original GeForce RTX series were no doubt in part due to a lack of any stabilizing force preventing the company from doing exactly what it did. AMD’s latest Radeons may not vie for the PC graphics performance or efficiency crowns, but they’re exceedingly solid products in their own rights. To see how they stack up value-wise, let’s consult the world-famous Tech Report scatter plots.
Features named with three- and four-letter acronyms aside, as a pure comparison of pixel-pushing prowess, the Radeon RX 5700 XT matches the GeForce RTX 2060 Super blow-for-blow. It is every bit as smooth, and just a hair faster overall, when compared to Nvidia’s card. In some titles it falls behind, and in others it punches above its weight class. Whether the result of Nvidia-specific game optimizations or poor driver code, some games just run particularly poorly on Radeons.
The most impressive thing to me is that, in many of these games, AMD’s cards actually put up a smoother experience than Nvidia’s. Once upon a time, AMD struggled with consistency, even when using a single graphics card. Now, those fortunes seem to have been completely reversed… until you play something like Mirror’s Edge Catalyst. Nvidia owns the majority of the PC gaming graphics card market, and that reality means that Radeon gamers are occasionally going to come across a title that just doesn’t work very well on their graphics card.
On the flip side of that argument, Nvidia seems keen to promote proprietary features that require game-specific support. Don’t be mistaken: I like RTX, and I don’t hate DLSS—at least in theory—but it’s hard to deny that they feel awfully limited compared to Radeon Image Sharpening and Radeon Anti-Lag, at least in terms of purview. Nvidia has its share of cool game-agnostic features too, to be sure. FreeStyle is basically ReShade, built right into GeForce Experience, and AMD’s Virtual Super Resolution is a poor imitation of Nvidia’s far-superior Dynamic Super Resolution. It’s just hard to put much value in Turing’s special sauce because there are so few opportunities to use it.
Greenbacks to graphics cards
After the announcement of the GeForce RTX Super cards, AMD dropped the price on the RX 5700 from $379 to $329. That’s a heck of a lot of video card for that price. In the $300-400 range, it’s difficult to imagine recommending anything besides the Radeon RX 5700. It absolutely mashes the GeForce GTX 1660 Ti, and creeps up on the RTX 2060 Super surprisingly often.
Once you get up to $400, it’s definitely a toss-up between the Radeon RX 5700XT and GeForce RTX 2060 Super. Both are excellent 2560×1440 gaming cards with the grunt to stretch their legs into 4K on some titles. The GeForce card comes with forward-looking RTX support and more consistent performance, but the Radeon card has slightly better raw performance and doesn’t require an online-enabled login to use all of its extra features. I wouldn’t fault anyone for choosing either card.
Moving up the price range, we next have the GeForce RTX 2070 Super at $550. This card, particularly our whisper-quiet triple-fan model from Gigabyte, is not difficult to recommend. GeForce GTX 1080 Ti-level performance, true single-card-4K-at-60-FPS performance, for just $549, is excellent. My singular quibble with the RTX 2070 Super is that it only has 8GB of memory. That might seem a silly complaint in a world where most games are barely using 5GB of video RAM on the most extreme settings, but that won’t be the case for much longer. After all, Monster Hunter World in 4K can already exceed 7.5GB of video RAM usage, just as one example.
The truly “super”
If 8GB of RAM is a little short on the RTX 2070 Super, it’s legitimately worrisome on the RTX 2080 Super. This is a card that pushes nearly 120 FPS average in Doom while playing in 4K. It would be tragic, almost embarrassing, if you had to turn down a game’s settings on this otherwise-monstrous graphics card because you ran out of video RAM. The idea smacks of the R9 Fury X and its 4GB of memory. For gamers like myself who use myriad monitors—I’m currently using five—the extra video memory on the “x80 Ti” cards offers some breathing room when games want to suck down over 7GB of graphics space.
As much as I’m impressed by the nearly-world-beating performance of the RTX 2080 Super, I can’t help but feel that it’s not quite enough of an upgrade from the RTX 2070 Super. Going over our benchmarks, there are only a few titles where the RTX 2080 Super is likely to give you a meaningfully-different experience compared to its little brother. Aside from the extra performance, they have the same capabilities, so it’s easy to argue that most buyers should just save $150 and pick up the cheaper card.
With all of that in mind, there’s certainly something to be said for the RTX 2080 Super’s no-compromise performance. For folks who have the money, this is surely the card to get, because the next step up from here is the $1050 RTX 2080 Ti. I’ll admit that that card might end up being the better buy in the long run thanks to its 11 GB of memory, but the extra $350 probably doesn’t buy you much more in terms of realized performance. If you’re not powering an HP Omen X Emperium, the extra speed is likely wasted anyway.
The matter of the pre-loved market and the previous-generation
When Nvidia priced the original GeForce RTX series the way it did, it seemed like a lot of gamers who were itching for an upgrade started looking once more at discounted and used 10-series parts. This time around, I think most users should go ahead and keep their eyes on the new models. You can bet that future optimizations in Radeon drivers will be targeted at the RDNA architecture in Navi, and these are the most efficient Radeons ever made. For those who bleed GeForce green, it’s difficult to recommend buying Pascal at this point for much the same reason.
First-generation GeForce RTX cards could be a good buy depending on pricing, but they’d need to be deeply discounted. The upgrades that Nvidia doled out with the Super series are very near to a straight step-up, tierwise, so the RTX 2060 Super seems similar to an original GeForce RTX 2070 in overall performance. Using the GTX 1080 Ti as a reference point, the RTX 2070 Super looks like it runs just behind the OG RTX 2080. If you could find an original RTX 2080 for less than the price of an RTX 2070 Super, then we might say “go for it,” but all the usual caveats of buying used hardware apply.
In the end, I actually think AMD and Nvidia both succeeded in their goals with these new products. For Nvidia’s part, all three GeForce RTX Super cards offer improved performance and improved performance-per-dollar, making them more attractive to consumers, which will in turn drive adoption of RTX. Nvidia has a firm grip on the high-end graphics card market, and AMD knows that, so it seems like these latest Radeons are an attempt to elbow into the high-midrange or mid-high-end of the market, where most graphics cards are sold. Navi is “competitive” at the very least and “impressive” more often than not, while RDNA is a much-needed upgrade for the time-worn GCN architecture. PC gamers won’t go wrong with any of these graphics cards.