Re-Evaluating the RTX 2070 at Its Proper $500 Price Tag

When Nvidia launched the RTX 2070 last week, we evaluated the card harshly, concluding that its price tag didn’t justify the upgrade while its capabilities in next-generation games that utilize DLSS and RTX were fundamentally unknown and potentially dangerously underpowered given that the only data points we have suggest the RTX 2080 Ti is required to reach 60fps with ray tracing enabled in BF5 and the impact of using ray tracing in the Northlight engine demo by Remedy.

I can tell you why I found those predictions believable. After the disastrous launch cycle of 2014, in which neither AMD nor Nvidia had acceptable inventory levels available for 3-5 months after launch, it wasn’t hard to believe that the same thing might happen again. But an explanation and an excuse are not the same thing, and I apologize to anyone who concluded the RTX 2070 would launch at $600 as a result of what I wrote. The decision to rely on the price predictions of other sites was mine and mine alone.

So let’s take another look at the GPU, this time at an assumed $500 price tag. The cheapest GTX 1070 you can buy today is $385 according to Newegg, the RTX 2070 is $500. The RTX 2070 delivers, according to Anandtech, a 1.36x increase in 1440p performance over the RTX 1070. We agree with Anand that 1440p is the reasonable sweet spot for the card. This works out to a 1.29x increase in price for a 1.36x increase in performance.

Historically, we typically expect GPUs in the $250 – $500 market to continue to offer at least a linear increase in performance for every dollar of additional spending. It’s true that at a certain point, the curve inevitably bends the other direction — luxury video cards don’t deliver a great price/performance ratio. In this case, knocking $100 off the assumed price of the RTX 2070 does improve the performance ratio enough to argue that if you can afford to drop $500 on a GPU, you’ll at least get proportionally faster performance out of doing so.

Does this make the RTX 2070 a better value than I previously believed it to be? Absolutely, yes. Does it make it a compelling GPU purchase? No — though it may be a reasonable one now, depending on your requirements (more on this in a moment).

Still Stuck on the RTX Question

If the RTX 2070 cost $385 today, I’d absolutely recommend it. As a drop-in replacement for the GTX 1070, it’s an easy pick. But again, the problem here is that only one RTX-compatible game is even expected to ship before the end of the year, with everything else arriving sometime in 2019. We have no idea how well the RTX 2070 will be able to handle ray tracing or DLSS — we only know that the RTX 2080 Ti struggled to deliver 60fps in BFV at 1080p, and the RTX 2070 has just 55 percent the rated RTX performance of the 2080 Ti.

Nvidia leaned on ray tracing to justify its price increases, yet all the performance indications we have to date (very few), indicate that ray tracing may put a very heavy load on an RTX card. Without knowing whether the RTX 2070 is going to be able to ray trace effectively, I literally can’t recommend the GPU. I’ve consistently said that I base my own GPU purchasing arguments overwhelmingly on what the GPU can deliver today, and what the RTX 2070 delivers today is a lot of promises and a performance jump a bit smaller than the size of its price increases.

I don’t spend cash on promises. I don’t recommend anyone else does, either.

There’s absolutely a way for Nvidia to make a case for the RTX family: Release better ray tracing performance information. Yes, it’s early days and yes, Nvidia has to work out the details of such releases with its partners, but there is nothing wrong with asking a company to provide proof that its products will perform like it says they will. If reviewers are being overly pessimistic in their evaluation of RTX’s performance hit, Nvidia can fix that simply by being more open about what their performance expectations are.

But Nvidia hasn’t made that case yet — at least, not to me. I’d love for the company to do so. I investigated ray tracing back in 2012 because I was curious about what the technology could do for gaming, and integrating it into rasterized engines could indeed allow for a ‘best of both worlds’ hybrid approach.

It wouldn’t surprise me if Nvidia let the GTX 1080 fall out of market while leaving the 1070 or 1070 Ti to anchor lower price points, which means gamers that don’t pick up a 1080 now ($439 at Newegg) may be stuck with an RTX 2070, period. It’s hard to get a sense for how well Turing might be selling — at Newegg, the top seller, as of this writing, is an RTX 2080. At Amazon, a GTX 1080 Ti is in Spot #8, while the first Turing card doesn’t show up until #18. We haven’t been monitoring either site over the last week and the previous days aren’t archived at the Internet Archive, but sales figures seem to point in opposite directions at Amazon versus Newegg, with the former overwhelmingly dominated by 10-series cards and the latter showing more competition for Turing, though the 10 series remains strong competition at Newegg as well.

Conclusion-ish

After reconsidering the RTX 2070 at the appropriate price point, our guidance is this: If you can afford the cost increase and do not care whether the RTX features pan out, the RTX 2070 will deliver a performance improvement over the 1070 that is mostly in-line with what you’ll be asked to pay for it.

If you actually care about ray tracing, can afford a $500 GPU, and wish to make certain that you purchase a card that can deliver this feature, we do not recommend the RTX 2070, for the sole reason that Nvidia has not yet demonstrated that it will. Once Nvidia has made that demonstration if the card’s price and features are attractive to you, pull the trigger.

If you absolutely in need of a new GPU and cannot afford a $500 card, we recommend either buying a used card (for short-term purposes while waiting to see if NV has 7nm GPUs coming in 2019 and/or what Turing RTX performance will actually look like) or buying the least expensive new card you can afford at a reasonable performance level and banking the rest towards a higher-end GPU once these questions are answered. And keep in mind, Pascal and AMD’s Polaris are both still capable of providing solid midrange performance, even if the 10-series and Vega no longer represent the latest and greatest in GPU technology.

Continue reading

Scientists 3D Print Microscopic USS Voyager With Its Own Propulsion

It doesn't have warp engines, but it can get along fine with the help of hydrogen peroxide and platinum.

In Leaked Memo, LG Proposes Withdrawing From Smartphone Market

A memo circulated at LG declares some hard choices are necessary after $4.5 billion in losses over the past five years. The memo lists several possible courses of action, including the end of LG's mobile business.

The 7 Planets of TRAPPIST-1 Have Surprisingly Similar Properties

Astronomers are getting very good at hunting for exoplanets with a little help from powerful ground and space-based telescopes. We're no longer finding one planet here and there — we're discovering entire solar systems. TRAPPIST-1 has been of particular interest with its system of seven planets, discovered in 2016 and 2017. A new study has confirmed that all these planets are small and rocky like Earth, and they're all surprisingly similar to each other.

New Study Proposes Warp Drive That Might Actually Work

Physicists have speculated on the possibility of a real warp drive for years, but a new paper lays out a vision for a warp drive that might actually work. We still don't know how to build it, but at least we know why we can't build it.