Nvidia kicked off this excruciatingly long week on Monday morning with an event in Germany to debut its upcoming GeForce RTX 20-series graphics cards. The RTX 2070, 2080, and 2080 Ti start rolling out in September, and I can’t stop thinking about the presentation and the less-than-positive reaction to it from some enthusiasts.
I’m excited about the RTX video cards and ray tracing, even though I’m not preordering one. The event had a lot of problems, but I understand why Nvidia chief executive officer Jensen Huang spent so much time trying to sell the audience on ray tracing.
The company is in a chicken-and-the-egg problem in which ray tracing is going to require developers to support it and hardware that can run it, and you don’t get one without the other. And since ray tracing is pretty expensive computationally, you can’t just sneak the feature into a regular upgrade — it has to justify a price increase.
Ray tracing is the future. Developers have wanted it for decades because it “just works.” That’s a phrase that Huang repeated multiple times, and a lot of people are skeptical of it. And they are right to judge the claim that it’s as easy as turning on a switch, but the broader argument that it should require significantly less labor to get graphics to look as beautiful (and probably better) than traditional illumination tools.
“Showing our upcoming game Enlisted, in 4K at over 90fps with our real-time global illumination solution on a new GeForce RTX looks amazing,” Gaijin Entertainment chief executive Anton Yudintsev said on Twitter. “We’re super-impressed with the Nvidia Vulkan ray tracing performance. We had development hardware for only three days — looking forward for more that we can achieve.”
So what is essentially happening here is Nvidia is trying to sell us a more expensive product that will save developers time and money. As it turns out, though, saving developers time and money is not our responsibility, so Nvidia’s No. 1 priority for its event should have been to make the case that ray tracing is something that we will all benefit from.
Or, at the very least, that the price makes sense.
I think Nvidia failed to do that in three key areas. Let’s start with the least offensive mistake and work our way up.
Nvidia talked too much about shadows
Real-time ray tracing is an incredible leap forward because it physically models the behavior of light. This means that developers no longer have to fudge objects like shadows or reflections. To demonstrate this, Nvidia brought Square Enix’s Shadow of the Tomb Raider onstage to show what its shadows looked like with RTX off and on.
Now that I’ve had time to go over this footage and really examine it, I can see how the edges of the shadows look more defined, and the video doesn’t have any weird banding or shimmering. But I did not notice any of this the first time I played the game.
My brain just isn’t looking for details in shadows. Think about it in real-world terms. Except for very specific circumstances involving infiltration of weirdly lit spy installations, why would you ever focus the 15-to-20 watts of brain power on what is effectively useless data? You don’t.
That’s why turning down shadow quality is one of the easiest decisions to make to improve the performance of your games. You get a ton of frames per second back, and you probably won’t notice the difference. It’s a huge bang-for-your-buck when it comes to framerate compared to turning down other visual effects.
And this is why Nvidia’s Tomb Raider demonstration was a mistake. We watch that and we can all say to ourselves, “If this is Huang’s pitch, we can safely dismiss RTX because it clearly doesn’t really matter.”
This doesn’t even get into whether or not RTX is going to hurt performance. The popular assumption among enthusiasts and commentators is that RTX is going to cause a significant dip in performance compared to when you turn it off. That seems lie a safe assumption, but I just don’t know for sure until I get my hands on it. But assuming that is the case, better shadows absolutely will not convince me that I need RTX if I’m losing frames in the process.
Announcing one price and then selling for another
I thought that the rest of the pitch for ray tracing was more effective, but your personal mileage is going to vary when it comes to reflections of lightsources in windows and car doors in a game like Battlefield V.
The awkward faux hype in the video aside, I think this looks phenomenal:
So let’s assume that Nvidia did convince you that ray tracing is something you want. I’m sure some people feel that way because I’m almost on board, and I’m fully aware that only a handful of games are going to support it.
For the people who were ready to buy the cards, Nvidia had other, more creative ways to lose them.
After hours of explanations and demonstrations, Huang finally started getting into the details about the cards themselves. They would start at $500 for the RTX 2070, $700 for the 2080, and $1,000 for the 2080 Ti. That’s a price increase when you compare them to their equivalent models from the GTX 10 series, but I’ll get to that in a moment. If that’s the price, then that’s the price. Fine.
Only those aren’t the prices for those cards. At least not yet. The 2070 is not up for preorder, but if you want the 2080, you need to pay $800. And if you want the 2080 Ti? Nvidia is only selling the overclocked-out-of-the-box Founder’s Edition version for $1,200. The “reference version,” which has the standard clock speeds, will come later.
Those premium early-adopter fees are difficult enough to accept, but what’s worse is that Nvidia didn’t warn anyone that this was going to happen. It had tens of thousands of people watching around the world, and it led them to believe that they would pay one price. And then when everyone loaded up the Nvidia shop page, they were shocked to find much higher prices instead.
And now that shock is all anyone is talking about. It also gives anyone who is on the fence a reason to start poking holes into ray tracing as a feature. Even for the true believers, you have to look at these cards and wonder why you would want them when you can just wait a few months and pay something closer to the announced price.
Undermining the Nvidia numbering convention
Potential RTX customers are also spending a lot of time talking about price increases from generation to generation. They are looking at a GeForce GTX 1080 Ti for $700 and a 2080 Ti for $1,000, and they start doing math.
That’s a 43 percent price increase. Am I going to get 43 percent more performance from the 2080 Ti? Even if I do, price shouldn’t scale with performance like that. Huang talked about Moore’s law where things get faster, smaller, and cheaper, so why did that break down this time?
The reasoning behind that price increase is probably multifaceted. The extra hardware that is powering the real-time ray tracing is probably significantly more expensive than the silicon that went into the Pascal architecture of the GTX 10-series cards. But more importantly, Nvidia probably thinks it can get away with resetting price expectations.
Nvidia, however, shouldn’t want its customers in their heads trying to do algebra to determine if something is wroth the price. It needs to appear as if it’s creating value and giving people a deal.
But you want to know what is especially wild about all of this? RTX 20-series is barely a price increase over GTX 10-series at all. What’s really happening here is that Nvidia switched the damn names around.
A 2080 Ti at $1,000 (let alone $1,200) seems off the charts compared to a 1080 Ti at $700, right? But the 2080 Ti is really the Titan model for this new Turing architecture — the Pascal Titan XP, for example, sells for $1,200. The 2080 at $700 is the same MSRP as a 1080 Ti. And a 2070 at $500 is actually a price drop from the $550 for a GTX 1080.
So instead of launching with the 2080, 2080 Ti, and Titan Turing at comparable prices to the 1080, 1080 Ti, and Titan XP, Nvidia abandoned its naming structure. I have no idea why it would do this. It could call these cards whatever it wants, so I don’t know why it would pick monikers that would invite unfavorable price comparisons.
If I had to speculate, I would guess that Nvidia doesn’t sell a lot of Titan cards. Its xx80 Ti models, however, have a reputation for giving you all the power you would ever need at a much more friendly price. Maybe Nvidia thinks that by putting the Ti name on what is essentially a Titan card, it can get more people to spring for the higher price. I have my doubts about that.
But whatever the reason, it was obviously a mistake because, again, it is giving people another reason to hesitate and take a closer look at ray tracing.
To get people hyped about the RTX cards, Nvidia had to do everything right. It is trying to sell us a distant future, and that’s always tough for creatures like humans that prioritize short-term gains. But the chip manufacturer didn’t do everything right.
If you’re wondering what happens next, I would bet on Nvidia’s stock going up. The truth is that Nvidia’s goal isn’t to get people excited, it’s to sell hardware. And I think these early Founder’s Edition cards are going to sell it fast. Enough people have more money than they know what to do with, and they will show up for the hottest new thing. That doesn’t even include the potential for a resurgence in the GPU-based cryptocurrency-mining market.
And to ensure it remains the market leader in video cards, Nvidia didn’t have to focus on real-world performance in today’s games. AMD is putting zero pressure on Huang’s company when it comes to that. The Vega cards are powerful and impressive. They are a viable alternative … and they are also a year old.
So what does AMD have in the works? Well, I’ve heard some second-hand rumors that it is taking its time, and that is why Nvidia took so long to introduce RTX. Without that competition, Nvidia could end up dominating the growing PC market by default.
That’s not to say that RTX 20 series cards will never turn into a great deal. If you look at the 2070 as the de facto successor to the 1080 and wait for the non-Founder’s Edition prices, things start to make a lot more sense. But I don’t think Nvidia made its case, and I want to see it get back out on a stage or something equivalent and convince us that it knows why we should want these GPUs.