Skip to main content

Nvidia Titan RTX: Everything you need to know

The Titan RTX graphics card is nearly here. Here's what you need to know

Image used with permission by copyright holder

Nvidia’s RTX series of graphics cards are all rather expensive and all rather powerful, but the most expensive and most powerful of them all, is the Titan RTX. Sitting one notch above the RTX 2080 Ti and targeted more at enterprise and prosumer users than gamers, the Titan RTX represents the pinnacle of graphics card power available today.

Should you buy one? Probably not, but it doesn’t hurt to indulge your imaginary bottomless wallet now and again. Here’s everything you need to know about the Nvidia Titan RTX.

Recommended Videos

Pricing and availability

The Titan RTX was first teased in a number of “accidental” reveals by the likes of Linus Tech Tips and the Slow Mo Guys at the end of November, but it didn’t appear to bother Nvidia too much. It followed up with the official announcement days later. It revealed the Titan RTX in all its gold-tinted glory, alongside its eye-watering price tag of $2,500.

Get your weekly teardown of the tech behind PC gaming
Check your inbox!

What we don’t know, is exactly when it will be released. Based on previous generation’s release dates for Titan cards, they typically hit the digital and real store shelves a couple of months after the main lineup. With AMD expected to make a big splash with Navi at CES in January too, we’d be surprised if Nvidia didn’t make these cards available before then.

Nvidia has suggested the Titan RTX should be available before the end of the year, so we’re going with that assumption. If it slipped into early 2019, though, perhaps to try to steal some CES thunder, it wouldn’t surprise us.

Architecture and performance

The Titan RTX is based on the same Turing architecture as other cards in the RTX range such as the 2070, 2080, and 2080 Ti. However, it is what could be classed as a ‘fully-unlocked’ RTX card, because it uses the same TU102 GPU core as the 2080 Ti, only with none of its CUDA cores disabled. Here’s how it compares to other cards in the RTX line up.

Titan RTX RTX 2080 Ti RTX 2080 RTX 2070
CUDA cores: 4,608 4,352 2,944 2,304
Tensor cores: 576 544 368 288
RT cores: 72 68 46 36
Base speed: 1,350MHz 1,350MHz 1,515MHz 1,410MHz
Boost speed: 1,770MHz 1,545MHz 1,710MHz 1,620MHz
Memory:
24GB GDDR6 11GB GDDR6 8GB GDDR6 8GB GDDR6
Memory speed: 14Gbps 14Gbps 14Gbps 14Gbps
Memory interface: 384-bit 352-bit 256-bit 256-bit
Memory bandwidth: 672GB/s 616GB/s 448GB/s 448GB/s
TDP: 280 watts 250 watts 215 watts 185 watts

While clock speeds and core counts are all welcome improvements over the 2080 Ti, the starkest difference between those two powerful cards is the memory. Where the 2080 Ti has 11GB of high-speed GDDR6, the Titan RTX sports 24GB. Considering few games will even take advantage of the full complement of memory that the 2080 Ti comes with, this should highlight the intended audience for the RTX Titan. Such mammoth quantities of memory lets it deal with enormous datasets (perfect for AI), and heavy computational tasks in datacenters.

The boosted hardware over the 2080 Ti delivers the kind of raw power bonuses that you might expect. In single and double precision tasks, the Titan RTX is 10-20 percent more capable than the 2080 Ti, as per Anandtech. Arguably more important though, is the tensor performance improvement. Even with a modest improvement in tensor core count, the Titan RTX can offer performance in excess of double that of the 2080 Ti when those cores are leveraged, delivering a total of 130 TFLOPs. That’s even in excess of the capabilities of the Titan V, which has a greater transistor count, as well as a greater number of CUDA and tensor cores.

What all of this means, is that the Titan RTX bridges the gap between the highest performing gaming hardware and the enterprise-focused Quadro cards. At $2,500, the Titan RTX is very expensive, but it’s well under half the cost of the Quadro RTX 6000, making it a viable option for those who can’t quite justify such expense. It should also work brilliantly as a strong workstation card that can also be used for high-end gaming, for that particularly niche audience.

In terms of the RTX Titan’s real world performance, we don’t know exactly what it can do. It should be better than just about any card we’ve ever seen when it comes to gaming performance, but almost no one will use it for that task, so that’s largely unimportant. Except if you love hardware, so we can’t wait to see what it can do when reviewers get their hands on it.

Ray tracing? DLSS?

Perhaps the most intriguing question to be answered by the release of the Titan RTX, is how well will it handle Nvidia’s new visual technologies: Ray tracing and deep learning super sampling (DLSS)? We’ve seen middling and downright poor performance of Nvidia’s new RTX cards when it comes to ray tracing in games like Battlefield V, and DLSS has real potential, but isn’t as groundbreaking as Nvidia might like you to believe.

In theory, the Titan RTX will be better at ray tracing than the 2080 Ti. It’s much faster clock speed should keep its head above water at higher resolutions and the additional four RT cores might keep frame rates bumped up enough to be considered playable, but it’s still hard to imagine even this powerhouse from being able to deliver playable frame rates with ray tracing enabled at 4K.

We’d love to see it happen, but we’re not holding our breath that even the Titan RTX can handle it.

Topics
Jon Martindale
Jon Martindale is a freelance evergreen writer and occasional section coordinator, covering how to guides, best-of lists, and…
Nvidia may have found a new way to bypass GPU export restrictions
The RTX 4090 graphics card sitting on a table with a dark green background.

Chinese gamers are expected to have limited access to Nvidia's best graphics card due to strict export restrictions. However, Nvidia may have found a way around it without cutting down its flagship GPU. A new leak suggests that the RTX 5090D will have the same hardware specifications as the worldwide version, and the solution lies in firmware adjustments.

As a quick refresher, the "D" in RTX 5090D stands for "Dragon." It marks GPUs made by Nvidia to bypass the export restrictions imposed by the U.S. on China, which limit the sale of high-performance graphics processors. We first saw it appear in the RTX 4090D as a remedy for the fact that the base RTX 4090 exceeds the performance thresholds set by these regulations.

Read more
MacOS update: everything you need to know
The MacBook Pro 16-inch on a table.

Whether you've owned one of the best Macs for years or are totally new to the Apple ecosystem, knowing your way around the latest updates to macOS is crucial. Thankfully, you don’t have to do so alone, since we’ve got everything you need to know about the latest macOS updates.

Should you upgrade to the newest macOS release, and how do you do that in the first place? How does it compare to older versions of macOS -- and, indeed, Microsoft’s Windows operating system? We’ve set out to answer all those questions and more, giving you all that you need to know when it comes to the newest macOS update.
macOS Sequoia

Read more
Nvidia CEO in 1997: ‘We need to kill Intel’
NVIDIA CEO Jensen Huang at GTC

Those headline above includes strong words from the maker of the best graphics cards you can buy, and they have extra significance considering where Nvidia sits today in relation to Intel. But in 1997, things were a bit different. The quote comes from the upcoming book The Nvidia Way, written by columnist Tae Kim, and was shared as part of an excerpt ahead of the book's release next month.

The words from Nvidia CEO Jensen Huang came as part of an all-hands meeting at the company in 1997 following the launch of the RIVA 128. This was prior to the release of the GeForce 256, when Nvidia finally coined the term "GPU," and it was a precarious time for the new company. Shortly following the release of the RIVA 128, Intel launched its own i740, which came with an 8MB frame buffer. The RIVA 128 came with only a 4MB frame buffer.

Read more