Mobile Gaming – New Life with Kepler
Clearly the NVIDIA Tegra K1 processor has the most impressive graphics system seen in a mobile processor to date, and if there is one area where I feel confident that NVIDIA can maintain that advantage, it is with the substantial graphics architecture IP it holds. But as we saw with the release of the NVIDIA SHIELD, great graphical computing capability doesn’t always mean great content – Android games continue to lag behind consoles, PCs, and dedicated handhelds when it comes to high quality games.
With the Kepler integration in Tegra K1, some very exciting avenues are opening up. Check out this comparison table (provided by NVIDIA) that looks at the performance of its latest SoC against the Xbox 360 and Playstation 3.
Comparing these architectures is difficult since they are based on very different hardware. The crucial compute data points show an impressive equalization of graphics capability across the three platforms. The Tegra K1 has a higher peak shader performance, a higher texture rate, a larger addressable memory capacity, and support for more recent shader models like DX11 and OpenGL 4.4. Even looking at the processor performance of the quad A15, you’ll see roughly 1.5x the performance in the K1 versus the Xbox 360! The only area in question is the memory bandwidth provided by the Tegra K1, but the consoles have the advantage of desktop-class memory buses and power consumption.
Just as important as the hardware performance is the API support of the hardware. Tegra K1 supports both DirectX 11 as well as OpenGL 4.4, though I think it is clear that OpenGL is going to become the key player for NVIDIA on this product. Games developed for the Playstation 3 were all built using OpenGL and upcoming titles built for the Playstation 4 will use OpenGL as well, leaving a door wide open for the Tegra K1 that other SoCs in the mobile ecosystem just cannot match. Imagine a market in which nearly all of the previous generation console titles are available for Tegra K1 based products thanks to the portability of the graphics API.
It may not be that far off of a dream – NVIDIA has always invested heavily in developer tools for gaming, and the introduction of the Tegra K1 means that these tools will address it as well. NVIDIA is actively working on perfecting the tools to enable game developers to easily port games from OpenGL on GeForce to OpenGL on Tegra K1. It is not a trivial feat, but NVIDIA did bring out several demonstrations to drive home the point that it’s possible. It is amazing when it works.
NVIDIA has ported a handful of Unreal Engine 4 demos to the Tegra K1 and I got to see them running in person on the reference platforms last month. The “room” demo and the shooter demo both worked flawlessly and produced gaming and image quality that I had not seen before in a mobile device. The number of games and developers working with Unreal Engine is staggering and NVIDIA is putting a lot of emphasis on perfecting the implementation on K1. In fact, you might have seen a tweet from Epic’s Mark Rein recently that was cleverly hiding excitement over Tegra K1…
Serious Sam 3, a game released in 2011, was shown working on the same reference platform.
NVIDIA’s Tony Tomasi, Senior VP of Content and Technology, was glowing with excitement over these hardware and software innovations and what it will mean for mobile gaming. NVIDIA is removing all the roadblocks of technology that are currently preventing developers from bringing world class content to mobile devices. Matching performance with the previous generation of consoles, while also supporting newer feature sets and APIs, could be a perfect combination to finally get NVIDIA’s Tegra chip in a place where it can succeed and thrive.
Trust me, there are still lots of questions about Tegra K1 that keep us from getting too excited. The features and performance as presented by NVIDIA during our previous briefings as well as during the press conference last night are truly impressive, but this company above all others knows how to market and present its products. I was hoping we would have a prototype reference Tegra K1 system prior to writing this article that would allow us to do much more in line with a full review rather than a feature and system preview. Alas, that didn’t happen, so we’ll wait for one to make it to our offices and give it a thorough analysis.
Speaking of availability of samples, as of this writing we don’t have any official retail ready products to announce based on the Tegra K1 at all. NVIDIA told me in December that they expect to have products for consumers based on K1 in the first half of 2014. Super phones were confirmed but not for the North America region so they might be relegated to outside the US once again.
It would be crazy for NVIDIA to not continue down the line with its own mobile devices in the Tegra line, so you will surely see an updated SHIELD using Tegra K1 and with the prototype K1 tablet already integrated into the Note 7 chassis, that’s a shoe in as well. In fact, we even had specifications of that prototype available to us beforehand; 1920×1200 resolution screen, 4GB of memory and of course the Tegra K1 SoC.
The Denver-based iteration of Tegra K1 brings a dramatic shift to NVIDIA place in the mobile market and moves them up to a very exclusive tier of ARM partner. Performance and effiency of these custom cores with the ARMv8 64-bit architecture is going to be very interesting to see later in 2014 and we'll be able to judge the first product from the engineers that NVIDIA has been scooping up for the last 5 years.
NVIDIA’s Tegra K1 is ambitious and is the first true iteration on the company’s initial goals for the Tegra line. Yet again, we see the beginnings of an SoC that could actually shift the direction of the industry; we just need to see implementations that are as exciting as the chip itself.