Nvidia demoes Logan for mobile: a big leap for mobile graphics

July 24, 2013
19 22 2 7
Nvidia showed the press the first demos of its upcoming Kepler-based mobile GPU, dubbed Logan. Due in 2014, Logan will bring feature parity between mobile GPUs, PCs, and consoles.

    ira logan nvidia kepler

    In spite of Nvidia’s illustrious pedigree in computer graphics, the GeForce GPUs contained in its Tegra SoCs have been consistently outshone by competing offers from ARM or Imagination Technologies. That will change soon.

    The Santa Clara-based company gave us today a tantalizing glimpse of the possibilities that Logan, the first mobile SoC to feature a Kepler-based GPU, will open starting next year.

    The two demoes below are rendered by a mobile SoC running on a development board. The first shows Ira, a highly detailed model of a human head rendered in real time at 1080p. You might remember Ira from the presentation that Nvidia’s CEO gave earlier this year when introducing Kepler. Back then, the demo was rendered on a GTX Titan, a $1000 PC GPU targeted at hardcore gamers and professionals. The video below is not as impressive, but Nvidia’s technical achievement is still amazing, considering that the company managed to bring power consumption down to watts levels, from the 250 watts consumed by the Titan.

    The second demo is dubbed Island and is meant to showcase hardware tessellation and the new OpenGL 4.3 API compliance.

    One of the most impressive things about Logan is that it will bring feature parity between mobile devices and PCs and consoles. That means that the mobile implementation of Kepler will support features found on GPUs powering PCs and consoles, such as OpenGL 4.4, DirectX 11, tessellation, and CUDA 5. Game developers will be able to develop games across platforms much easier than before, which is hugely important for the growth of mobile gaming.

    Finally, consider for a second Nvidia’s claim that the 2014 Logan will offer more raw horsepower than a PS3 or a 8800 GTX PC graphics card, all while maintaining a level of power consumption that allows it to run on smartphones and tablets.

    NVIDIA_Siggraph_Mobile_HR_21-500x312

    Hopefully, we’ll see Logan come to market by mid-2014 as part of Tegra 5, if everything goes to plan; knowing Nvidia, though, we wouldn’t hold our breath. The first Tegra 4 product, the Shield console, will only launch next week.

    Nvidia announced recently that it’s open to license Kepler technology to anyone interested, meaning that we might see more mobile SoCs come to market with console-like graphics performance in the next 18 months.

    For a more detailed presentation on Logan check out this post from AnandTech.

    Comments

    • Jean-Claude

      WE’LL BE ABLE TO RUN CRYSIS??

      • guest1110

        fo sure!

    • Roberto Tomás

      a big problem is that it is only coming in 2014, and it is 28nm. Every first-tier ARM manufacturer will be moving to 20nm then. They will have no clients for the IP, and basically only Logan parts will ship with it — otherwise, it is really a quality product. Power hungry at max, but nice.

      • guest1110

        20nm parts will NOT have good yields. NVIDIA will WIN this round. And going directly to 16nm or 14nm, will be costly.

        • Roberto Tomás

          that is interesting … 32nm was having so many problems most companies jumped straight to 28nm. 28nm had such problems they lost a lot of main contracts, and spent an extra year on the node. now you think 20, 16, and 14 will all continue to have problems? what is this, the end of computation? :P

      • http://AndroidAuthority.com/ Bogdan Petrovan

        Why do you think they will have no clients for the IP?

        • Roberto Tomás

          Over at Ubergizmo the moderator there gave me a very good reason to doubt my statement. He said that just because they distribute the IP at 28nm doesn’t mean that clients will be bound to use it at 28nm. So, assuming that is the case, sure, this will work wonders and I can’t wait to see it. but if it is stuck at 28nm, you’re not likely to see any top-tier manufacturer use it except NVidia itself because of the time frame —the last 28nm designs will be coming out later this year and early next year: after that it will only be budget firms. So the question they will have to ask themselves about purchasing this IP is “why switch to a design you know will only last 3 months?”

    Popular

    Latest