Nvidia's first SIGGRAPH announcement, the launch of the Quadro K6000 board, may have been somewhat niche, but last night's presentation was anything but: a sneak-preview of its next-generation Tegra family, codenamed 'Logan.'
Continuing the theme of using superheroes' public personae as codenames - 'Logan' representing X-Men's Wolverine, of course - Logan is due to take over from the recently-launched Tegra 4 in 2014 when it will launch as the Tegra 5. While it will include the usual tweaks to the ARM-based general-purpose processing cores, Nvidia's teaser event at SIGGRAPH in Anaheim last night concentrated, naturally, on the graphics side of things with the news that Logan will represent a standardisation on the Kepler GPU architecture across Nvidia's entire product offering.
'Our mission with Project Logan was to scale [Kepler] technology down to the mobile power envelope,' explained Nvidia's Jonah Alben of the SIGGRAPH announcement, 'creating new configurations that we could both deploy in the Logan mobile SoC [System on Chip] and license to others. We took Kepler’s efficient processing cores and added a new low-power inter-unit interconnect and extensive new optimizations, both specifically for mobile. With this design, mobile Kepler uses less than one-third the power of GPUs in leading tablets, such as the iPad 4, while performing the same rendering. And it gives us enormous performance and clocking headroom to scale up.'
That headroom, Nvidia claims, will see the mobile version of the Kepler architecture in Logan rocket past anything else on the market. A slide used at the show highlights this in particularly stark fashion: a slow ramping in graphics performance becomes a cliff as Logan rises above the iPad 4, the PlayStation 3, and even Nvidia's desktop GeForce 8800 GTX GPU - a chip which, when it launched in 2006, was the fastest around and had a thermal design profile (TDP) of 185W.
Okay, so it's approaching seven years old now - but for a mobile chip, in a hand-held device that runs on an internal battery, to beat the speed of even an outdated 185W desktop graphics card is no mean feat.
'We achieved this efficiency without compromising graphics capability,' claimed Alben. '[Mobile] Kepler supports the full spectrum of OpenGL – including the just-announced OpenGL 4.4 full-featured graphics specification and the OpenGL ES 3.0 embedded standard. It also supports DirectX 11, Microsoft’s latest graphics API. These advanced APIs will allow developers to use more efficient, visually compelling rendering approaches than were previously possible in mobile.'
Indeed, Alben name-checked features more normally associated with desktop GPUs, including tessellation, compute-based deferred rendering, new anti-aliasing and post-processing algorithms, and even GPU-accelerated physics and other simulation engines with Logan promising full support for the CUDA 5.0 GPGPU offload language - the same feature set as Nvidia's desktop GPUs, in other words, but in a 2W TDP.
To show off the capabilities of the chip, Nvidia treated attendees to its previously-released FaceWorks 'Ira' demonstration runnning in real-time - but whereas the previous demo, at GTC 2013, required half the power of the company's flagship Titan graphics card, this version runs entirely on a Logan development board.
If you're curious as to just how that works, Nvidia's demonstration is reproduced below - and, for now, represents the last little tidbit the company is willing to share regarding Logan until closer to its 2014 début.
Continuing the theme of using superheroes' public personae as codenames - 'Logan' representing X-Men's Wolverine, of course - Logan is due to take over from the recently-launched Tegra 4 in 2014 when it will launch as the Tegra 5. While it will include the usual tweaks to the ARM-based general-purpose processing cores, Nvidia's teaser event at SIGGRAPH in Anaheim last night concentrated, naturally, on the graphics side of things with the news that Logan will represent a standardisation on the Kepler GPU architecture across Nvidia's entire product offering.
'Our mission with Project Logan was to scale [Kepler] technology down to the mobile power envelope,' explained Nvidia's Jonah Alben of the SIGGRAPH announcement, 'creating new configurations that we could both deploy in the Logan mobile SoC [System on Chip] and license to others. We took Kepler’s efficient processing cores and added a new low-power inter-unit interconnect and extensive new optimizations, both specifically for mobile. With this design, mobile Kepler uses less than one-third the power of GPUs in leading tablets, such as the iPad 4, while performing the same rendering. And it gives us enormous performance and clocking headroom to scale up.'
That headroom, Nvidia claims, will see the mobile version of the Kepler architecture in Logan rocket past anything else on the market. A slide used at the show highlights this in particularly stark fashion: a slow ramping in graphics performance becomes a cliff as Logan rises above the iPad 4, the PlayStation 3, and even Nvidia's desktop GeForce 8800 GTX GPU - a chip which, when it launched in 2006, was the fastest around and had a thermal design profile (TDP) of 185W.
Okay, so it's approaching seven years old now - but for a mobile chip, in a hand-held device that runs on an internal battery, to beat the speed of even an outdated 185W desktop graphics card is no mean feat.
'We achieved this efficiency without compromising graphics capability,' claimed Alben. '[Mobile] Kepler supports the full spectrum of OpenGL – including the just-announced OpenGL 4.4 full-featured graphics specification and the OpenGL ES 3.0 embedded standard. It also supports DirectX 11, Microsoft’s latest graphics API. These advanced APIs will allow developers to use more efficient, visually compelling rendering approaches than were previously possible in mobile.'
Indeed, Alben name-checked features more normally associated with desktop GPUs, including tessellation, compute-based deferred rendering, new anti-aliasing and post-processing algorithms, and even GPU-accelerated physics and other simulation engines with Logan promising full support for the CUDA 5.0 GPGPU offload language - the same feature set as Nvidia's desktop GPUs, in other words, but in a 2W TDP.
To show off the capabilities of the chip, Nvidia treated attendees to its previously-released FaceWorks 'Ira' demonstration runnning in real-time - but whereas the previous demo, at GTC 2013, required half the power of the company's flagship Titan graphics card, this version runs entirely on a Logan development board.
If you're curious as to just how that works, Nvidia's demonstration is reproduced below - and, for now, represents the last little tidbit the company is willing to share regarding Logan until closer to its 2014 début.
MSI MPG Velox 100R Chassis Review
October 14 2021 | 15:04
Want to comment? Please log in.