The uber-expensive GeForce GTX Titan Z PC card - designed for next-generation 5K and multi-monitor gaming - has two Kepler GPUs (5,760 CUDA cores) and 12GB of memory with up to 8 teraflops of performance.
But there is a lot more to this company. Co-founder and recently appointed CEO, Jen-Hsun Huang, announced a raft of new products and directions at its GPU Technology conference aiming to “take us from a chip maker to a visual computing company.”
NVidia is Latin for envy and over the past 20 years – since 1993 - its GeForce GPUs have been envied by serious gamers. Times change and it leveraged its chip design and manufacture skills to make ARM based system on a chip (SOC) called Tegra for mobile devices. Its products are excellent but it faces huge competition from Qualcomm, Intel, AMD (purchased competitor ATI), and dozens of smaller ARM foundries. Bottom line - its revenues and profits are under threat.
Next generation Pascal GPU
Huang introduced the next generation GPU called Pascal. He lamented that Nvidia could make even faster video cards but current computer architecture was holding it back. “GPU has memory bandwidth of up to 288GB per second (in the current GeForce GTX Titan), but the CPU memory has bandwidth up to 60GB per second, and the PCI-Express bus can handle only 16GB per second. If we want to increase performance, we’re going to have solve these PC bottlenecks,” Huang said.
Pascal will incorporate two new technologies to assist.
First is NVLink - a high-speed interconnect that will increase the bandwidth between the CPU and multiple GPUs by five to 12 times the interconnect speed of PCIe 3.0.
Second is a new chip packaging technology that places 3D stacks of DRAM memory chips, connected with through-silicon-vias (TSVs), in the same package with the GPU. Nvidia says it can double the memory capacity, deliver up to 1,000 GB per second memory bandwidth, and increase energy efficiency by up to four times.
Scheduled to ship in 2016, Pascal replaces the Volta GPU with stacked DRAM that it announced in 2013.
Huang emphasized that Pascal was well suited to other applications - machine learning for image detection, face and gesture recognition, video search and analytics, speech recognition and translation, recommendation engines, and imaging and search. Adobe, Baidu, Yahoo’s Flickr, IBM, Netflix, Yandex are among the companies using GPU acceleration to enable these kinds of applications.
Nvidia is also developing new lower cost GPUs codenamed Maxwell that supports Microsoft’s DirectX 12 standards. It announced two lower end GPUs – the GeForce GTX 750/i (US$120 with 1GB and $150 with 2GB). These compliment the GeForce 800M – mobile series.
Next generation Tegra K1
Huang also acknowledged how hard it was to play in the ARMs race. “The current Tegra 4/i have struggled to find major design wins and market share in smartphones and tablets,” he said.
The K1 has 192 CUDA cores, 328 gigaflops of power, and four times better energy efficiency than an ARM Cortex-A15. “It will find a place in embedded applications such as computer vision for driver assistance technology,” he said.
The next mobile GPU, known as Tegra Erista is based on the Maxwell GPU architecture and is due in 2015. No details were released.
Great news for cloud users
Ben Fathi, VMWare’s Chief Technology Officer, announced that VMWare’s Horizon Desktop-as-a-Service platform will use Nvidia’s GRID “GPU in the cloud” for applications that are hard to virtualize because they require low latency and detailed 3D graphics. Nvidia already has a similar vGPU (virtual GPU) arrangement with Citrix.
VMWare’s server virtualization technology is very popular in enterprise but is has not been as widely adopted because of cost and a poor user experience – graphics are too slow. Nvidia says its vGPU technology will solve these end-to-end virtualization graphics issues.
Source: This article was compiled from various reports and official releases from Nvida newsroom. . It is solely a brief overview to update iTWire readers on significant industry news.