Details
NVIDIA Tesla K40 is the world’s highest performing accelerator ever built, delivering extreme performance to a widening range of scientific, engineering, high performance computing (HPC) and business applications. Providing double the specific memory and up to 40 percent higher total performance than its predecessor, the Tesla K20X GRAPHICS gas, and 10 times higher performance than today’s swiftest CPU, the Tesla K40 GPU will be the world’s first and highest-performance accelerator increased for big data analytics and large-scale scientific workloads.
Featuring clever NVIDIA GPU Boost technologies, which converts power headroom into a user-controlled overall performance boost, the Tesla K40 GPU accelerator allows customers to unlock the untapped performance of the considerable range of applications.
“GPU accelerators have gone nicely known inside HPC and supercomputing industries, enabling technical engineers and researchers to continually generate innovation and scientific discovery” said Sumit Gupta, general manager regarding Tesla Accelerated Computing goods from NVIDIA. “With the particular breakthrough performance and higher memory capacity from the Tesla K40 GPU, enterprise customers can quickly crunch by way of massive volumes of details made by their big info analytics applications. ”
Greatest Performance for Science, Massive Information
Based on the NVIDIA Kepler compute framework – the best efficiency, the majority of efficient architecture ever built : the Tesla K40 GRAPHICS accelerator surpasses almost all other accelerators on two common measures of computational performance: 4. 29 teraflops single-precision and one 43 teraflops double-precision peak suspended point performance.
Key features of the Tesla K40 GRAPHICS accelerator include:
12GB associated with ultra-fast GDDR5 memory allows users to process 2X larger datasets, enabling them to rapidly analyse massive quantities associated with data.
2, 880 CUDA parallel processing cores provide software acceleration by upward to be able to 10X compared to manage to by using a CPU by yourself.
Dynamic Parallelism permits GRAPHICS posts to dynamically offspring new threads, enabling consumers to quickly and very easily meltdown through adaptive in addition to powerful data structures.
Pci-e Gen-3 interconnect support accelerates information movement by 2X in comparison to PCIe Gen-2 technological innovation.
NVIDIA said the processor would be quad-core in the beginning stages but they denied revealing any of the finer details of the processor. A couple of days back NVIDIA published white papers pertaining to the architecture of the processor. The papers reveal that the processor actually has a fifth core running parallel with the other four cores. The secret fifth core is called "companion core". All the five processor are ARM Cortex A9 processors and can be separately activated or deactivated based on work load.
Tesla K40 GPU is going to work on CUDA parallel processing cores. CUDA parallel processing cores technology can be used for optimising the CPU in multi-threading scenario and also introduce multi-tasking functions for efficient performance and better power utilisation. The companion core i.e. the fifth secret core is built transparent to the OS which means that the OS and the applications running will not be aware if the fifth core but can take advantage of it during run time.
The operating frequency of companion core is limited to 500MHz while all other four main cores are left to function at their full throttle of 1.5GHz. The companion core is mainly meant for less demanding applications that run in the background like animated background, Twitter app, Facebook, push mail service, background apps etc. Thus the architecture leaves the four main cores to do their work at full potential for maximum results while doing system heavy tasks like gaming, video editing, image processing etc.
Menu
