We are working with a company that has developed a specific in-memory, GPU accelerated database, which is capable of delivering truly real-time actionable intelligence on large, complex data sets, providing 100x faster performance at 1/10 of the hardware requirement of traditional databases. Your database could potentially address up to 3584 GPU cores available on GPU, compared to the normally capped 22 per processor on X86 CPU 

Organisations are using Kinetica to simultaneously ingest, explore, analyse and visualise streaming data within milliseconds to make critical decisions and find efficiencies, monitor real-time trends and improve customer experiences. This large scale data analytics opens the opportunity for machine learning/artificial intelligence libraries such as TensorFlow, BIDMach, Caffe, and Torch to run in-database alongside, and converged with, BI workloads. 

Via user-defined functions (UDFs), it also allows for the availability of in-database analytics. This industry-first capability makes the parallel processing power of the GPU accessible to custom analytics functions, where data scientists can write their code and register it in the GPU database, eliminating the need to move data sets or orchestrate logic to make it run. Currently, the UDFs support API bindings in C, C++, and Java. Support for Python will be available shortly.  

The company has recently delivered a performance boost through utilisation of video memory, or VRAM. The database resides in regular RAM, but can benefit from a speed-up, by pinning some data to VRAM, thereby bypassing the need to send data down the PCI bus. This boost will complement the NVLink communication protocol also being developed by NVIDIA is in full production, which will further deliver a 4X performance boost over the PCI bus, increasing data delivery speeds from about 20GBps to 80GBp  

We can arrange a 'real-time' demonstration, so feel free to get in touch! If you would like to receive further technical details and a document on today's database challenges, a download is available below.