Edition 4.5
October 22, 2017
There are a few large Technology trends that – we think – are all about to smash into each other in 2018 and beyond.
Source: Internet media
- Analytics is becoming a standard feature across tech: Analytics is morphing from a standalone sub-segment of the software industry to an embedded function inside every application, software, and machine, i.e. ‘table stakes’ in tech.
- The age of AI/ML takes analytics to the next level: AI/ML is an attempt to use existing data analytics to move up the curve to predictive analytics.
- The supply side of available compute power is going places the industry has never gone to before: We have been in a fairly heavy investment cycle in high end computing, including CPU, ASIC, FPGA, and GPU.
What will it all mean?
[Note: Rmax - Maximal LINPACK performance achieved. The LINPACK Benchmarks are a measure of a system's floating point computing power. Introduced by Jack Dongarra, they measure how fast a computer solves a dense n by n system of linear equations Ax = b, which is a common task in engineering.]
The slide above comes from an investor presentation by Intel on a May 2011 European NDR. The NDR featured an Intel server fellow that we had not heard of before, nor since, and this slide is a prediction of the amount of maximum computing power available from the top 500 most powerful commercially available computer systems.
The growth of maximum computation power seems to be coincident with the rise of available analytics (including predictive/ML) embedded as a standard feature across the technology landscape. We think the next few years will see an explosion of analytics. The supply side is already prepared: CPU/GPU are in place + software programs are written and widely available. The stocks of the enablers have already gone insane.
So who wins from the next round, namely, a period when large scale analytics combined with a big jump in available computation are all readily available? What applications will suddenly take shape and alight – think: things that could only be done previously by a large mainframe plus a serious amount of custom software, will now be available maybe at the push of a button from multiple sources, and maybe with instant results? Who are the winners in that field?
Source: Convergence of Machine Learning, Big Data, and Supercomputing, by Dr. Jeremy Kepner MIT Lincoln Laboratory Fellow
