The Flatlining of Supercomputing...And What It Means
Have you ever thought about supercomputers? How about the trend they are on?
This is something important because it shows how things operate in the world of technology. By following the progress, along with alterations to the norm, we can surmise what will take place.
In this article we will cover the trends and then focus upon how things will change in the future.
Supercomputing Trends
Supercomputing shows far things have come. We also see some of the shifts that are taking place. This is data from Top 500 which tracks supercomputing stats.
The middle set means little. It is the top computer in the world.
When we look at the blue line, this tells a story that could be read in a couple different ways. The first is the number 500 supercomputer is getting less powerful against the trend.
We could see this as a shift from the norm. This could be an issue or it might be a normal course of action. When we look at supercomputing-as-a-service, this makes a great deal of sense. Smaller companies do not require as much power. If these supercomputers, at this end of the market, are used by these entities, then the drop-off makes some sense.
What is imperative from this chart if the top line which is the total computing power of the top 500 supercomputers. Here we see a huge change from the past trendline. What this tells us is we are seeing a slowdown in the computing.
Gap To Close
Here is the same chart with a couple of trendlines added. It shows the gap we are dealing with.
It looks small on the chart but we have to understand how big a difference this makes. We are dealing with a logarithmic scale.
Does this mean we are headed for a slowdown in the growth of computing power? Not at all.
In fact, this chart tells us that we are likely to see an acceleration. The reason for this is because we are going to see a return to the norm. Technological trends have a reversion to the mean. So while Moore's Law could be slowing down, we will see another shift in the future. Moore's Law is not the first period of computing, it is actually the 4th paradigm.
This means we can expect number 5 around the corner.
We are dealing with a chart that is more than 30 years old. This is not a short term trendline.
Here is what the chart will look like with the "recovery".
At some point, we have to see a turn like this. Whatever the technology, the long term trendline will be met. It is likely going to take another paradigm shift for us to arrive at that.
This is par for the course for technology. It means we are in for an adjustment in the upwards direction. The drop means we are losing a great deal of economic productivity. Imagine what could be done if we had the additional compute. That is what will happen when it is recaptured.
So while the chart, at first, looks like a setback, it is only part of the larger picture. Technology never fades in aggregate. Instead, it is always forging ahead, even if isolated technologies do fade (and get replaced).
This is a part of what will usher in the Age of Abundance.
Posted Using InLeo Alpha
So supercomputers provide the opportunity for us to increase productivity. I'd anticipate every modern nation will be trying to lead this tech to advance their GDPs and economic success.
They already are although supercomputers are no longer the realm of governments. Tesla is working on one. Google and Meta both ordered a ton of H-100s last year from NVIDIA. I have no idea if they are building a supercomputer cluster with them or they apply to separate data centers.
Supercomputing is an indicator of overall computing. We see distributed computing picking up steam as technology advances. However, the big boys are still a metric to follow of how the industry overall is advancing.
I wonder if some resources have been pulled away from the supercomputer race to fund the AI race and they are betting on that AI to then benefit the supercomputer once things are at a certain level on the AI side. Could be an interesting thought!
There is no difference. AI supercomputers are possible. It is only a matter of the chip. Is it a general purpose or AI chip.
Tesla is building out Dojo, a supercomputer designed for neural network training. It will be a supercomputer because of the total amount of processing yet is not general purpose.
time to have a super computer named taskmaster O.O
It's like a different iteration of supercomputing is ahead of us, one that will be more advanced and sophisticated than we have now. I think it will merge well with future AIs.