Blogpost by Joonas Laitinen, Technical Account Manager, Nextron Finland.
It is back to school time for many of us, so let’s do first post bit easier to understand for all, not depending your computing expertise. This time we are going to focus on the benefits of going for newer and more powerful machinery. What are the actual main benefits for that? To answer it simply, it is only to save time. The only resource that is limited for everyone. This time we are going to go for a very simple example.
Let’s say that we have quite big company, +100M€ / year company. They want to develop their research capabilities, maybe branch bit in to AI or something else compute intensive. So, now they have 3 options on how to actually do the computing.
Option 1, that is quite common is doing the stuff on cloud. That will side-step some of the problems, but it is often quite costly. I have already covered this topic here: https://nextron.fi/blog/onprem-vs-cloud so I am not going to do that here.
Option 2, is the next most common, and the one actually under scrutiny today. That option is to use existing computing resources that are already in place. That does quite often mean rather old machines with rather limited calculation power. That leads to couple things: 1. This is highly power inefficient, since you have to run the calculations for really long periods. 2. You will waste time, time that could be used to gain advantages over your competitors or leverage completely new products or improve your service models based on automated processes.
Also, you are going to have your datascientists and other highly specialized personnel to just wait for machine to do its work, before they can do the next set of calculations. That might be really bad, since there certainly are takers for people highly skilled with AI related tasks. If they feel that they are not working in the best possible environment, they might go somewhere else, and you are now even deeper in the hole than you were before. Even if those people stay, it won’t be smart to pay them to wait for the calculations to proceed. If you saved like 50 000€ for the acquiring the new hardware for the research people, calculate how fast it would save its own costs due to people doing more actual work, and less twiddling their thumbs and waiting for the calculations to be complete. You might be surprised how quickly you are going to brake even.
Option 3, that I suggest to consider: look at the costs of new Hardware. It is not that expensive, if you are factoring in the stuff listed under Option 2. Right now, there are more developments in the computing field than never before.