Using NLPP to find the best valued Intel Broadwell-EP processor in 30s
Intel just published pricing for the new Broadwell-EP (Xeon E5–2600 v4) generation of server processors.
Since we are experts in price / value analysis I just wanted to know which processor type gives the best value for the price. Of course, I used our NLPP (Non-Linear Performance Pricing) analytics solution for this.
Here I show you how I have identified the processor type with the best value for the money in less than 30s.
1. Import Data Preparation (10s)
The input data was scraped from here and looks like this:
To get the data took about 5s. Mark, copy, paste into a Google sheet, done. Then I just split column C with the SPLIT function to get the clock frequencies as numbers. Overall about 5s.
2. Prepare Analysis (13s)
Next I imported the data, which took about 3s and defined the performance drivers (or price drivers) which I can later use in different calculations, which took 10s. Here is what it looks like in NLPP:
3. Calculate Model (2,34s)
Since NLPP finds the best regression model automatically and eliminates price drivers that have no impact on the model quality automatically, I just selected all performance drivers and let NLPP do the job.
This is the ultimate & optimized best mathematical model one can get from the data. It’s a non-linear model. Overall this step took about 2s to select the price drivers to be taken into account and 0,34s to find the best model.
4. Find processor with the best performance price ratio (3s)
Since we want to find out, which processor model gives the most value (=performance) for the money, we just need to spot the type that has the biggest discount to the should cost benchmark (blue line).
It’s the one indicated in the center of the screen. The red line shows the worst practice benchmark (nothing should cost more than this) and the green shows the best practice benchmark (everything below this line is a very good deal).
To be faster I used the detailed result table, sorted by savings (the biggest savings on top as a negative number, biggest extra cost at the bottom as a positive number) and clicked the last record. This highlights the data point in the graph. Here is the table:
The 4th column shows the absolute potential per type. The marked line means, that the “XeonE5–2680 v4” is priced at $1,745 but should cost $276.633 more based on the pricing for all Xeon types.
And this means, that with this processor type you get a performance that is worth $276.633 more than what you actually pay.
To read the explanation takes much longer than it took me to identify this specific type. Which was about 3s.
Conclusion
It took me about 30s from end-to-end analysis to find out that the “XeonE5–2680 v4” gives the best value for the money.
That’s why NLPP is such a powerful tool. It’s fast, very fast.
Oh, and, it works for all kinds of parts you can imagine. Just give it a try…