Motherboard Finder

Disclaimer

Please do not email me about this project without reading the disclaimer above.

This program helps you find motherboards for sale that best meet your criteria including price and lets you easily compare any two products. It can automatically generate feature comparison matrices, highlighting the differences. The same program could be used to find/compare other types of computer equipment such as SSDs (Solid State Disks), CPUs (Central Processing Units), hard disks, laptops… It could be used on anything that has specifications including cars, heavy equipment, cellphones, laptops. It could even be used as part of a website that advertised or sold used equipment.

This project has two parts. Collecting data about motherboards from manufacturers and retailers and providing an Applet to the public to summarise that data.

You need to screen scrape websites to collect information about various motherboards. The best sources are the motherboard manufacturers themselves: ASUS, BioStar, ECS Elite Group, Gigabyte, MSI and Tyan (Thunder, Tiger, Trinity, Tomcat, Toledo). You can use the HTTP package to scan the websites and collect the raw data. You can use regex expressions to extract what you need. You can use the entities package to DeEntify or flatten the HTML (Hypertext Markup Language). The facts you are after include: form factor, CPU (Central Processing Unit) socket, CPUs supported, max RAM (Random Access Memory), type of ram, RAM speed, type of onboard video, number of USB-2 ports, number of USB-3 ports. plain or theatre sound, number of SATA-3 ports, number of SATA-III(6G) ports, major vendors carrying the board and price. What types sizes and speeds of cache come with the motherboard? How much power does the MB consume? This data must be kept up to date by periodically rescanning to find new products and new specifications and adjusting your webscraping techniques to keep up with the web page formats. You might lobby for manufacturers to provide the data in computer-friendly CSV (Comma-Separated Value) or XML (extensible Markup Language) SOAP (Simple Object Access Protocol) format. I think Amazon already has a computer-friendly API (Application Programming Interface) you can sign up for. You might use Google to help you research retailers. You might consider collecting photos as well.

To use this database, the user tells you facts about what they want: e.g.

• What is your budget range?
• Desired form factor e.g. ATX (Advanced Technology Extended), microATX.
• Do you have a preference for AMD (Advanced Micro Devices) or Intel?
• Are you planning to use this for gaming, accounting, business, word processing etc.?
• Do you need theater sound?
• Questions for the more technically oriented: minimum USB-2 ports, minimum USB-3 ports, minimum firewire ports, is DDR-3 RAM a necessity? how much RAM capacity do you need? Preference for video chip vendor.

It then orders boards by bang per buck, by bang or by price that meet your requirements. The user then might adjust the budget range up or down or add or drop requirements when he sees just what you can get for what price.

To make money from this, you need to be an affiliate/associate with various retailers such as Amazon.com or Tiger Direct who will give you a referral fee if someone buys a motherboard, not necessarily the one your program recommended, after you refer them to their site. You can take your visitor directly to the retailer’s page on the manufacturer or retailer’s site using Applet.getAppletContext().showDocument( url );

Make sure you give a disclaimer about the accuracy and timeliness of the information, reminding the user to double check the information about the most likely candidates with the manufacturer.

You could also show two motherboards side by side, with the better specs of each board highlighted. You could consider this as motherboard comparison shopper if you had enough retailer prices included. There are so many almost identical motherboards out there, you need automated help pointing out the differences. You could optionally suppress identical data.

You could build a prototype for this project by manually collecting data for some small subset of motherboards e.g. microATX AMD from Amazon.com or TigerDirect.ca.

ASUS has designed their website to make screenscraping as difficult as possible. They use a potent combination of JavaScript and SSL (Secure Sockets Layer) encryption to defend it from screenscraping. I wrote them asking them to make their data available in computer-friendly form. They wrote back saying they are working on a major revamp of the website. BioStar, MSI and Tyan have made theirs quite easy. Gigabyte have the complication of using JavaScript/Ajax to format structured data. You can sort out how it works fairly easily with Wireshark. ECS Elite Group is of intermediate difficulty since the specs are written in narrative English rather than point form.

You might store your collected data in XML files one per motherboard or in an SQL (Standard Query Language) database. You need a way to easily handle missing data and a way to introduce new fields later. I suggest you prepare a resource made up of gzipped Motherboard objects in an array each with a dependent array of Price objects. Each price object contains a vendor name, URL (Uniform Resource Locator) where the vendor displays the product (one that can be easily converted to an affiliate URL that gives you a finder’s fee), a price, a currency that the price is in and a date when the price was current. See PowCost for code to convert prices into a common daily local currency.

You might publish your list on a website as a CSV file, showing only the best price for each item. If you created a SQL database, you could implement this as a Servlet, so that the end user would not need to install Java locally to use it.

If this project caught on, you could conceivably have smaller vendors coming to you, asking to be listed. You could say, just email me with an attachment that looks like this from time to time, telling me your price list of items, as a CSV file: Motherboard manufacturer, model, price, currency, url, instock? and an XML file describing information about the retailer, e.g. location, phone, website. You could eventually list price lists of even corner store retailers.

The trickiest part of this project is inventing a bang number for each motherboard to use in computing bang/buck. You have no benchmark data to use. You might do it with multi-regression analysis on the price and its factors. You might use a neural net to come up with an average price, or you might try to come up with some ad-hoc formula. You might not even try to create one, just display actual price. You might assign it one based on the average price of the ten most similar boards.

Feature Matrix

You will find vendor or manufacturer websites that lets you compare two or three of one manufacturer’s products (e.g. motherboards, SSDs, hard disks…) They generate a grid with one column for each products and one row for each feature. The typical problems with these displays are:

• It is up to you as customer to notice the similarities and differences. The differences are not highlighted in any way. You still have to read all the fine print in all the boxes.
• It is useless for comparing products from two different manufacturers.
• Often even after you read the entire list, you can’t find any differences. Their software is not self-proofreading to avoid this blunder.

How could you avoid some of those disadvantages?

• Highlight differences in columns with a different colour, font, size etc. in the detail text
• Display a ( same ) / ( different ) / ( better ) icon for each feature between each pair of columns.
• Use a vendor-independent MDL (Motherboard Descriptor Language) to describe the products.
• Write a proofreading program to ensure you have collected enough data about the products to ensure all products in the database have unique attributes and that you have correctly coded each attribute for each product. You could do a sanity check on cost vs cost estimated from features. A product with better features should probably cost more. You could do this in a rough generic way by assigning relative cost importance to each feature.

If a query brings up no matches, you might automatically relax constraints until you get at least one match. In other words, matches are not binary yes/no, but a goodness of fit, so you show the best matches in order.

Once you have tackled motherboards, you might take on video cards, headsets/headphones, keyboards, HDTVs (High Definition Televisions), routers, cellphones… You want products where there are an overwhelming numbers of brand/model choices and simple specs to categorise them.

standard footer

	This page is posted on the web at:	http://mindprod.com/project/motherboardfinder.html
	Optional Replicator mirror of mindprod.com on local hard disk J:	J:\mindprod\project\motherboardfinder.html
	Please read the feedback from other visitors, or send your own feedback about the site. Contact Roedy. Please feel free to link to this page without explicit permission.
	Canadian Mind Products IP:[65.110.21.43] Your face IP:[3.22.68.29]
Feedback	You are visitor number