One of the more unfortunate post-MHz war repercussions was the empowerment of number-loving marketing folks to name product lines however they damn well pleased. Coupled with the increased market segmentation and binning practices we have grown accustomed to, consumer CPU shopping has become a research laden nightmare for the layman, and in the server space where you have even fewer options to choose from they’ve somehow made it even harder.
The cubicle dwellers at Intel (INTC) HQ have hatched a new scheme that attempts to cram even more information into a single model number. According to an Intel blog post, the new scheme breaks down as follows: [ Using the following model number as an example: Intel Xeon E7-4820L v2 ]
- Brand – Xeon
- Product Line – (E3, E5, E7)… or eSomethingMeaningless if you prefer
- Product Family – (A four digit number that breaks down into several parts, 4820 in this example)
- The ‘4’ represents how many sockets are supported (can be 1,2,4, or 8)
- The ‘8’ represents socket type where the higher the number, the more capabilities that socket supports, such as memory support and I/O speeds.
- The ‘20’ at the end represents an almost arbitrary SKU number such as 20, 25, 30, 35, etc.)
- Optionally they might paste an ‘L’ at the end of it all to indicate a low power variant
- Version number – (v2), this is similar to appending the ‘2’ in front of Sandy Bridge model numbers to indicate that they are second generation Core iSomethingMeaningless chips.
Sounds kind of logical so far, and if you have this cheat sheet in front of you while chip shopping it might not be so bad. The inclusion of features such as Hyper-threading and Turbo-boost seems to span the entire E3/E5/E7 range, with the other variations not included in the new naming scheme coming in the forms of core count, clock speeds, TDP, and cache size.
The part that intrigues me is the second digit in the product family, or socket type, which should translate across generations. For example if a v3 (Ivy bridge based Xeon) is released with the same socket type number it means it should be a drop-in compatible upgrade. After a several generations of I/O and socket changes they will inevitably run out of numbers, but that will be a few years down the road when we’re all running ARM based servers, flying around in hover cars, and ordering our tea from a food replicator. Earl Grey, hot.
From the outset of this article I wanted to lampoon the naming scheme employed, but after studying it, it makes sense to me. From a business perspective, if your system admins can’t be bothered to research the chips going into your critical equipment, you’ve likely got bigger problems than a silly sounding processor name. If this naming scheme was employed on consumer chips however it would be a whole different story.
The only major criticism I must offer is the omission of core count in the model names. The name is so long already why not add a simple -4, -6, -8, -10, etc to the end of the product family to indicate core count and actually be of some practical use? I for one welcome our new Xeon E7-4820L-10 v2 XTX-PE GT3 RS overlords. S|A