Reviewing Intel's CPU branding approach

INTEL LEADS the CPU market, full stop. More than ever, perhaps. So, why change its perfectly good branding then? After all, Core 2 - and Core 3, and so on - does sound better than Pentium, Hexium or Sexium, and definitely far better than Itanium, all names sounding like some evil Big Pharma drugs.

Aside from a lot of extra marketing and sales work, resulting in marketeers keeping their jobs in these tight times, there doesn't seem to be any real reason to rehash the branding just because of the Nehalem arrival.

The new chips, rather than being called say Core 3, got the brand new Core i7 moniker, in a way reminding us of the old P7 codename. Now, the dual-channel mainstream Nehalems will be called Core i5, and the low-end integrated-graphics parts might bear the burden of the Core i3 brand. And, just like the current i7 with the brand new three-digit product numbers instead of the old four-digit ones, you'll have an interesting time comparing chip models.

So, when you do your holiday system shopping later this year, you could, for instance, choose between the old Core i7 965, the new Core i7 960 or the brand new Core i5 XXX (I didn't say '860'). While all three are internally basically the same CPUs and run at a 3.2GHz clock, the differences will be there. The first one is the old 2008-launch part with unlocked multiplier but C0 stepping. The second one is to be the new part late this year, with locked multiplier but newer, more efficent D or even E stepping. And, finally, the last part will have two memory channels and the LGA1156 versus LGA1366 socket, but faster Turbo mode and of course cheaper P55-based mainboards. Love the confusion?

Many publications commented on the new branding approach, with mixed reactions. There is no clear connection to the old Core 2 branding, and even the product numbering was completely rehashed. Core 2 Quad Extreme QX9770 and Core i7 Extreme 965 cover the same market segment and run at the same default clock and, in fact, aren't that far apart in performance. But there's no correlation at all in the naming. The old Pentium, Pentium 2, Pentium 3 approach was, in this respect, more consistent.

Why not look at something like that, since Intel already went with this BMW-style numbering? The "7" series is the high end, the "5" series is the mainstream, and the "3" series is doing the basic work. An obligatory "X" could be added at the end of any Extreme part in the "7" and "5" series, to avoid having to use different basic numbers for otherwise same-clocked extreme and normal, that is, locked parts. A similar "L" could be added for the low power parts, and an "M" for the mobile parts. Plus, of course, a "G" for the graphics-enriched ones.

After all, at the Xeon front, that's the case already. The "W" parts are top bin workstation CPUs, the "X" parts are for high end servers, the "E" parts are the mainstream offerings, and the "L" parts are the low power workhorses for dense and green computing.

Then, there should be enough numbering in reserve to accommodate the 32nm 'Westmere' parts without changing the i7-i5-i3 sequence. Right now this scheme seems to be a bit tight for the i7 series as we'd only have the 980, 985, 990 and 995 numbers available before hitting the four digits, and that has to take care of the next 20 months at least. Aside from that, the possible i8-i6-i4 sequence could then be left for the Sandy Bridge and Haswell generations.

Talking about numbers, in Chinese, eight is a very lucky number, but four isn't. The last murdered Alpha CPU was codenamed EV8, but was supposed to be really called, umm, the 21464. So, maybe, let's skip any future 'i4' at the low end, eh? µ