The Inquirer-Home

Definition of true quad cores proposed

Unofficial official Inq test enclosed
Thu Feb 01 2007, 08:49
ONE OF THE more interesting debates of 2006 was what criteria made a quad core chip. This debate went backwards and forwards with no real resolution.

All sides picked the high points and ignored the lows. Intel was far closer than anyone and AMD took potshots. Intel had something AMD didn't, but forget them. What is a real quad core?

The AMD argument is that it is, purely and simply, a single piece of silicon with all the cores on it. At least one person at Intel would jump out of his chair and shout 'bollocks' at that, countering that if you buy a CPU as one piece, it doesn't really matter how it is arranged under the heat spreader. You buy a CPU, the OS sees a CPU, and you can't take it apart.

Both have merits and down sides. If the MCM approach is valid, then why not FX-7x series, you buy them as a pair, you put them in as a pair, and, well, you can separate them. If you are making any distinction other than single piece of silicon, why stop at the MCM level?

To make matters worse, what about taking two adjacent dies uncut and putting them on a carrier even though they are seen as two different CPUs electrically? They share nothing other than an area that should be cut, and if it did, you would have two functional single cores. Intel did this on some of the early Pentium 4 dual cores, and they fit AMD's current definition quite well.

After much thought, I came up with a two pronged test for it. Well two two pronged tests. First is to take the people arguing this on political grounds out to dinner and stick them with a two pronged fork repeatedly. You can find these at fancy restaurants that won't let us in the door with t-shits proclaiming how society hates us. This test is satisfying, but does not settle much of the debate.

The second one is the real two parter. The first part is whether you could separate the parts into multiple functional chips. Can you make two Woodcrests out of a Clovertown? The other question is a little more esoteric. Can all the cores talk to the other cores without going out over the bus? This is not necessarily a question of 'does' but one of 'can'.

If the answer to the first question is no and the second is yes, then you have a 'true' multi-core product. It can be dual, quad, octal, or more, but it needs to fit those definitions to call it a true X core product.

Right now, everything on the market is a dual core, there are no half-assed duals anymore. There are also no true quads, but Intel is the only game in town, so does it really matter? To a buyer, the answer is no, you buy X cores, and you get Y performance for $Z. To them, Intel is selling quads like hotcakes.

To the geek, it probably matters about as much as the relative difference between Warp 9 on the Enterprise classic vs Enterprise E, IE deadly important. Since this is about the tech and not the marketing, I will stick with my definition. Feel free to make your own, if you must. µ


Share this:

blog comments powered by Disqus
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

INQ Poll

Dead electronic devices to be banned on US-bound flights

Will the new rules banning uncharged devices be effective?