Dual-Core Processors

Computers don't function by magic. As you know, it is the CPU that does the real work.

Over the past few years, we have seen processors speeding up from Mega-Hertz to Giga-Hertz. But for sometime, the manufacturers of these chips have known that this cannot continue for long. When we, as desktop computer users, find the need for a higher-end processor, we end up spending a pot of money on acquiring a new processor and sometimes a new motherboard. They have been trying to find alternative ways of providing us with that extra speed we require for our ever-growing software, at a lower cost.The latest buzzword in this direction is 'dual core'.

Dual core is not really something very new. It has been around, on the scale of multi-processor computers. It is only that now it is seen as a viable alternative to increasing processor speed. Let's examine what this is all about and what it can do for the desktop computer users, whether on the Intel, Mac or other platforms.

However, dual core isn't the end of the road either, some manufacturers have come up with quad and even 16-core CPUs.

Basics first



The processor's internal memory called 'cache' stores the immediately next set of instructions and the data necessary to process what it is running. It also has some special locations called 'registers' where the instructions are executed. Very often, the data in the cache and the registers need to be 'refreshed' and this is controlled by its own clock. All this together is called its 'core'. The pathway between the different parts is called the 'bus'.

The faster the computer needs to run, the faster all of the core and bus have to work. Speed generates heat which can sabotage other components. This is where the heat sinks and fans you see inside a computer's cabinet come into play. And

heat puts a limit on why present-day desktop CPUs cannot infinitely speed up.

Enter 'Dual Core'. The answer to this puzzle lies in combining the cores of many individual processors into one single chip. This has given rise to dual, triple, quad and even multi-core processors.

How it works



Manufacturers have different names for their technologies. Essentially, this is a similar technology that is used to make multi-processor systems work. The technology combines the cores of two CPUs that would have normally run in a multi-processor setup, where all the buses between them that would have connected to the second chip are joined. This also solves the heating problem since the number and lengths of the buses required are lesser. Costs also come down, as you need to put less material onto the new chip than into two single core CPUs.

Apparently, this picture is not very rosy. In a computer, access and manipulation of its memory is controlled by a 'memory controller'. Since you have two caches, it is obvious that it would have their corresponding (ie, two) memory controllers. However, it has been found to be impossible to leave both of them active. Therefore, it is necessary to disable one of these controllers on a dual-core processor. What this translates into is a slight performance handicap, making it not as fast as a real two-CPU system.

Some questions/answers



The first question that crops in our minds is whether it will work with what we already have. Depending on a particular manufacturer's technology and future vision, some of these new chips have the same size and number of pins on their underside as their nearest single-core cousins. If they are, they should work on the same old motherboards. Intel says that its new chips will work, if you have the Intel 915 or 925 series motherboards. While AMD recommends that you have the 940-pin socket used with their Opteron processor range. On the IBM PowerPC platform (used on the Apple Macintosh), however, because of the size of their new chip (although based on the 970 series) it will not even fit into the same boards.



In order for them to really make a difference in our lives, the software must be able to identify and recognize the two cores as two different processors. At least the OS we use should be able to do this. Luckily, most software that will require the new power already use similar techniques and not many changes will need to be made to them-Windows, Linux, RDBMS (Oracle, SQL Server), graphics tools (Adobe Photoshop), animation and rendering tools (3DSMax) to name a few.

The present definition of a 'CPU' could change yet, because of the way both computer and software vendors will look at the dual-core processors. A lot of them price their products according to the number of CPUs. Till now, this picture has been clear with a 'CPU' meaning a single chip. Should individual cores be treated as seperate CPUs? Or, should we treat them as a single CPU simply because they reside on a single die regardless of the number of cores?

If they treat it as a single processor the prices should remain independent of this debate. Interesting developments have already sparked off in this respect and are worth monitoring over the next year. Companies such as Microsoft and Novell have assured customers that they will see the new chips as single processors. Oracle insists on seeing them as two processors meaning higher software prices. RedHat has pulled in a monkey by saying it will license it at one and a half times the current prices.

So get set for an interesting year, since end-user versions of these new chips will be available only by the end of 2005 or early 2006. And what all of us have learnt about CPUs and have been brought up with, is going for a toss along with the licensing policies. Is it worth the upgrade? Perhaps they will iron out the performance problems and come up with some measurable

advantages over the single-core systems. For now, it remains a buzz.

Sujay V. Sarma and Sushil Oswal