Thursday, May 01, 2008 Ng: Development of processors By Wilson Ng Wired Desktop
WHEN the personal computer was introduced, the first was an 8-bit computing of the original Apple II. Then there was the 16-bit computing with the IBM PC, and ultimately scaling to 32 bit with the 80386 and higher processors.
While it took less than 10 years to scale from 8 bit to 32 bit, we have been stuck in 32 bit in the last 20 or so years.
Years ago, people predicted we would go to 64 bit. It would be more powerful and scaleable. However, while Intel for instance, introduced their Itanium pure 64 bit years ago, the migration did not happen as predicted.
There are reasons for this. One of them is that until it hit volume production, it was still very expensive. But mainly, we would need to re-compile software (with some minimal changes) in order for it to run in Itanium.
Then AMD, Intel’s rival, aced the latter by introducing the Athlon and Opteron, which were hybrid 32/64 bit processors. This means it was not pure 64 bit (internally it was 32 bit), but it could address a 64 bit memory space.
To regress a little, a 32 bit computer—like the Pentium—has a 32 bit address space, which, theoretically, means that your computer was limited to a maximum of four gigabytes of memory.
These new processors took out the limitation, and allowed it more (I have now seen computers with 16 gigabytes, and even 32 or 64 gigabytes of memory).
So the user was able to enjoy one of the benefits of 64 bit computing, which was more memory address space. What’s more, you could use your existing software without change or re-compilation.
But still it could not be denied that while you have more memory, you also need more power. What the processor manufacturer did was to introduce dual core and quad core. This means there are several ways to boost speed. One is to use more memory, and the other is to increase the cache me-mory. Both were done. The other, of course, is to have a faster processor.
What the manufacturers did was that instead of making the processor faster, they put two or four processors into one core.
So if you buy an Intel dual core processor, this means it has two processors inside. If the unit says that it is an Intel core 2 duo, it does not only have two processors, but it has x2 addressing space. That means, combined with the right software (64 bit Windows), you can now use the computer with more than four gigabytes of memory, and it would work.
The strategy has been well received. However, AMD squandered some of its lead when its Barcelona 4 core processors were delayed several months due to some glitches.
How fast was the adoption of multi-core processors? Well, Intel recently announced that it has sold over 10 million quad core processors already.
Can you imagine how far we have gone? We are now using processors that are thousands of times faster than the original IBM PC.
In a recent seminar, I was informed that the latest Intel processors have more than 820 million transistors that are less than the size of a one peso coin. Each circuit is the size of 45 over 1 billionth of a meter! The other stat I got was that the AM radio has only less than 10 transistors for it to work. So a processor is technically almost a hundred million times more complicated and powerful!
A few weeks or months from now, we are almost sure to get something even more powerful. I am sure it will not be long when a billion transistors would be there. Not so long ago, we could hardly imagine having terabyte hard disks (one terabyte is 1,000 gigabytes), and now they’re here.
