Kinda agree with you guys.

The power consumption of a multi-core processor isn't always necessarily higher. Allow me to explain. Remember that the clock speed of a quad-core cannot be as high like a dual-core because of thermal reasons and the TDP constraints. The objective in a multi-core processor is not about extremely high clock speed but rather multiple cores of reasonable clock speed.

With the progress of silicon fabrication and newer architecture, its the efficiency of the processor architecture and how good the software utilize these facilities that matters. So a reasonable clock speed in a multi-core environment coupled with advance fabrication process and modern architecture will not always consume more power than say an extremely high clock speed of a dual-core processor.

This is synonymous with Intel's Core (1 or 2) architecture. Wide execution and higher performance per watt ratio. We can already spot a trend over here. The architecture design is slowly moving towards more transistor budget allocated for caches and lesser for execution units. From the early Pentiums to Intel Core 2 Duo, cache increment was very obvious. Look at the development trend of software. Now look at the console gaming hardware. Its a very ideal platform to test the market and your technologies.

XBOX360 has a tri-core general purpose in-order with an extremely simplified execution unit architecture that is at the same time multi-core and yet extremely high clock speed. Look at PS3, a single general purpose execution unit (almost similar to XBOX360) and with up to 8 (only 7 active) dedicated processing units which is again extremely simplified cores.

Its all about parallel processing from now on. Making a general purpose processor more parallel with dedicated purpose processing in its execution units for acceleration. Its all about simplified and efficient architecture and multi-core in nature. IBM and Sun's UltraSparc T1 processor has been leading the way to parallelism. AMD is going with its Fusion and we can see Intel will definitely progressing to it too with its simplified 80-core experimental processor. Its going to be multi-core with multiple acceleration units for the future. Just remember that a DSP is more powerful and consumes lesser power than any general processing platform for a given dedicated task.

This post has been edited by davidmak: Jan 28 2008, 08:25 AM

The most important thing is the segment of cheap performance per price ratio will always be there from both AMD and Intel. Doesn't matter if today Q6600 is cheap today and obsolete tomorrow. There will always be such a product available for consumers who want something new for a good deal. The value for price product will always be consumer's highest priority and the manufacturers will not leave it out of their offerings. 101 marketing rule, do not reduce price but introduce higher value by giving more features.

During the introduction of dual-core processor, Intel offered Pentium D 805 which was best bang for the buck and OC-able. Then we have a few value for money dual-core processor which was on the Core 2 architecture. Today, that same segment is now represented by Q6600 for quad-core processor product. In a few months, we will have Penryn replacements for quad-core which will be very popular.

Today, software for multi-core (not only quad-core) support (at least for what I am doing) is matured. By the time it is widespread to consumer software, I would have something new to replace Q6600 because its gonna be cheap as chips. Again, I can still hold on a few months or a year or two with that Q6600 because at least I'm still using it, squeezing every last value from it. I don't think you could do the same with a single-core or a low-end dual-core processor though. You'll be on the verge of upgrade sooner. I don't care if my Q6600 is obsolete and become of no value, at least I am still using it for its intended purpose, multi-core processing. At least I will be satisfying the requirement of my work.

I think this is the confusion with the software written for single thread or multiple thread processing. Like you say software designed for single core processing and multiple core processing. I really think there's a good way to solve all these confusions. Right now we have problem with single thread processing software not scaling effectively with a multi-core processing. This problem can be worked around with 'reversed Hyper Threading' concept.

Hyper Threading was a concept to make the system acknowledge 2 processing cores where in actual fact there is only 1 real physical core. The fact it could handle more than one thread is because of its additional set of registers to hold an extra thread. The processing still goes one by one but with Hyper Threading you could ensure less idle states in the registers when work is actually done.

Reversed Hyper Threading however is still a very theoretical concept in which to make the system acknowledge multiple processing cores are one. Where two or more cores could combine to speed up a single threaded process. Some of the program instructions sets that goes into multiple level of processing could be done in parallel with the cores assigned. Or the system is smart enough to assign a few cores for this process while the remaining does the rest.

One of Intel's testing approach is the Core 2 architecture where it attempts to combined a few similar instruction sets to be executed as one. This is pretty ironic because the CPU always breaks instructions into micro instructions for 'easy digestion' but these micro instructions can be combined to be executed at once.

Yeah agree. Especially the current AMD owners using AM2/AM2+ motherboards. Its a cheap upgrade to the quad-core experience and the associating performance boost. Although the performance isn't as what we all expect to be, but its a nice performance boost. Significant I would say, nothing fantastic however.