What's new, pussycat ?
The 3.2GHz CPU is rather plainly faster than the 3.0GHz and 3.06GHz (533FSB) parts on pure clock speed alone. How else, though, have Intel managed to effectively increase the efficiency of these very latest CPUs when compared to the earlier 133FSB variety ?. Firstly, as you may suspect, Hyper-Threading is present here. In simple terms, Hyper-Threading allows the CPU to more efficiently use its resources, assuming one is using a H-T-aware OS. To see our take on the P4's H-T technology, head here.We've touched on the fact that CPU's FSB is now at 200MHz (or 800MHz if you buy into Intel's method of describing their processors. 800MHz is derived from the CPU being able to send data at 4x per clock cycle. Requests to system memory, however, in the form of address requests are still limited to 400MHz). The raising of the CPU's FSB and the ample number of motherboards that are able to run this new speed with performance-boosting dual-channel memory will allow all 'C' processors to tap into a theoretical 6.4GB/s of memory bandwidth. Compare that to 2.13GB/s and 4.26GB/s for single and dual channel (533FSB P4s) motherboard respectively and you see why pure MHz clock speed isn't everything.
Performance can be thought of as the sum of clock speed and efficiency, and a higher CPU and motherboard FSB, especially when the motherboard can run in dual-channel mode, allows greater performance per given MHz. That's exactly why the 3.0 'C' (800FSB), when run on a decent Springdale and Canterwood board, can be significantly faster than the nominally quicker 3.06GHz (533FSB) on either a single (i845PE) or dual-channel (Granite Bay) motherboard. Memory bandwidth plays a key part in a number of real-world applications.
So it's a question of ramping up the total work done by the processor / motherboard combination. Intel can sing and dance about the effectiveness of a 200FSB and dual-channel chipset running, but Joe Public doesn't care much for behind-the-scenes efficiency, really. What they care about is a big, fat number for the processor. Intel oblige with a mind-numbing 3.2GHz. What we must remember is that this processor is still using the same manufacturing process that debuted with the 1.6GHz Pentium 4 Northwood; namely a 0.13-micron fabrication. You can guess just how much heat these 3GHz+ processors put out by looking at the changes to the retail Intel Pentium 4 cooler. It started off as a slim, quiet affair, yet it now resembles the more impressive aftermarket coolers with its copper-bottomed construction and variable fan speeds. We're looking at the wrong side of 80w now, so we don't expect there to be many more, if any, speed increases from the Northwood core that put Intel back in the benchmarking business.
We can somewhat temper this with the knowledge that each and every 3.06GHz / 3.0GHz Pentium 4 we have tested has run past 3.2GHz with default voltage. Intel took their time in launching a 3.2GHz CPU. They believed, rightly or wrongly, that 3GHz clock speed was enough to see of their main challenger, AMD, in the performance stakes. That's why it has been over 6 months since the launch of the previous clock champion, the 3.06GHz (533FSB).
This processor may well mark the demarcation line for the Northwood core, and it will soon be replaced by the much-vaunted Prescott core. With a new, revised 0.09-micron manufacturing process that should see it scale to 4GHz and beyond, and an increased L2 cache of 1MB with 16kb of L1 Trace Execution cache, it should help Intel to combat the very real threat of the upcoming AMD Hammer CPUs, especially if a newer, more impressive H-T tech. is implemented. For now, though, we're stuck with the Northwood, which is no bad thing.
