The Bottom Line: While the Pentium 4 provides speed, new features, and a wad of performance boosts, the price, and lack of support for feature sets make the P4 an extreme disappointment.

Pros: The P4 makes the first foray into a range of new technologies.
Cons: Price, performance problems, and instability frustrations.

As we move on in computer technology, there will inevitably be a sour lemon in the line of a components company. For 3dfx it was Voodoo 3, for Cyrix it was every processor they attempted to build, and for Intel, I am afraid this might well be thier first truly failing foray into the processor markets.

Overview
My first contact with the processor was in its design stages, two of my contacts were recipients of beta systems and they allowed me a rare peek at the system and its performance. True, I was amazed. Used to working on an old AMD K6-2 450 machine at work, this computer with a brand-spanking-new pentium 4 beta was phenominal and quick but with the usual beta bugs (massive instability, system freezes, short hangs, etc.). As the project progressed and the processor was finally released, I was more and more impressed with the speeds, feature set and capabilities that Intel was boasting -- until I got a post-release P4 into a desktop test station at work.

Features
Now, in order to find out if the P4 would be a viable upgrade option for our work computers, we needed to know exacly what we could expect the pentium 4 to support and how that would benefit the company (after all, if it has fancy bells and whistles but they dont do anything, what good are they?).

The first feature to be noticed was RDRAM support, the ability to support the memory DIMMS on a proprietary system supporting 2bits per clock cycle through put, running at 600MHz (the OEM shipped with the 600MHz type and there was no way to upgrade it pre-shipped). We will come back to the memory itself later.

The floating point units on the Pentium 4 were about mediocre. While I really cannot say much as to the weakness of them, I also cannot rave as to thier strength. While they do provide adequate support for "normal" application needs, they are well short of the revolutionary power that intel has been boasting thier processor would be able to sustain.

The next feature to be looked at was the new SSE2 (Secure socket encryption featureset 2) which is spec'd to provide processor-level encryption of data transmission for stronger algorithms, quicker encryption times, and generally more secure data transfer. The potential advatages were obvious from the start. SSE2 support meant that the big check cutters in the business office could finally order office products online with ease, without the worry of hackers getting access to our expense accounts -- or so we thought. The hardware was a good idea and even works as spec'd (which is actually unusual for Intel) but there are not yet many internet applications that support coding for the SSE2 instruction set. Thus, it may be a while before having the ability to use SSE2 will become more of a reality.

Coming down from our disappointment over the realization that SSE2 really wouldnt do much for me as yet, I went back and worked with the memory. I took out the slower 600MHz RDRAM and replaced it with the faster 800MHz RDRAM, figuring that I would console myself with the faster speed.
The speeds were delievered as promised, a full 800MHz ram speed which noticeably increased processor-memory intensive tasks but did very little for any task where the speedy ram information needed to be crossloaded to a PCI, USB, or AGP peripheral. (The bottleneck problem: though the memory is able to run at speeds of 800MHz sending information accross the bus, the bus runs about 300 or 400 cycles per second slower and then the peripheral busses run about 200-300 cycles slower than THAT per second. So even though front end speeds are fantastic, the physical abilities of the subsystems really cannot take advantage of the extra speed).

Speed
The processor actually is able to put out between 1.3GHz (that is 1 billion, 300 million instruction cycles per second) or 1.5GHz speeds. Benchmaks (Benchmark 2000) indicated, however, only a 32% speed boost under windows 98se, and only a 34$ speed boost running in the windows 2000 environment over a system running a 933MHz Athlon. The speed is quite good but disappointing since the processor is supposed to be able to do much better.

Price
I was unable to order a Pentium 4 processor as is and instead had to buy a full OEM package through Dell which was over $1600 for a "budget" system that was only running a 1.3GHz Pentium 4. The prices are quite high right now and since the Pentium 4 requires RDRAM (which is extremely expensive right now itself), I highly doubt that we will be implementing many of these desktop systems into our office in the near future.

--- Update 11/15/01 ---
I have recently come by some industry rumors that Intel plans to dump the original socket set employed by the Pentium 4.

Apparently the current number of pins is too low to accomidate the throughput levels neccessary for 2GHz+ processors due out 4th quarter this year. According to my source, Intel will release a second iteration of current-speed P4 CPUs for the new socket set in order to offer P4s to lower-end PC markets while still allowing for upgradeablility. This change should be much like the socket change for the Athlon to the Athlon Thunderbird.

It would be my hope that Intel will be able to emulate the perfomance increase that the thunderbird showed over the original Athlon. Otherwise, throughput boost or not, the Pentium 4 will still be plagued by the same faults as the current iteration is now.