Intel Keeps Its Promise with Prescott

By Rob Enderle  |  Posted 2004-02-06

Intel Keeps Its Promise with Prescott

From an IT perspective, there isnt much not to like in Intels new "Prescott" microprocessor. Its more efficient than the old processor and it slips, almost unnoticed, into an existing system architecture, which means that there are no image changes required. And while it is always prudent to do a qualification test, the risk if you dont is about as small as it has ever been. This promise kept to IT is one of the things that continues to differentiate Intel in the IT space (only NVIDIA is making an effort to match this promise.)

Kind of like a car that gets a turbocharger, the changes to the chip have little to do with how code is processed, but rather with how quickly things are done. The most obvious change is the move to a faster clock speed, at 3.4 GHz, which shouldnt surprise anyone; however, it is the "turbo-charger" part that makes things a little more interesting.

This is the first mainstream desktop processor from the company using the 90-nm process. This process puts the key components closer together, increasing efficiency and reducing cost, which should translate into lower prices over time for a given level of performance. Thats one nice thing about this market—you continually are getting more for the same price and often for even less.

The next version of Intels Pentium M (the chip used in the companys Centrino bundle), slated for next quarter, will also use this 90-nm process.

This also doubles the cache (level 2, if anyone cares) on the chip, which means more instructions can be retained closer to the processor in ultra-fast memory—increasing the potential performance substantially. This is typically what differentiates a workstation Intel chip (Xeon) from a desktop chip, and workstation chips have a lot of L2 cache. Last year in November Intel released its Extreme Edition processor, which took on die cache, L3 in this case, to 2MB, and the performance jump was stunning. You were basically getting a $3K processor for a third of the price. Granted that price was still $1K or more than many desktop systems sold today, but cutting-edge performance is never cheap.

Here is where things get interesting, if not a little confusing. There will be a number of Pentium 4 parts on the market with unique identifiers likely to confuse you while Intel shifts to the new manufacturing process and new processor.

Next page: Get the scoop on these new parts


At the high end of the line remains the P4 Extreme Edition; however, it is still based on the old manufacturing process, doesnt have the new instructions and remains relatively expensive. On the other hand, it is still a workstation-class chip at desktop prices, and the price is more reasonable then a workstation part for modeling, financial analysis, multi-media and graphics work.

Next down is the new Prescott part designated P4E, which includes the new process, 1MB L2 cache, and the new instruction set for gaming. Strangely enough, for gamers, this is likely a better value now then the Extreme Edition part because it will anticipate future games with the instruction, and at the same price to the OEM as the P4C described below, it likely is a good value at volume (in low quantities the P4C may be cheaper, depending on relative inventories and OEM pricing.) This will change by mid-year, but early on this part should used where there isnt an absolute need to keep hardware specifications constant.

If you must keep a constant hardware specification, the P4C is likely the part for you, although it is based on the old process and has half the L2 cache. Basically this is the old P4 with HT technology part and with a new "C" added to the name.

Following that is the P4A, basically a P4E without hyperthreading and with a slower front-side bus (533 MHz vs. 800 MHz.) Hyperthreading is one of the big technologies that differentiates Intel in the desktop space; it allows one processor, under certain circumstances, to act like two. This becomes most noticeable with applications like Microsoft Outlook XP (Outlook 2003 doesnt seem to need this capability as much) and background utilities like virus checkers. With the latest Trojan virus disrupting the world as I write this, the processors with hyperthreading just became much more important and for a marginal additional cost (and no compatibility penalty that Im aware of) the better choice.

Finally we have the basic vanilla P4, a workshorse chip, with the old process and no hyperthreading. Think of it as a Celeron with more cache. If you see a P4 system for under $600, it probably has this chip, which is best for those who just need the basics. (Though, from Intel, Celeron systems are likely an even better value at this end of the market).

Manufacturers, particularly those using the old parts, may decide not to use the letter designations. So to spec this part properly call out the L2 cache size. You could call out hyperthreading, but this is an Intel only-technology and you dont want to lock out AMD-based systems because even if you dont choose them their inclusion will drive down the bid price. Further, Athlon64 systems will also have 1MB of L2 cache and be competitively priced with systems based on Intels new chip.

Now with any new part from Intel there are two shoes; the new processor is the first, the second is a new chipset, and that is where the amazing changes are expected. Intels new chipset promises instant-on, improved system security and will generally come coupled with Gigabit Ethernet components and better power management. It will provide the closest thing to instant on-and-off we have had to date. However, this will also be a major change, requiring full compliance testing and, as always, those who get their images from OEMs will experience far fewer problems with these new systems then those who do their own imaging. The latter group will need to wait, likely until 2005, before deploying the complete system widely.

In addition, in 2005, we will begin testing the next new operating system from Microsoft, and this complete new platform will be the best way to see just how good this new OS actually is, since the two will go through tuning to best work together. But that is the future and, in the present, for most, that means this announcement, while interesting, isnt disruptive for 2004. And given the disruptions we face on a day-to-day basis, maybe that is the strongest value of all with this new line.

Rob Enderle is the principal analyst for the Enderle Group, a company specializing in emerging personal technology.

Rocket Fuel