drop shadow

Page 1 of 1 pages for this article

Intel’s Prescott Meets SFF – What’s That Burning Smell?
Title Gradient

Virtually ever since the rumors of Intel?s Prescott processor first began to surface, we?ve been pondering whether current small form factor systems can handle the temperature and power requirements of Intel?s new 800 lb. gorilla CPU. Opinions abound, from "there?s no way" to "why not?", but only a definitive test could answer the question. Early reviews pouring in over the net have shown varied temperature, power, and performance numbers for the Prescott processor within standard desktop systems, but few, if any, have attempted to test the new beast within the confines of a small form factor box. Perhaps their insurance premiums weren?t paid up so they didn?t want to take the risk. Nevertheless, the question still stands, do Prescott and SFF mix?

With power dissipation rumored to be in excess of 100 Watts, how could an SFF system handle such an energy hog? And even if it could, how would you keep it from burning up? Many are still wondering if current SFF Bios?s and motherboards will even handle the new CPU. Most manufacturers have been rather quiet, but what?s the real story?

Today we aim to answer these questions and more by stress testing our review sample Intel Prescott 3.2 GHz E processor within the Shuttle SB75G2 XPC. As you?ll see in the review, we showed that Prescott has essentially the same performance as Intel?s Northwood core and was actually slower in some tests. This coupled with the excessive heat made us conclude that, at this time, the Intel Northwood should still be the processor of choice. Nevertheless Intel plans to push forward with the Prescott core and that means our SFF readers will want to upgrade. Should they?

Let?s just say you?re in for quite a rollercoaster of a ride. Unfortunately, the ending isn?t very pretty, nor did it smell good. But like most stories, everything begins on a positive note?

Meet Prescott

Meet Prescott

Launched just over two weeks ago on February 2nd, the Intel Prescott represents the microprocessor giant?s latest and greatest desktop/workstation CPU. Reflecting Intel?s new 90 nanometer process technology, the CPU crams more transistors per millimeter than any other processor on the market today. At 125 million transistors, exactly what does the Prescott buy you?

Prior to release, rumors abounded from special "low-d dialectics" to reduce temperature and improve speed, increased manufacturing yield which should help reduce costs, larger caches for increased performance, special SS3 instruction set support, improved HyperThreading, and much more... But... If you followed the launch, or read our Sudhian Prescott Review, you know that this "monster of a chip" had a louder bark than bite.

Despite its less than glowing thermal characteristics (you could fry an egg on this sucker), curiously poor performance, and incredible power requirements, the Prescott may represent the future of the Pentium 4 for some time to come.

The die, shown below, looks huge and complex. Some have said that it contains 64-bit functionality, akin to the Athlon64, which Intel is secretly keeping deactivated for the time being. With over 2X the transistors of its younger sibling, the P4 Northwood, one may wonder just what all those 125 million transistors are up to. Others point out that they are simply accounted for in the increased cache. For whatever the reason, more transistors inevitably means more power and more heat, as we?ll see later on in our benchmarking.

Before we move on, let me point out that this review isn?t meant to provide you with all the details surrounding the Prescott processor itself. For more detailed specifications on the processor as well as detailed benchmarking and thermal analysis data, please see our Sudhian Prescott Review.

Back on the subject of Prescott within the confines of an SFF system, our biggest concerns were heat and power. With most SFF?s today shipping with PSU?s ranging from 180 to 220 Watts, we knew it was going to be a stretch. Under load, the Prescott can consume well over 100 Watts of power. Combine this with a high performance AGP video card, 7200 RPM or greater hard-drive, an optical disk, and active cooling system, and you?re risking serious meltdown.

Power is one challenge, but even if you can power the system, heat dissipation in an SFF has always been a challenge. Even with standard Pentium 4 Northwood processors, one generally finds them to run 10 degrees or more hotter within the confines of a small form factor box versus the larger, roomier, and often cooler confines of a desktop tower case. Could an SFF cube really dissipate heat fast enough to keep the processor cool?

Initial rumors abound regarding one of the most popular SFF manufacturers, Shuttle, and Prescott support. Images such as the own below started circulating last year.

Exactly what the heck is that goliath of a cooling system affixed to the rear of that XPC? Fortunately this prototype unit never made it further than just that, a prototype. Gear of this sort isn?t required to use the Prescott processor with a Shuttle XPC. For those wondering, Shuttle has confirmed with us that the Prescott processor is fully supported in their SB75G2, ST61G4, and ST62K XPC?s ? with the latest Bios upgrades. For those who need a quick refresher, the SB75 is powered by the Intel i875 Canterwood chipset and represents Shuttle?s current top-of-the-line XPC in terms of performance. The ST61G4 and ST62K systems feature the new ATI RS300 chipset including the on-board 9100IGP video solution.

Before we move on to building our test environment, let me take a moment to comment on the Prescott in the ST62K Zen XPC. In a word? Crazy. The Zen ships with an external 180W PSU. On the positive side, the PSU will certainly not contribute to the internal case temperature given its external form factor, but powering an entire system on 180W with the Prescott consuming potentially more than half of that power is alarming at best. I?d recommend holding off on this experiment.

Moving on, let?s discuss our testbench setup before we move into benchmarking? No smoke yet.





(1 of 5) - next >


Page 1 of 1 pages for this article

Search

Advanced Search


Newsletter Signup