Baltimore Sun Spots MHz Myth, Brings P4 Slowness To Light

This just in from our iI told you soi department: Mike Himowitz, a columnist for the Baltimore Sun, was puzzled as to why it seemed that his shiny new Pentium 4 1+Ghz systems running Win2k and XP only ran marginally faster then his older P3 running Windows 98. You see, Mr. Himowitz has a killer spreadsheet that takes several minutes for his P3 machine to munch. When he fed the same spreadsheet to his new, "faster" P4 computers they crunched the number only 12% faster. Puzzled, Mr. Himowitz asked the computing public why. The answers he got surprised him. Hereis what he has to say in his article titled, "The vaunted Pentium 4: more hype than zip?":

Although I was taught never to repeat the original error when making a correction, I have to do it here. I wrote about my experience running repeated calculations on a 25,000-row spreadsheet of Census data. Each number crunch took several minutes to process, and since Iive rarely waited more than a second for a spreadsheet to recalculate in recent years, I realized I had a test of grunt computing power on my hands.

I was surprised when a new Pentium 4-based computer running at 1.8 GHz crunched the numbers only 12 percent faster than my 18-month-old Pentium III machine, which runs at 1 GHz. Another new P4-based machine was actually 12 percent slower than the old PC. All the computers had plenty of memory available.

The results certainly werenit what I expected. So I blamed the difference on the operating system - the P4 machines were running Windows XP and Win 2000 (the newest versions of Microsoftis operating system), compared with Windows 98 on the old computer.

My hardware gurus told me I was wrong. Sure, they said, Windows 2000 and XP arenit as quick as Windows 98 for many jobs, but the real problem is the Pentium 4 itself.

As they explained it, the P4 was Intelis first major redesign in several years. (The Pentium III was much like its predecessors at its core.)

Intel likes to sell the clock speed of its chips, since itis something consumers think they understand. That speed is measured in cycles per second - actually in megahertz, or millions of cycles per second (MHz) and gigahertz (billions of cycles per second). The problem is that different processors do different amounts of work in each clock cycle. Thatis why itis hard to compare the true speed of chips from different manufacturers based on their specifications alone.

Apple has had this problem for years. The Motorola-built G4 processor in Appleis high-end Macintoshes is often faster than Intelis P4 in real life. But because it processes instructions differently (and more efficiently in some cases), the G4is official clock speed is far lower. AMDis Athlon chips, which compete directly with Intel in the PC market, are comparable to the P4 in performance but also operate at slower clock speeds.

As it turns out, the P4 - released in November 2000 - was designed to run at a higher clock speed than its predecessor. But it does less work with each clock cycle than the old PIII. Thatis because it breaks down the execution of each instruction in a program into multiple steps, a process known as pipelining. This isnit bad design; in fact, it works very well if youire dealing with repetitive and highly predictable instructions - the kind it takes to process digital video or drive the graphic display in a high-resolution action game

Mr. Himowitz has a lot more to say so click on over to the Baltimore Sun and check out the article.