Contrary to popular belief, megahertz is not the ultimate benchmark of computer speed.
Despite the popular perception, rating computer speeds by megahertz (MHz) isn’t the same as comparing miles per hour for automotive speed–the latter is a benchmark that can be applied equally to all moving vehicles. But in a world of Intel, AMD, PowerPC, and other chips, it’s not that simple. Comparing the megahertz of, say, a Pentium III chip to that of a PowerPC G4 simply doesn’t work well. Still, it happens all the time. Call it the myth of megahertz.
Megahertz is shorthand for one million cycles per second. It’s used to measure the transmission speed of electronic devices, including channels, buses, and the computer’s internal clock. But megahertz is most often used to measure the speed of the central processing unit (CPU) of a computer. And CPU speed is only one variable in a complex formula that measures PC speed. And that formula may yield vastly different results for you, depending on which program you’re running.
When you’re talking about the nitty-gritty, real-world performance of computers, there are several factors to consider. When an architect is designing a microprocessor chip, she has to make certain decisions about how to achieve optimum performance–and megahertz is only one of these decisions.
“Megahertz is generally achieved by making the pipeline–the term for the sequence of calculation units and the stages of those calculations–longer if you want to run faster,” say Peter N. Glaskowsky, analyst with the Microprocessor Report www.chipanalyst.com, a Sunnyvale, Calif.-based microprocessor resource that regularly publishes performance information. “What this means is that each stage in the pipeline does a little bit less work, that there are more stages working at the same time.”
In the early days of microprocessors, according to Glaskowsky, chip makers did things in two stages: They pulled and calculated data, then wrote it into the processor. Eventually, technology advanced so that multiple instructions could be executed at any time.
“You could run the clock faster because you were doing less work on each part of the process,” Glaskowsky says. “You could do two things separately. This doubled clock speed, but that in itself doesn’t double the amount of work you’re getting done. What doubles it is how many different things are going on at the same time–what’s called parallelism. That’s the other major factor in getting a certain performance from a CPU, and it’s often glossed over when only megahertz is being discussed.”
For the majority of things that most computers do (integer operations for drawing programs, accepting keystrokes, running spell checkers, etc.), different chip architectures do about the same amount of work per clock period. When you look at it in those terms, chips that run PCs are faster than chips that run Macs in proportion to their processors, Glaskowsky says.
“For a lot of functions, a Pentium III running at one 1GHz really is twice as fast as, say, a 500MHz G4,” Glaskowsky says. But for most tasks, extra speed beyond a certain point doesn’t really make much difference. The difference between 500MHz and 1GHz isn’t terribly important for most programs. The emphasis on performance today has shifted to multimedia processing that involves such things as digital video and MP3 encoding.”
It’s the graphics, stupid
The three major processor families all do graphics differently. This makes up the major difference between PC and Mac systems. For example, both the Pentium III and G4 have extra circuitry designed to accelerate multimedia functions: MMX extensions on the Intel and the AltiVec/Velocity Engine on the PowerPC.
“MMX and AltiVec are very different,” Glaskowsky says. “And for the things these units do, the AltiVec does about twice as much work per clock period as the Pentium III does. In some ways, it’s even a little more efficient than that. This means that for a lot of multimedia tasks, the G4 at 500MHz is effectively just as fast as a Pentium III at 1GHz. But you can find examples of code that may be better for one or the other.”
All other things being equal, AltiVec allows Mac users to run graphics programs like Photoshop faster on lower clock speeds than does its MMX counterparts. But, as we shall see, all things are rarely equal.
Memory beats clock speed
Any processor can be made to run at any speed, but it might not run effectively. David Read, a technical writer for Ziff-Davis Media and International Data Group publications, says the modern processor is more like a symphony than a sonata. As in a symphony, different elements come together at certain times.
Memory is one of the most important components of a computer, and one of the most overlooked aspects of having a zippy system, Read says. “It’s like trying to run a steam engine without enough coal to shovel into the engine,” he explains. “You can never have too much RAM or RAM that’s too fast. The microprocessor reads and writes to RAM, so a lack of it can create a double bottleneck.”
You can also boost system performance in other ways. Some tasks can be offloaded to such graphics processors as the PowerPC’s AltiVec, or to the graphics processor on a video card.
“The good thing about the computer processor is that it can do basically everything, but it can’t do everything extremely well,” Read says. “Some things it takes hits on, some it doesn’t.”
That’s why it’s almost impossible to accurately compare the different types of processors: Different ones excel at different tasks. Processor architects have to engage in a series of trade-offs. Boosting clock speed may speed up some jobs, while enhancing graphics capabilities will speed up others. In this respect, systems that offload some tasks to embedded chips on graphics and sound cards can run faster for such applications as games and streaming audio.
Read says that some embedded systems–such as those from ATI and nVidia–may seem slow by clock-speed standards, but because they are asked to perform only one kind of task, they do it very efficiently. “They’re usually running at 180 or 250MHz,” he says. “In terms of megahertz, they’re not that fast, but they’re efficient at what they do, and they have good memory systems next to them to help them ‘shove’ data.”
The memory bus itself–a common pathway, or channel, between multiple devices–is critical because it determines how fast the processor “talks” to the computer memory. In fact, it’s second only in importance to the processor itself, Read says.
Other components that affect speed include the overhead of your particular operating system. This includes the chipset architecture (the data-handling part of the system), the amount of cache (a portion of your computer’s RAM set aside to store frequently used data), and the type of system bus (the pathway between memory, chipset, and processor).
What’s more, microprocessors are really just lots of little processors brought together. They tackle instructions that can be simple or complex. Complex instructions take more cycles. And there are major differences in the various microprocessor architectures, a fact that can make an across-the-board megahertz comparison less than totally accurate.
“There are so many variables when considering computer speed, but unfortunately, most people only care about the speed of the CPU,” Read says.
The real tests of speed
So if MHz isn’t very effective in comparing the real-world performance of different processor architectures, what is? Sadly, there’s no good answer. There are several different types of benchmarks out there. The folks at the Microprocessor Report prefer those from Standard Performance Evaluation Corp. www.spec.org, which attempts to compare one architecture to another.
“The specs aren’t just a single number, but a collection of a different programs that perform different kinds of functions,” Glaskowsky explains. “Take a single number away from the benchmarks, and they’re not very useful to the end-user.”
The best way for end-users to find out which computer offers the speed they need is to simply test the programs they’ll use most often on the systems they’re considering. If you use Photoshop, see which machine runs it the best. If you run lots of Excel spreadsheets, see which system handles it with the most grace. And if you’re doing things like word processing, surfing the Net, and e-mailing, the simple truth is that any new computer will be plenty fast. But that doesn’t make the process of comparing and contrasting chips any less tricky.
“Megahertz is proportional to clock speed for a lot of different tasks, so in some ways, comparing megahertz works pretty well for comparing processors of the same vintage, though it isn’t the final word,” Glaskowsky says. “As an approximation, it’s OK, but that perspective glosses over a lot of things. It’s difficult to say that, in general, a PowerPC chip is faster or slower than an Intel chip.”