Slashdot Mirror
Why Do We Still Use Clock Frequencies?
Mr. Sketch asks: "With all the multiple pipelines, prefetching, caching, etc., that goes on in modern the (Bogo?)MIPS be a more accurate measure of a processors speed? If this is the case why don't chip manufacturers rate and advertise their chips with the MIPS value speed, but it seems like nowadays the MHz value is pretty much meaningless and we (as well as chip manufacturers) need to be using something else to get an accurate measure of the speed of a processor." I agree that clock frequency is next to meaningless when it comes to discussing the real speed of today's processors, but would MIPS really be a better replacement?
Measuring the performance of machines is way to complicated of an issue to use something like a MIPS rating. Because MIPS factors out the instruction count needed to get something done, you can inflate your rating by doing a large load of useless instructions really quickly.
MIPS = (InstructionCount) / (ExecutionTime *10^6)
ExecutionTime = (InstructionCount * AvgClockCyclesPerInstr * CycleTime)
The InstructionCount's cancel out, leaving
MIPS = 1 / (AvgClockCyclesPerInstr * CycleTime *10^6)
So if another computer can do the same amount of work with ten times less instructions, it doesn't show up in a MIPS rating.
Measuring performance just isn't as simple as looking at a single numeric rating. Sometimes you are interested in measuring responsiveness, somtimes throughput, and a lot of it depends on the specific applications you want to run. Just asking "how fast" is a processor is almost meaningless.
To my knowledge, the spec benchmarks, while not free, are the best standardized benchmarks out there. For integer performance alone, it tests data compression (gzip and bzip I think), FPGA circuit placement, compiling c code, chess, running perl, ray tracing, database stuff, etc; I can't even remember all of the stuff it tests for floating point performance. Obviously, because it isn't free you probably won't be using it to test your home linux box, but if you are doing serious bench marks, the money would be worth it.
This sig is false.
The basic reason is because the market is still young and immature. We used to talk about the number of barrels in a carb, cubic inches, rear end ratio, quarter-mile/0-60 times/speeds and so forth. These things are easy to measure, easy to quantify. Now we talk about handling qualities, anti-lock brakes, cubic-feet of cargo, etc. What happened was we got to the point where we actually knew how fast was fast enough and suddenly realized that there was more to making a good car than simply straight line accelleration.
;-)
So...for now we talk about MHz, MIPs, MBs, GBs, bits in a bus, bandwidth, frame rates and so forth because they are easy to measure, easy to quantify, and we really don't know what it takes to make a good computer.
In some places, mostly the embedded world, you hardly ever hear of MHz. Who cares how many MHz the proc inside your camera is? Unless of course you want to build a Beowolf cluster of cameras
Good judgement comes from experience, and experience comes from bad judgement.
Good judgement comes from experience, and experience comes from bad judgement.
- W. Wriston, former Citibank CEO
Answer: Use a number that does in some vague way represent a speed difference and sounds really sexy, MHz. The more MHz you got the faster you are baby!
Sure your neighbor has a 700 MHz box but for only a couple grand you can buy this new 1.2 *GHz* box and wipe him off of the map! We sold you cars this way, why not computers? Reality - pshaw - who cares? Joe Sixpack knows thay want "MHz" and MHz we'll sell him.
AMD tried convincing folks their 300 MHZ was just as fast as Intel's 400 MHz chips (or whatever the exact speeds were.) Didn't work, "Processor Class" went away & when AMD surged in speeds they never looked back. Apple tries to convince everyone that their 500 MHz PowerPC is comparable to a 700 MHz Intel PIII and while it may well be no one cares - columnist after columnist sneers at Apple for it's poky 500 MHz (or dual 500 MHz) chips.
Big-iron folks know, mini-folks know, workstation folks know, but the general computer buyer doesn't know that there's a dozen or so variables that affect the speed of a consumer box & CPU speed is only one of them. Motherboard speed, RAM speed, cache size & speed, hard drive speed, so many basic issues affect the 'speed' of a computer but are ignored for the MHz rating.
So know you want to communicate this information to Joe Sixpack who just wants to come into the store, drop a few grand to get a fast box tricked out with today's must-have technology and be back home in an hour? Or to Savvy Shopper who's bought a dozen geek mags in the past week, read over every one yet still has no clue of what any term means & will want the 19 year-old community college part-time clerk to try & explain it all?
No slams here but it aint gonna happen. Folks know MHz, they understand MHz are faster, they want MHz (or now GHz.) Sure they might buy a 1.2 GHz machine with crappy slow RAM & a 5400 RPM hard drive but it's gonna be FAST 'cause it's *1.2* *GHz*!
"It goes to *11*, man! Not just "10" like everyone else but to *11*!"
The rest of us roll our eyes but hell, that's the way the world works. You're not going to find another intrinsic value that communicates the speed of a computer to the general public better then MHz and there's no chance of getting everyone to agree to an artificial one.
I don't read ACs: If a post isn't worth so much as a nom de plume to its author then I wont bother either.
But then Marketing has to have the ability to teach the buyer. Not the easiest task in the world for something so technical.
Clock speed is deathly easy to understand. Faster clock speed means faster computer. (Of course the clock speed argument breaks down when you look at different processors/different buses/supporting architectures/etc)
My main point is: clock speed is the easiest and most effective sell for marketing.
We geeks know the difference.
Neither MIPS or MHz is a truly accurate measure of the speed of a system, the problem is that no-one can agree on a standard. There are other much more mathematically rigorous measures, but which one will you choose? Are you going to measure integer, floating point, or vector processing? Bus speed? Cache hit rate? Unfortunately, the processing requirements for different applications vary as much as the applications themselves. Serving up web pages is a much different task than doing a 2-D FFT on a 16Kx16K floating point array. And processors and systems can be configured differently to do each efficiently.
Consistency is key- if website A compares processors with suite X of speed tests, you can only compare that test with other things done with suite X (and probably only with website A). MHz is a much easier and quicker way to compare. Just one number. Just not an accurate number.
As everyone should know by now, MIPS stands for Meaningless Indicator of Processor Speed. It's completely arbitrary because of the different interpretations of the term "Instruction" between different CPU architectures. A better way would be to use real performance mesaurements, but I can't see Intel's marketing department going for that one: "buy our new Pentium IV (specint 57, specfp 94)". They're alway going to prefer "buy our new Pentium IV/1333 -- it's 33% faster than the Pentium IV/1000."
"The invisible and the non-existent look very much alike." -- Delos B. McKown