It seems my understanding is out the window. I’m currently on the hunt for a laptop. My budget puts me in Core i3 territory, which seems to involve clock speeds of 2.1 to 2.6GHz. My 5+ year old desktop is a Celeron D with a clock speed of 3.2GHz. Why do I feel like I’m being duped in buying a new machine with a slower clock speed? It has more processors… okay, does clock speed not matter then? What is the measure of performance?
Clock speeds don’t mean much except when you’re comparing different parts from the same family/model. The i3 can do a lot more work per clock and may have more cores and cache than the old Celeron D, so that i3 can probably get the same work done faster.
Across different processor architectures (x86 vs. ARM vs. PPC vs. MIPS…) it’s even less useful.
As Andrewm indicates, measuring performance by a single criteria is difficult between generations of processor. The i3 has multiple cores (so it can handle more tasks at once), much lower power consumption (important for a laptop), much lower heat output (same), internal GPU (may or may not matter depending on rest of the system), more cache, supports more RAM, etc. It’s individual cores benchmark better than the Celeron as well because of upgrades to the technology.
That said, comparisons between the two indicate that while the i3 is “better”, it may not be significantly so to a point where you’d say “Wow, this is much better!” simply because the i3 is an entry level CPU. Important note here: I was comparing the Celeron D against a the lowest range 2.1GHz speed i3 (I also guessed at models based on clock speeds). As you increase the clock speeds, the gap will widen. Your best bet is to use similar websites to match up your current processor with the actual model of the one you plan to buy.
It’s also worth noting that with current chip technology a little over 3GHz is the fastest practical speed- further gains in performance are mainly due to better architecture. So depending on machine and application, anything in the 2+GHz range is going to offer comparable performance.
It may interest you to know that the Celeron-D was anincredible piece of crap, performance-wise.
IIRC from my computer architecture classes, there is a device on the chip called a clock, which produces an oscillating signal between 0 and 1. The clock speed is how many times this cycles in a second.
A helicopter was flying around above Seattle when an electrical malfunction disabled all of the aircraft’s electronic navigation and communications qquipment. Due to the clouds and haze, the pilot could not determine the helicopter’s position and course to fly to the airport. The pilot saw a tall building, flew toward it, circled, drew a handwritten sign, and held it in the helicopter’s window. The pilot’s sign said “WHERE AM I?” in large letters. People in the tall building quickly responded to the aircraft, drew a large sign and held it in a building window. Their sign read: “YOU ARE IN A HELICOPTER.” The pilot smiled, waved, looked at her map, determined the course to steer to SEATAC airport, and landed safely. After they were on the ground, the co-pilot asked the pilot how the “YOU ARE IN A HELICOPTER” sign helped determine their position. The pilot responded “I knew that had to be the Microsoft building because, like their technical support, online help and product documentation, the response they gave me was technically correct, but completely useless.”
While you are correct that the differences are due to architecture, two processors that have a similar clock speed can have drastically different performance.
When comparing processor speeds, the best thing to look at are benchmarks, like those on Tom’s Hardware Guide and other sites. Some processors are better at some types of operations where other processors are better at different types of operations, and there are different benchmark charts for those different types of things (single threaded vs multi-threaded applications, for example). Look up the benchmarks for the types of things that you use your computer for and go by that.
Another key area of performance these days is in the graphics card. Modern graphics cards are very smart, and often allow the processor to offload a bunch of work to it, which frees up the CPU to do other things. This makes a big difference in things like game performance in particular. If you have two computers with identical CPUs and motherboards but different graphics cards, one of those could easily run at two or three times the frame rate of the other in a graphically intense game. Again, there are benchmarking sites that you can use to compare graphics cards, and again there are different types of benchmarks depending on what you use your computer for. There are some charts that even compare the amount of noise that the graphics card makes while you are playing a game. Some of them have rather large cooling fans on them and can be quite noisy.
For laptops, you are also going to want to look at the processor’s power consumption. Not only does this affect battery life, but it also affects the computer’s long term reliability. Laptops squeeze 14 pounds of crap into a 4 pound bag, and the one thing that ends up suffering in most cases is heat dissipation. A processor that uses two or three times the power will tend to cook itself to death faster than a much lower power processor. What you are probably going to find as you look at laptops is that there is a big tradeoff between performance and heat. A laptop with a processor that scores high on benchmarks is going to score low on battery life and heat. Focus on battery life and heat, and the processor benchmarks go down.
Passmark does has software that they give out to PC users to test their own machines, in order to get actual testing data from computers in the wild. Based on their testing , even the 2.1 Ghz i3 is a lot better. It has a score of 2182 on their tests, while the Celeron-D you describe has a score of 311. Now, you’re not going to feel an 8x increase in speed, but it’s definitely a lot faster.
(The 2.6 Ghz i3 is just a bit better, at a 2248. I’m assuming, of course, that you are talking multi-core. Otherwise the score drops pretty much by half.)
What’s really important is to not only base your decisions on benchmarks, but to choose a benchmark that’s similar to what you actually intend to use the computer for. Most advancement in computer speeds nowadays is in parallelization, but some problems are more parallelizable than others. For some problems, adding more processors just doesn’t do any good at all. If that’s the case, then you want the fastest single processor you can get. Others can benefit from two or three processors working in parallel, but not really from more than that. In that case, you want the fastest two or three processors you can get, but three moderately fast ones might be better than the one fastest one. Yet other problems, you can pretty much throw as many processors at them as you want, and performance will just keep on scaling. In this case, you might be able to get literally thousands of processors, and not care that any one of them individually is rather slow.
Nobody’s really explained what the clock speed IS, at least not in a layman-friendly way.
Imagine that you have a bunch of slaves on your massive galley. Some pull oars, some pump water, some hoist sails, etc… They’re all doing their things to the beat of the guy with the drum. The faster the drum goes, the faster they row, pump or hoist the sails up or down.
So if you take that metaphor and translate it into computer terms, your processor is the galley, the slaves are the GPU, CPU, registers, etc… and the drum beat is the clock, with the tempo of the drum as measured in megahertz or gigahertz is your clock speed.
All else being equal, a processor with a faster clock speed will both outperform a slower one, and it’ll also generate more heat (more work done in shorter period = more heat per unit time). Rarely is all else equal though, and that’s what posters like Chronos, **engineer_comp_geek **and Jophiel are describing.
to refine what bump said, the “drum beat” inside the system is a PLL (phase locked loop) circuit which is fed by a quartz crystal oscillator. The PLL generates a “clock signal” which is a series of precisely timed electrical pulses. so for example, if the “clock speed” of the chip is 100 MHz, then the PLL is generating one million of those electrical pulses per second. These pulses go through a wire on the chip which is its “clock input” signal, and everything inside the chip is synchronized to those clock pulses. every transistor in that chip will switch state in time with the clock pulse. If you increase the number of clock pulses per second, the CPU will run faster (up to a point.)
in the old days, everything in the system ran on the same clock frequency. For example, on the original IBM PC, the system clock frequency was 4.77 MHz. Which meant the CPU operated at 4.77 MHz, the memory operated at 4.77 MHz, the external bus ran at 4.77 MHz, etc. modern systems normally have a PLL chip on the mainboard as a clock source, then other chips have their own built-in PLL circuits to multiply or divide the system clock signal as needed. So for example, on something like a Pentium 4 system, the PLL chip on the mainboard will fire out a 133 MHz clock. The CPU will take that input clock signal and run it through its own PLL, multiply it by e.g. 20, and use its own internally-generated clock to run at 2.66 GHz.
The above answers adequately explain what the clock in a cpu is, but don’t explain how the clock relates to actual operations:
The Instructions per Cycle (IpC) (or the inverse, Cycles per instruction) measures how much a processor architecture can do per clock cycle. For RISC architectures (like the ARM chip in your phone), IpC is high, but you need more instructions to do complex things. For an CISC (Intel/AMD CPU) architecture, IpC will be lower. Both modern CISC and RISC architectures use pipelining and simultaneous multithreading to dramatically increase the overall IpC.
So the difference for the OP is that his Celeron D at 3.2Gh has a much lower IpC than the i3 with a lower clock speed.
As noted, in the end, relevant benchmarking is the only way to verify and compare performance.
Only tangential, but a quite interesting article from The Register on PCI-E, SATA, Northbridges, Southbridges, Front Side Buses, and indeed buses all round.
I had no idea the Northbridge was gone. I will mourn it — mostly because one gets used to the general prevalence of different bits of kit and associated softwares, and never realizes until too late how solely of their era they were.
Winsock, anyone ?
Need speed? Then PCIe it is – server power without the politics 14 April 2015