Using a CPU that is under (artificial) load and using slow CPU are extremely different. The responsiveness comes from properly working scheduler, not overpowered CPU.
agreed, although its also worth noting that for the majority of people, paying for the bleeding edge of anything in their computer isn't going to really be tangible or worth the premium they'd pay.
CPU contention is one of the least noticeable overload situations on a computer because schedulers have hidden it well since forever. Even if your video playback and your benchmark are at the same nice-value, the scheduler notices that the benchmark is using lots more CPU, so it gets effectively "niced" compared to the video playback, which then gets the CPU whenever it's runnable.
If someone swapped out your RAM for something less than the working set of the programs you have running -- you would notice. Also, if it were an I/O benchmark instead of a CPU benchmark -- you would notice. (Yes, there are I/O schedulers now, that help somewhat.) I/O and memory are not as trivial for a kernel to "make room" in by kicking out other programs.
I'd only notice if I didn't fire up any games or any major CPU-intensive workload, which are few and far between most of the time. I run powerd on my 3.4GHz 4-core workstation, and the CPU spends 50% of its time between 800 and 1100 MHz, and it often goes all the way down to 100MHz if I shut down Firefox for the night.
Whether a person will notice depends very much on how they're using the machine, and where the bottlenecks are.
I'd notice if you swapped out my CPU for a slower one or a faster one, because it's my main bottleneck in SC2. But if I was only browsing the web, I wouldn't notice. I have friends whose main performance hits are coming from hard drive access, from too little RAM (and therefore paging/swapping), from underpowered video cards, from a weak CPU, and from having too many resource-wasting processes running in the background.
So, to expand on the guy's point: you might or might not notice the difference between a faster and slower CPU, because it might or might not be what's holding you back. If you're going to upgrade, make the right upgrade.
I know that Hector Garcia-Molina pioneered the re-purposing of exotic hardware to be used as a space heater, but really, there are cheaper ways of doing this (and arguably less likely to impact the lifespan of your computer fans in a negative way).
Normally speaking 'orthos' would run at a priority low enough that it would not interfere with other foreground tasks but you can actually set it to be 'bad', and then you would definitely notice.
I was thinking about this today. In math class I got bored and started playing with my calculator (TI83) and a friends calculator (TI84). I wrote the same program, essentially
x=1
while true
print x
inc x
end while
on both calculators. Starting both at the same time, the TI84 naturally counted faster than the TI83. But while the TI84 was faster, I've used both models a lot and have never really noticed a huge difference.
On the other hand, a TI89 that I use for a different class and performs noticeably faster at stuff (and when I ran the program on it later, seemed to count much faster).
I have a script in my ~/bin that does some very inefficient calculations on some numbers from /dev/random. It is called heat.py and used to get run in the winter, when I lived in a less than ideal place with a barely working heater and poor insulation. Definitely made a difference.
Can't right now, as apparently my home internet connection is down. Off the top of my head tho, basically I used python's random module, and got a random int, and a random float, then did a bunch of pow(rfloat, rint) and multiplications in loops and bit shifts and so on. Basically i tried to utilize all the parts of the CPU I could think of, and keep the cpu usage pegged.
I used to live over a garage and during the winter it would get freezing cold in my room even with the heat on. I left my tower on all the time running World Community Grid to keep the place warm.
Conversely, during the Summer it would become unbearable to even operate my tower so it was a great incentive to turn it off, save electricity, and go outside.
If you permanently undervolted my CPU, I wouldn't notice until I go to run numerical computations or load up star craft. But I would always notice a slower hard disk and slower ram.
I would notice if my DVD Rips were taking longer to process, or I was no longer able to sustain my live streaming video feeds at church on Sunday morning.
Sorry for the self promotion, but here is a blog post that argues that the latest microprocessors now overshoot the needs of the vast majority of customers and how this spells trouble for Intel: http://www.fernstrategy.com/?p=229. I've appended the key parts of the blog post below for convenience:
In 1965, Gordon Moore, the co-founder of Intel, proposed that the number of transistors on a chip (forming a microprocessor) would double approximately every two years (Moore’s Law). Transistor count is important because the number of transistors on a chip affects a microprocessor’s performance; i.e., the number of instructions performed within a given period of time. For the last 45 years, this doubling trend has continued unabated, and the latest generation of microprocessors from Intel contain about 2-billion transistors. Now combine this rate of technology development with the movement towards cloud computing. Cloud computing reduces the need for high performance (bulky and expensive) computers at every desk because complex processing tasks can now be handled remotely. In short, the rate of technology development combined with the advent of cloud computing has given way to microprocessors that now overshoot the needs of the vast majority of customers. Because many customers are content with older generation microprocessors, they are holding on to their computers for longer periods of time, or if purchasing new computers, are seeking out machines that contain lower performing and less expensive microprocessors. The result for Intel is lower revenues and profitability.
The next issue for Intel is that cloud computing is facilitating an entirely new crop of mobile computing devices, such as netbooks, tablets, and smartphones. The issue for Intel is that many of these mobile devices use the ARM architecture, a competing technology that is more energy efficient. And it’s very difficult for Intel to compete directly with ARM because of ARM’s unique strategy. Unlike Intel, ARM does not produce and sell microprocessors based on its technology; rather it licenses the technology to companies. If Intel were to follow a similar strategy, its revenues and profitability would drop significantly.
In sum, Intel is facing a big squeeze; as Intel’s revenues and profits are getting squeezed at the top because the average customer no longer needs or wants the latest generation Intel chip (due to Moore’s Law and cloud computing), it’s also getting squeezed at the bottom as OEMs of mobile devices are adopting the ARM architecture in droves. And, like many disruptive technologies, the ARM architecture is now moving up-market, beyond mobile devices. For instance, the company “Smooth-Stone” recently raised $48m in venture funding to produce high performance, lower powered chips based on ARM technology to be used in servers and data centers.
The comparison between Intel and ARM is like comparing Microsoft and Apple. On a customer facing level, they may appear similar, but they're different beasts, and arguably in different markets.
Either the CPU-consuming process was running at an idle priority -- meaning the OS would always push it aside for other activities that needed CPU, such that when the author was browsing the web or navigating emails, those more important processes got the run of the processor -- or the author is just incredibly tolerant. If there was something actually competing on an equal basis for the same cycles (instead of only taking the available cycles), the impact is obvious to anyone.
While for some reason "browsing the web" always comes up as a low need activity, it happens to be the area that really differentiates machines. It is not a low demand activity, and hasn't been for years.
