Do we really need faster processors?

[ramollari]

Research in operating systems has shown that most of the time in the computer is spent with I/O operations, like disk write/read, network I/O, etc., while the processors' speeds have advanced so much that they mainly stay idle. Even CPU intensive operations, like compiling and long calculations are done relatively quickly. So why would we need much faster CPUs, let's say a CPU of some THz clock rate? The effect would have been almost ignorable, knowing that most of the time would be spent with peripherals. Don't we have to improve all the hardware in parallel before we think of major advances in CPU speeds? Why for example would one have to buy the expensive 3GHz+ processors while a much cheaper 2GHz processor would suffice?

 

[hitssquad]

Processing ability is not measurable by clock rate or any other discrete processor characteristics. This is explained in terms of the g factor and the human brain in Arthur Jensen's The g Factor.

 

[graphic7]

It very much depends on the application of the computer. If you're doing any sort of scientific computing, involving numerical computations, the CPU is the most important componet, and speed/optimizations do help quite a bit. If the application does any sort of disk or memory swapping, and graphics, then of course the most important componet will not be the processor. As I said, it strictly depends on what the application of the system is.

As stated in another thread, the amount of cycles/second a processor can do is not the only factor when one evaluates the overall "speed" of a processor. Many compilers are geared towards specific processors because of the specific optimizations they have. And the converse, as well, many processors are geared towards certain compilers. Overgeneralizing is common by just referring to x86 processors not gaining any compiler performance for minute increases in their frequency.

Keep in mind, emphasizing that *all* CPU manufacturer's have a focus increasing their CPU's frequency is simply not true - the x86 manufacturers are doing this, stressing frequency over optimization and processor design. Remember, most consumers care about the number of Hz a processor can do over the processor design. Not that I'm criticizing Intel or AMD, but the PowerPC and SPARC are well, well below 2gHz and both can compare reasonably well to x86 chips running at ~3gHz in floating point and integer arithmetic.

 

[exequor]

depending on what you use you computer would justify whether someone needs a faster computer or not. the thing is i would like faster processors whether we really need them or not, i am dying for the day when the first quantum processor is used.

 

[ramollari]

i am dying for the day when the first quantum processor is used.

I would rather think an optical processor that uses light impulses than electricity. But what's the deal, after all. If we don't improve hard disks, network connections, bus speed, memories, and so on, there won't be a significant increase in performance.

 

[ramollari]

depending on what you use you computer would justify whether someone needs a faster computer or not.

There are two things one should consider. First of all, any computer today is perfoming I/O operations. That's the general pattern. Second of all, any operating system implements virtual memory, so disk read/writes have to be performed continually.

 

[chroot]

ramollari:

Many programs (such as scientific computations) involve very little I/O. Second, you don't necessarily have to make use of an OS's virtual memory services; nor do you need to use an OS that offers them. In some cases, you don't even need (or want) a distinct operating system at all.

- Warren

 

[exequor]

i guess everyone (except chroot) is relating their posts to OSs like windows. Anyway i totally agree with improving the peripherals. i think the usb and firewire have been improvements in that area though.

 

[graphic7]

i guess everyone (except chroot) is relating their posts to OSs like windows. Anyway i totally agree with improving the peripherals. i think the usb and firewire have been improvements in that area though.

I fail to see how I related to Windows.

 

[ceptimus]

People like things to be fast. Even though I only spend an average of two minutes a day microwaving stuff, I'd still like a faster microwave.

Same thing goes for my PC.

 

[ramollari]

People like things to be fast. Even though I only spend an average of two minutes a day microwaving stuff, I'd still like a faster microwave.

Same thing goes for my PC.

Faster CPUs do not necessarily ensure a faster PC overall.

 

[PerennialII]

It very much depends on the application of the computer. If you're doing any sort of scientific computing, involving numerical computations, the CPU is the most important componet, and speed/optimizations do help quite a bit. If the application does any sort of disk or memory swapping, and graphics, then of course the most important componet will not be the processor. As I said, it strictly depends on what the application of the system is.

I second this ... also when doing numerical comps the moment the analysis hits swap or serious disk activity the analysis is often lost ... in "large" analyses you've pretty much got to keep it in memory & let the CPU do its magic, otherwise throughput times go off the ceiling. The fact that in number crunching the trend has been ever more towards parallel systems is an indication that higher power CPUs are required, at minimum you need a small cluster or a high end multi-CPU server to get anything real done.

 

[russ_watters]

Faster CPUs do not necessarily ensure a faster PC overall.

As others have said, it depends an awful lot on what you use the computer for. I'm transfering my family VHS movies to DVD and some more processor for the encoding would be nice.