danlavry wrote on Sat, 29 April 2006 00:37 |
Am I missing something?
Regards Dan Lavry
|
Hi Dan,
I'd say big improvements in systems over the last ten years include: Front side Buss speed, Memory speed, serial ATA, PCI-Express, Pci-X, 64 bit systems (more Ram), more efficient memory management, extended instruction sets for APIs, smaller physical size of chips etc. etc.
I know that this discussion is centred on the actual CPU speed but improvements in computer systems as a whole should be taken into account if overall speed is the salient factor.
As with all computer systems YMMV as regards the application of these improvements on your software's performance. Some older programs may not be optimised to take advantage of the system's improvements so you may see little gain.
Moore's law states "the complexity of an integrated circuit, with respect to minimum component cost, will double in about 18 months.". Processor Speed should be determined not in clock speed but in ips (instructions per second) and flops (floating operations per second). Consider this regarding cost in Moore's Law: in 1997 it cost $30,000 dollars per GFLOP (Giga flop) with two 16-Pentium-Pro–processor Beowulf-class computers. In 2006 it costs 1$ per GFLOP with an ATI PC add-in graphics card (X1900 architecture).
Currently, the second fastest computer in the world is the IBM Blue Gene/L supercomputer at 280 GFLOPS but the fastest is still the Human being at an estimated 10 quadrillion FLOPS.
So we're safe for the moment!
Regards,
Karl