As always, there are many reasons. 1. While specific operations are three or four orders of magnitude faster than the old computers, the underlying operations

As always, there are many reasons. 1. While specific operations are three or four orders of magnitude faster than the old computers, the underlying operations are only small multiples faster. When the computer salesmen of the world quote numbers, they are the most optimistic numbers possible. MIPS == Misleading Information Provided by Salesmen... a definition that goes back 30 years or more and still is being practiced today. For instance, when a CPU is rated at 3GHz, it means it can do 3x10^9 NOPs a second, but a missed branch prediction will stall the pipeline for tens to hundreds of (3GHz) clock cycles. Tens if the target of the missed branch is in L1 cache. Hundreds if the target is in SDRAM in a closed page. Excruciatingly slow if the target has to be swapped in off of disk. Another example is (DDR3) SDRAM: it is rated at a phenomenal number of GB per second, but that is only because (a) the GB/sec rating is only if it is streaming sequential data perfectly and (b) it is a lot of bytes wide. If you look at how long it takes to get the first byte of data from SDRAM on an initial (non streaming) transaction, it is not that much faster than the old PC-AT (8MHz clock). Disk drives are similar: the rotational speeds have not increased significantly. The actuators move faster and the data transport is faster, but if it takes 6mSec (10,000 RPM) worst case for a bit to show up under the read head, it isn't that significant if it is 1nSec or 1uSec to get that byte to the CPU. 2. Computers are doing more, and that ain't necessarily a bad thing. 3. Aggravated by #1, the assumption that computers are essentially infinitely fast, computer programmer have become sloppy either through carelessness or through the use of libraries (time to market, y'know). An interpreted language cannot be faster than finely tuned assembly language... except it is practically impossible to do the amount of things expected of a computer today, all coded in perfectly optimized assembly language, and get it done before the computer is obsolete (and the programmers go insane). tl;dr: The speeds quoted by the computer salesmen are assuming a perfect world. High speed caches cannot hide all the imperfections of the real world. Mel's blackjack on a RPC-4000 will beat a 3GHz Xeon processor running a sloppy program written in an interpreted language... but it will only ever run on a RPC-4000 and it is impossible to "fix." Ref: http://catb.org/jargon/html/story-of-mel.html http://catb.org/jargon/html/story-of-mel.html