I started computing when a home computer was programmed in machine code and had 256 bytes (yes bytes) of RAM. We progressed to assembler programs that saved us having to know the hex code for each instruction a processor was capable of and commercial computer systems with 16K of RAM and on into today's world of human intelligible language compilers. I spent nearly 10 years programming for a system which finally reached the luxurious level of 4Mb of RAM. We couldn't use that much for programs (the application didn't need it) so I wrote a disk cache to make use of 2Mb of it. This system only supported 60 concurrent users.

The quality of the compiler has a huge influence on the efficiency of the code executed. A piece of 'generalised' code to multiply two numbers is likely to run to many orders of magnitude more instructions than would be required if the programmer had used assembler with a comprehensive understanding of what sized numbers the code might be asked to execute.

What I am rambling on about is that computers can be made to execute the tasks we want done much faster by revisiting the compilers used to convert the programmers high level language into what the CPU actually performs. Could this be the way for Apple to be going?

Actually,
u guys shld check out mac.com and find out why a g4 mac processor is equaivalent or even better than p4s... something abt velocity engine and the bus. Anyway the apple's Airport feature is realli cool too.. AND imac is the onli model i have seen that has an adjustable LCD screen...

I've looked at "mac.com" and what I've seen is lies, evasion, and deliberate distortion by a company attempting to deceive in order to cover up the fact that their machines are distinctly behind the curve, overpriced and horribly underpowered, because their main supplier of processors isn't staying competitive. I see a company attempting to convince the gullible that up is down, sideways is straight ahead, and the G4 is always faster than the P4 no matter what their relative clock rates.

And I'm sad to see that some people fall for it.

The "Velocity engine", also known as "Altivec", is the PPC's implemention of SIMD. In the P4 it's known as SSE2, and they're basically equivalent now in effective capability. Altivec can perform more per clock cycle, but SSE2 gets in a lot more cycles, and they do the same amout of work in real time for all practical purposes. But irrespective of that, SIMD systems are by their nature very specialized and can only be used to solve certain specific kinds of problems. For most things SIMD (Single Instruction, Multiple Data) is totally useless and can't be applied.

The G4's bus, meanwhile, is actually the reason why the latest Macs are completely crippled. The newest Mac is an attempt to build a city and supply it completely over a dirt road filled with potholes. Apple is about a year overdue for a total redesign of that bus, but it can't be done without a total redesign of the CPU, and Motorola doesn't seem interested in doing that.

One thing in particular that's interesting: benchmarks have shown that the new dual-processor Macs slow down considerably when running two unrelated programs simultaneously which are both CPU bound. That's not what is supposed to happen in a dual system, since each program should be running on a separate CPU, and on a decently designed system both programs will run about the same speed as they would alone. But on the Macs, what this does is to overstress that marvelous bus to which you refer, leaving both processors starved for ability to access the RAM. As a result, both CPUs end up wasting a lot of cycles in wait-states because they need to retrieve memory and the bus is busy serving the other CPU. http://www.barefeats.com/pmddr2.html

More disturbing is that Rob has found that the new dual-G4 DDR 1GHz Mac is actually slower at most things than the older dual-G4 SDRAM 1GHz Mac. The new machine has DDR-SDRAM and a 167 MHz bus, while the old machine has SDRAM at a 133 MHz bus. But the old machine has twice the L3 cache, and that seems to be the difference.

The new machines use DDR-SDRAM, but they're feeding data onto a bus which is not double datarate. The reason Apple went with DDR was that it can be purchased to run at 167 MHz, while SDRAM can't be. But they're using the DDR as if it was SDRAM; they're not using the "double data rate" because the G4's bus can't handle it.

Any attempt to claim that the G4's bus is any kind of advantage is a total crock. It's the single weakest part of the entire modern Mac system, and it is what is holding the performance of all modern Macs backward. It is a terrible bottleneck, but Apple can't fix it without Motorola's help, and Motorola no longer cares.

As to the "adjustable LCD screen", check this out: http://www.gateway.com/home/prod/hm_profile4x_ProdDetail.shtml

That iMac may well look pretty cool, but the actual electronics in it is damned shitty, even by Apple's standards, and as a computer (rather than an objet d'art) it's a waste of time (http://www.etestinglabs.com/main/reports/gateway.pdf). Meanwhile, the single pylon support for the screen looks really cool from a distance, but it was an extremely poor design choice in practice. What people have been finding is that on the majority of the LCD iMacs, the screen flops to the left or right and can't be adjusted to stay straight. That was bad with the 15" version, but it's worse with the 17" because the screen is bigger and the lateral forces are greater. Gateway's design, with two supports, won't have that weakness. It's true that it also won't be quite as "adjustable", but sometimes that's a good thing given the iMac's tendency to spontaneously adjust itself by flopping its screen over to the side.

IanAdd, I'm afraid that compiler technology today is actually extremely good. Most modern compilers actually produce code which is tighter than humans could, because the compiler can use approaches humans would try to avoid. (If a human used them the chance of inducing bugs would be too high, but the compiler can keep things straight and avoid the pitfalls.) Unfortunately, the problem of code efficiency is multidimensional, and it turns out that there are legitimate reasons why you don't want to try to force your programmers to keep their code unreasonably efficient if you can avoid it.

In summary: efficient code takes too long to write and has too many bugs. Inefficient code can be delivered sooner, costs less to develop, and will be more reliable. And given the increases in processor speed available to everyone (except Mac users), it's a reasonable tradeoff: I'd rather have an inefficient piece of code available to buy today than a promise of efficient code to be delivered in five years, maybe. I can compensate for inefficiency with faster processors, but no one can run code which hasn't yet been delivered.

I wrote a long series of articles which discussed some of those issues, here: http://denbeste.nu/cd_log_entries/2002/07/OpenSourcepart1.shtml http://denbeste.nu/cd_log_entries/2002/07/OpenSourcepart2.shtml http://denbeste.nu/cd_log_entries/2002/07/OpenSourcepart3.shtml http://denbeste.nu/cd_log_entries/2002/07/OpenSourcepart4.shtml http://denbeste.nu/cd_log_entries/2002/07/OpenSourcepart5.shtml

MAN, am I sick of the RISC vs. CISC, G4 vs. P4 vs. Athlon debates. And the argument that Macs run better because Apple controls both the hardware and the OS, leading to better total integration between the two is hogwash.

I remember just a few short years ago and the ultra-hyped release of OS-8. Initial public releases of OS-8 installed on Performa 6500 series using the 603e CPU had a nasty tendency to trash the hard drives' boot sectors, making them unbootable and in fact, rendering all the data on them unrecoverable for all practical intents and purposes.

That Apple released for public sale an OS that acted like a boot-sector virus on hardware that they themselves made is a gaffosaurus if ever there was one. Didn't Apple even try the OS on all the machines on which they expected it to be run? Why not? They made every single one of them. They must have had examples of them on hand.

On the other hand, Microsoft has been pillaried for far less egregious errors even though they don't control or make ANY of the literally millions of possible hardware configurations on which their OS is to be installed.

How about the original iMac keyboards and mice? Both were ergonomic atrocities. But they sure photographed well.

I love the new OS-X. But I hate Apple for their proprietary hardware. A few years ago, Power Computing was building faster and cheaper Macs than Apple. Apple's solution was to jerk the license to their OS, buy out the company and dismantle it. God forbid there be any competition. So MacOS lovers are forced again to buy Apple hardware at a higher price than they should have to pay were it a free market.

I also hate Apple because for some reason, they think MacOS users want the computing equivalent of designer jewelry when in fact, most of us just want the computing equivalent of computers! Just because we appreciate the virtues of MacOS doesn't mean we want computers that compromise affordability, serviceability, expandability, cooling efficiency, legacy compatibility and ergonomics for the sake of making them look like high-tech tissue dispensers, table lamps, and toilet seats.

And just because we bought an overpriced new Mac doesn't mean there is lots of money in the wallet left to buy all new peripherals, just because Apple decides it no longer likes SCSI ports and 8-pin serial ports and AppleTalk networking. The great god Jobs not only takes these ports away, in most of Apple's newest and biggest-selling offerings, there are not even open slots to retrofit support for these older standards. So now all our external hard drives, scanners, zip drives, network cards, serial and localtalk printers no longer work. Yes, it's true that almost all PC's use USB for peripherals now, too. But most new PC's still also have DB9 serial ports and parallel ports, for those printers and modems we still want to use that need them.

Adapters? Get real. USB-to-SCSI? (YUK!) Serial-to-USB? localtalk/ethernet bridges? By the time you buy all this kludgeware, you may as well have bought new peripherals. And if you do buy all these adapters and manage to install and configure them, your system will become a nightmare of a patchwork camel - not exactly the image that Apple promotes, is it?

Apple seems hell-bent on building for us the perfect, beautiful computers they think we ought to have, rather than the low-cost, adaptable, upgradeable, USEFUL computers we WANT.

Why should MacOS users be trapped into either putting up with Apple's bullshit or holding our noses and switching to Wintel/Lintel? The reason is simple. Proprietary hardware. As much as I like Apple's operating system, I refuse to be trapped by their hardware platform. So I use Windows XP, DAMN IT!! Because it works and I can get my work done and my computer doesn't cost me an arm and a leg and I can upgrade it cheaply and to my satisfaction from a myriad of free-market options.