Originally posted by: Mday
problem is, 32bit processors run 64bit programs slower than 64bit processors. the hammer series of processors from amd will be able to run 32bit programs faster than existing 32bit processors.
anyway
i dont have many 16bit programs anymore. most common 16 bit programs have been made 32bit. there may still be some program out there that doesnt have a 32bit varient, but WHO CARES.
would you drive a brand new mercedes that can go 200 miles per hour on a 30 mile street. or an old pinto that can go (i forgot what speed) on a 40 mile street.
Actually, I'd probably drive the pinto; Don't want my Mercedes jacked...
That, and I'm one of those people who couldn't care less about luxury cars.
As for the topic:
4GHz ALU's have been in existence for sometime. The Pentium 4 uses dual double-pumped 16-bit arithmetic units. Apparently, a new algorithm is needed to ramp up 32-bit equivalents to the same speed if Intel's documentation is to be believed.
The adoption of 32-bit processing, if I remember correctly, occurred only because backwards compatibility with 16-bit was guaranteed. I believe the 486 could switch in and out of protected mode while the 386 could switch to but not out, which meant that once you ran a newfangled program in protected mode, your machine required a reboot to run the old programs.
Oh, yes, and direct memory access beyond 24-bit is useful, too.
The problem with processor design today is that nobody wants to ditch old code. Thanks to IBM, backwards compatibility has become the norm in desktops. Intel decided to try and ditch x86 in favor of IA64, but remember that they were not stupid enough to not provide a compatibility mode. Granted, it runs a hella lot slower, but that's similar to Trasmeta chips.
Buying a 32-bit processor today or even next year is not a stupid decision. While there are several reasons to keep data width small, there are only a few reasons one extends bitness. Chief among them is the ability to process larger numbers. If you look around, only a handful of users can honestly say they're itching for 64 and now that Opteron's available, some can clamor for 128 with a straight face. If Joe Sixpack can honestly say his accounting software is having trouble processing his expenses in 32-bit, he can fund the revolution himself.
Probably the second biggest reason for increasing bits is memory access, and truth be told, I'm not even sure that's required at this stage. Heck, the new opterons only address 40-bits or so at this point simply because it's less hardware to implement and the amount of memory it addresses should be enough for the next 10 years or so. Those who are clamoring for more memory are a relatively small minority. Some of them wouldn't mind having the full 64-bits and accompanying hardware. Some wouldn't mind 128, either.
The point is, only a small number of users are going to have valid reasons for Opterons in the next year or so. Much as I hate to say it, as long as Microsoft doesn't have a 64-bit OS for the masses, the vast majority of the population won't be buying AthlonFX's.
Consumer electronics are not bought like mutual funds.
Facebook
Twitter