64 Bit or 32 Bit?

[RefaelBA]

Thanks for the info. I'll go with the 64Bit version then.

[zombie@computer]

Nobody should be using 32-bit nowadays unless they absolutely have to.

i agree. The stance should be "64bit unless..."

The unless bit being:
-shortage of drivers (32 bit drivers dont work on 64 bit OS). Most hardware has 64 bit drivers nowadays but especially older hardware (particularly tv-cards) lack these drivers and are thus rendered obsolete.
-shortage of 64 bit cpu.

Really:
64 bit is faster. Os is faster, the cpu is faster (it has a fair number of optimizations it can only use in 64 bit mode), allthough usually only slightly noticable. It can adress more memory (though 32-bit programs are still limited to the first 4 gb of ram). 64 bit is cooler (like chuck norris versus your dad). Lastly, once we finally decide to drop 32 bit completely (which offcourse will happen eventually), you havent got a single reason to despair.

I wonder if the unix time is going to become a 64 bit integer as well... Perhaps that can avoid Y2K38...

[Dr. Delta]

-shortage of drivers (32 bit drivers dont work on 64 bit OS). Most hardware has 64 bit drivers nowadays but especially older hardware (particularly tv-cards) lack these drivers and are thus rendered obsolete.

If you use Windows 7 there's a good chance you won't even need to install drivers. :>

[whiffen]

Yep, I would add to this but most of everything has been said. Absolutely no reason to go 32 bit anymore. Except for maybe drivers. However I'm on Win7-64 bit and I have had no problems whatsoever involving them. In fact its mostly already done for you

[zombie@computer]

-shortage of drivers (32 bit drivers dont work on 64 bit OS). Most hardware has 64 bit drivers nowadays but especially older hardware (particularly tv-cards) lack these drivers and are thus rendered obsolete.

If you use Windows 7 there's a good chance you won't even need to install drivers. :>

if a manufacturer doesnt make a 64 bit driver then windows doesn't have the drivers either. Just because you dont have to install any doesn't mean you don't need them

[RefaelBA]

A new situation has emerged!

I followed this great guide to installing Win7 through a USB stick. The procedure in general is:
1. Turn the USB stick into a bootable drive
2. Get the Win7 installation on your USB stick.

I did everything it said until I came upon this message:

If you’re currently running 32-bit Windows Vista or 7, Bootsect will only work if you use the files from the 32-bit Windows 7 install disc. The Bootsect executable from the 64-bit version will not run in 32-bit Vista.

I have both x86 and x64 images of my Win7 install. So if I use Bootsec.exe from the x86 install, and then copy the Win7 x64 install files onto it, will the installation work?

[zombie@computer]

worth a try.

[coder0xff]

To clarify: 32 address lines (32 bits) can address exactly 4GibiBytes of address space. Be it RAM, video card address space, or any other memory mapped I/O or cache. Manufacturers have been using the convention that 1000MB is 1GB, to try and boost their numbers. A "350GB" drive is prolly only 350000MB. They are deliberately confusing the base-10 metric prefix with the base-2 binary one. And I can say that it is confused, and not just a difference of opinion, because while they use this convention to go from 1000MB to 1GB, they still maintain that 1024KB is 1MB and 1024Bytes is 1KB, in the same computation. Indeed if they were to apply this deception across the whole computation, a 350GB (actual) drive would suddenly be labeled a 375GB drive. At least just stretching only the Gigabyte part is a bit more modest. (358.4GB)

ANYWAY, the "standards guys" have tried to eliminate this problem by coining new verbiage such as Gibibyte (Gi), the bi being for binary. Though the industry has been slow to adopt because, well, they like lying to make money.

P.S. 64 bits can address 18446744073709551616 bytes, or 16777216TB (or Ti), or 16ExbiBytes (EB or Ei). 5 Ei is supposedly enough memory to store all words ever spoken by human beings (though still not enough to store a complete MD5 crack dictionary). I think 64 bits should last a while xD. But did you know that 32 bit machines have 64 data lines just like 64 bit machines (well all DDR versions anyway), meaning the memory bus transfers 16 bytes at a time. And the Pentium II and later has 36 address lines, even though it's still considered a 32 bit processor. I wonder if that means it could actually memory map into an upper address space letting you use more than ~3 GB of RAM. http://www.pcguide.com/ref/cpu/arch/ext ... ize-c.html
[EDIT]
Each address space refers to one byte (8 bits) NOT individual bits.

[TicTac]

so coder, in layman's terms every gigof HDD space is probably a few mb shy of 1 real gb?

P.S. 64 bits can address 18446744073709551616 bytes

^did you do the math for that when you typed it, or is this just a number that you know? o.O

[coder0xff]

so coder, in layman's terms every gigof HDD space is probably a few mb shy of 1 real gb?

P.S. 64 bits can address 18446744073709551616 bytes

^did you do the math for that when you typed it, or is this just a number that you know? o.O

I only have the powers of two memorized up to 16, so math. lol In all my programming endeavors I never needed more than 16. It is funny though that you bring that up though. Someone asked what the square of 256 (2^8) was once, and I dutifully spouted off the answer like it was trivial (2^16). In the middle of math class nonetheless, and I just laid my head back down on my desk and tried to go back to sleep. You should have seen the looks on every ones faces. Priceless. Got the first 25 digits of Pi memorized too - used to know the first 50.

[TicTac]

2^16 would be a LOT more than 256, do you mean 16^2? :/

[xoqolatl]

But did you know that 32 bit machines have 64 data lines just like 64 bit machines (well all DDR versions anyway), meaning the memory bus transfers 16 bytes at a time. And the Pentium II and later has 36 address lines, even though it's still considered a 32 bit processor. I wonder if that means it could actually memory map into an upper address space letting you use more than ~3 GB of RAM.

It can, via Physical Address Extension. It's considered to be a 32-bit processor because it's general purpose registers are 32 bit wide. Modern CPUs use 36-bit physical addresses, but all GPRs and address generation units are 64-bit wide.

[coder0xff]

2^16 would be a LOT more than 256, do you mean 16^2? :/

I didn't say 2^16 is 256. In fact I didn't say what it is at all. It's 65536. Though yes, 16^2 is also 256, just like 2^8.

So if PAE was available (I only knew about PAE when we were still hanging around 24 address lines) then why do 32 bit versions of windows have this limitation? Is it a performance issue.... or perhaps that software in one window of the PAE wouldn't be able to access memory mapped devices in other windows anyway, making it a moot point? (or at least can efficiently blit between two memory windows?)

Also have 64 bit processors moved on to 64 bit wide registers... wait... I think they do... I remember being happy when I saw that Vista x64 could properly do 2^65 when previous versions couldn't... but that doesn't make sense now that I think about. Because that should be on the FPU anyway, not the general purpose registers - simply because I don't see any reason calculator would choose to use the integer math unit for integers. Unless of course it knows it has to compensate for the fact the a double precision number of that says looses accuracy, and it knows the integer unit could handle it... but does the integer unit even have an exponentiation function? How the fuck is calculator doing it? long doubles? FPUs are native 80 bits anyway, right? I haven't really looked that closely at x64 assembler.

[xoqolatl]

So if PAE was available (I only knew about PAE when we were still hanging around 24 address lines) then why do 32 bit versions of windows have this limitation? Is it a performance issue.... or perhaps that software in one window of the PAE wouldn't be able to access memory mapped devices in other windows anyway, making it a moot point? (or at least can efficiently blit between two memory windows?)

Good read about that.

Because that should be on the FPU anyway, not the general purpose registers

FPUs can operate on GPRs, that's why they are called "general purpose". Seeing as in modern CPUs FPUs also do SSE operations, and most compilers know that, data for floating point ops is often stored in SSE registers, which are 128-bit wide. Software doesn't care where is something calculated - if I want to multiply two floats, compiler translates this to a FMUL instruction and take care about loading proper data from memory into proper registers (lets say they are rax and rbx) at proper time. Then CPU scheduler translates FMUL rax, rbx into micro ops (one microop in case of FMUL) and issues them to execution units. Unless you code in a specific way, you won't even know which execution unit your instruction is gonna end in.