I'm a little confused about this and i hope some of you can help me.

What does the "bitness" of a cpu mean? Like when someone says a 16 bits cpu, what does that mean? Is a 16 bit cpu twice as fast as an 8 bit one or does it relate to something else?

thanks for the help.

I think its the amount of data flow. Like think of 8 bit as a two lane road, 16 bit would be a 4 lane road and moves more cars along easier than a 2 lane, etc...

The technical meaning of bit is "binary digit", so 8 bits would be an 8 number sequence (1 & 0's) and 16 bits would be a 16 number sequence respectively

Dont quote me on that, I'm a little rusty with comp sci stuff...

[Edit]
But you Can Quote this...talking about the difference between 32 and 64 bit computers (modern bitrates):

"By using 32 bits your processor can represent numbers from 0 to 4,294,967,295 while a 64-bit machine can represent numbers from 0 to 18,446,744,073,709,551,615. Obviously this means your computer can do math with larger numbers, and be more efficient with smaller numbers."

^

Basically, yeah.

As I understand a 32-bit cpu can process 32 bits of information at once, 64-64 and so forth. All software is currently written for 32-bit processors, so 64-bit chips dont offer "double the processing power" - yet.

The big 4-billion number mentioned above has something to do with the total amount of RAM a CPU can address, too. Not sure exactly how it works, but 32-bit processors can only recognise a maximum of 4 gigs (4,294,967,295 bits) or RAM.

Currently you can only have 3 gigs of RAM I believe. With 64-bit systems, it's going to be 8 terabytes. If I remember right.

Some mobos support 4GB of RAM.

OK, the number stuff kind of makes sense, but I don't get the maximum RAM thing. Someone said it can only have 4 GB but my friend told me there's a server at his work with 8. Can anyone clear this up?

There are not any boards currently (biostar, soya, asus, aopen and MSI currently) that go above the 4 gig, because most boards have 4 DIMM slots, and there is not above a gig Ram stick avaliable yet. This is not the limit, just the limit for the standard right now. Soya's DRAGON II mainboard only supports 4 gigs, and its supposedly the 'fastest' motherboard on the market.

Although 64-bit boards may be able to eventually support more, regestered ram is a pain in the neck. A few hundread dollars for a gig total, you have to have some deep pockets. I built a K8 64 bit last week with a gig of regestered ram for a guy who wanted "something not bad" , but I digress

However, there is a server mainboard that can support up to 12 gigs of ram. Its a bit on the pricy side I have to say

Anything over 1.5gigs is just a waste anyway. (right now) That statement is gonna look so dumb in 10 years.

Plenty of server boards support way over 4 GB of RAM...

Pentium Pro and later processors support up to 64 GB of RAM, but can only address 4 GB at a time. The address space itself hasn't changed since the 386 introduced 32-bit protected mode, virtual memory, etc., so the extra memory can only be accessed using a new feature called Paging Address Extensions.

The way it works isn't too complicated--the MMU simply maps 32-bit virtual addresses to 36-bit physical ones--but it's a bit beyond the scope of this discussion. People who remember DOS can think of PAE as similar to EMS, except that it doesn't suck as much.

1.5 GB of ram is a nice ammount, but you may be happy you have more in the near future when games get more and more resource demanding.

Two raised to to the thirty-second power is 4,294,967,296. Before the advent of the 80386 processor, processors can only handle two to the sixteenth power of data or only 65,535. But that is where your supposed theoretical RAM limit comes from. Although higher amounts can be accomplished through like paging and the sort. The 64-bit processor can process unsigned integers as large as 18,446,744,073,709,551,616. That is way more than 8 Terabytes. That would be 18 Exabytes

On top of this entire mess is the fact that there are not one, but three (or four, if you view this first one as two different ones) differents variables in a CPU that can be assigned a "bit" rating. The easiest to understand are the internal and external data path widths. (which are what have been currently discussed) Here's a list of what processors have what data path width (I'm almost certainly to make errors here):

• Intel 808x - 16 bit
• Intel/AMD 80286 - 16 bit
• Intel/AMD 80386 - 32 bit
• Intel/AMD 80486 - 32 bit
• Intel Pentium/AMD K5 - 64 bit
• Intel Pentium II/AMD K6 - 64 bit
• Intel Pentium III - 64 bit
• Intel Pentium IV (Northwood Core) - 128 bit
• AMD Athlon/Athlon XP - 128 bit
• Intel Pentium IV Prescott - 256 bit
• AMD Athlon 64 - 256 bit

Feel free to note corrections needed to this list; I lose accuracy after the PII/K6.

The other two variables are the calculation precision of both integer and foating-point mathematic operations. It is one of these that manufacturers are usually referring to when they say something like "The world's only 64-bit Windows-compatable desktop processor", or the "the Playstation 2 has a 128-bit processor". Usually, the precision level for floating-point seems to be double its precision of integrer operations. In this case, Here's a list of integer precision:

• Intel/AMD 808x, 80286 - 8 bit
• Intel/AMD 80386, 80486 - 16 bit
• Intel Pentium (all versions, from original to Prescott) - 32 bit
• AMD K5, K6, Athlon (K7, Thunderbird, XP) - 32 bit
• AMD Athlon 64 - 64 bit
• Intel Itanium 2, Xeon, AMD Opteron - 64 bit

What the hell is the AMD Opteron cpu anyway? Is it a server CPU chip like Intel's Xenon?

The 80286 can do 16-bit
The 80386 and 80486 can do 32-bit.

For those familiar with assembly language, the 80386 introducted us to eax, ebx, ecx, edx, ebp, uhh, esp (stack pointer)? registers. Prior to the 80386, you just had 16-bit registers (ax, bx, cx, dx, bp, sp, etc). The 8085 was only 8-bit. It had registers A, B, C, D, H, L, SP. But they could be combined to form 16-bit register pairs. HL, BC are such parings. The 8085 has an instruction LXI which loads a 16-bit number into a register pair. Doing LXI H, 0F5Dh will put 0Fh into H and 5Dh into L. And there's your random fact about 8085 processors for the day...and proof that I learned something in Physics 302 :p

OK thanks for everyone who's replied. One question though. The last poster says that at least part of that list of cpu bits is wrong. And I don't know about in computers, but in maths 'integer precision' doesn't really make sense. So is any of the info in that post with the lists right or is it all wrong? No offence to the guy who posted it but I don't want to pick up any wrong information.

Its the amount of extremly short people you can cram in it....

Kinda like in-between the Xeon and Itanium, actually... It's a "64-bit" processor designed primarily for workstations, and also for servers (though that is secondary)

JediGandalf: So what made the Pentium the first "32-bit" processor, if the 80386 could perform 32-bit integer instructions, as well as having a 32-bit wide data path? Perhaps I'm mistaken on the precision of their floating-point operations ("double the precision of their intger operations"), and that's the measure of their precision...

The x87 series of FPU's has had 80-bit registers since ~1980.

The Pentium wasn't the first 32-bit processor by any metric. The major difference between it and the 486 is that it's superscalar.

Not really. Or should I say, not at all. The Xeon is Intel's only TRUE 64-bit processor, if I remember right, for servers, and the Opteron until just recently beat it hands down.