x86-64bit vs x64

[BestBOTEU]

which one is better or it doesn't matter?

[AlwaysFSX]

It's the same thing..

[rtpb5642]

Same thing

[79 Potatoes]

which one is better or it doesn't matter?

It's the same

[Cyrus49]

x86 is 32bit, which has a memory limit on the os. [somewhere between 3-4GB]

 

x64 is 64 bit, which has almost infinite memory.

[Kherm]

x86 is 32bit, which has a memory limit on the os. [somewhere between 3-4GB]

 

x64 is 64 bit, which has almost infinite memory.

You're right, but x86-64 IS x64

[Cyrus49]

You're right, but x86-64 IS x64

Oh, i skimmed over the title quickly...

[Art Vandelay]

x86-64 is probably what Intel invented because they didn't like the term AMD64

x64 is what some guy invented because typing 86- is a lot of effort.

[Litargirio]

x86-64 is probably what Intel invented because they didn't like the term AMD64

x64 is what some guy invented because typing 86- is a lot of effort.

 

And what does the number 86 refer to?

[79 Potatoes]

And what does the number 86 refer to?

x86 is the name of a instruction set

[Art Vandelay]

And what does the number 86 refer to?

86 comes from the 8086 processor, from which this architecture was an evolution.

[GoodBytes]

"x86" is an architecture name from Intel. They could have called it "BestBOTEU Ultimate CPU architecture", but they have chosen "x86". The reason has to do with the processor names which was directly assigned to the architecture of the processor. Marketing names like "Core i7" didn't really exists back then, if I am not mistaken, for old processors.

So, you had 8085, 8080, from Intel as different architecture attempt.

8086 is the one that stuck on, due to the power. It was a 16-bit processor, had an impressive 1 MB RAM (compared to 64KB from the 8085), had lots of compelling features, and was running fast up to 10MHz. It was also backward compatible (although the programs had to be recompiled for it), 8085, and 8080. Then marketing came in, and released 8087, and 8088, but those was the same architecture as the 8086.

Add many years later, 80186 (or the 186 for short), was born, a far more better processor still 16-bit, which is still based on that old x86 architecture. Then you had the 286, 386 and 486, which are models you probably heard about, at least once.

Ignoring the higher end model which increased the last digit or added letters afters, the architecture model/CPU model was "something"86, so Intel went with "x86". Well they did that way back when already. So, it is just a name.

x86-<something> was used to identified additional abilities of the original x86

What I mean is that you have:

x86-16 Which means that the CPU is x86 in origin, and the "16-bit" add-on to it, which additional circuitry was added to be able to work with and process 16-bit long instructions (that is 16 0 or 1's in a row). This allowed support for more RAM, and allowed new operation code to be created to do things faster as well. Much like how we have 64-bit processors today where we can address natively more than 4GB of RAM, the same was back then. 16-bit allowed more memory access than a 8-bit processor.

Over time, as processor model came and go, the 16-bit add-on got fully integrated in the CPU x86 architecture design. There is no benefit from this, just not considered as an "add-on", and rather "part" of the x86 architecture, as if it was natively there, and you won't have model without it.

Then you had x86-32, the x86 architecture that can address 32-bit long instructions, and refereed as 32-bit CPU, but like the previous 16-bit processor that it replaces (x86-16) and replaced later on with 64-bit or x86-64. The Intel 486 processor was the first 32-bit processor from Intel.

Once again, it allowed more RAM support, and new operation code to do things better/faster/safer, and from there, new Intel processor continues to be a x86 based, but bunch of stuff added to it, and expended.

Why did we stop at 32-bit processor for the longest time?

Excellent question. Because at the time, having 4GB of RAM, was like me saying that the 64-bit processor that we all have in our system supports 16 EX (Exabyte. That is ~1000 Petabyte, and that is ~1000 Terabytes (TB), which is ~1000 Gigabytes (GB)), and the saying of how we reached the top, and never needing more, as it was soo much, and room for newer operation code was plenty. Well we did need more, at least for RAM, 'a couple' of years later.

And 64-bit processor already exists before our desktop CPU were 64-bit. The Nintendo N64 was using a 64-bit processor, same for the Dreamcast.If you want, you can build a 128-bit CPU. Of course, cost is an issue, and more bits doesn't necessary mean more performance. As you can see the N64 performance was very limited. Under such case, 64-bit was used for certain algorithms to be able to do them faster, (algorithm affecting games), COMPARED to the same processor 32-bit version of the same speed. It was a way to get more performance without cashing out even more money (a lot more, plus you needed to consider a cooling solution that you need at this point. The N64 had no fans) to get a 32-bit processor more performance out of it. Of course, different architecture has ups and downs, so a 320bit processor from a different manufacture (like what we have in our desktop, destroys the N64 CPU, even at the time in all ways)

To better understand, for simplicity sakes, (this is not a good example, but it is to show different architecture benefits and drawbacks)

-> You have architecture "ARCH 1000" It is a 32-bit CPU, and it can do additions really fast. Takes 0.1ms to do one. But Subtractions, it takes 0.5ms to perform.

-> You have architecture "MONARC 5000" It is also a 32-bit CPU, but additions are done more slowly. Takes 0.7ms to do one. But subtractions, it takes 0.001ms to perform.

Both run at 5MHz

MONARC Inc, figured out a way to accelerate it's processor for additions to 0.5ms on average for performing additions. It introduces a 64-bit variant, still at the same speed. If you do multiple additions at once (for example: 2+2+3+5), on average it will take 0.5ms, but a single one (2+2), it will take 0.7ms. It can achieve this new way of execute multiple additions at once, because it support multiple operation code per instructions. Because of this, instructions needs to be longer to support the same number value size as a simple addition, and why they needed to add more circuitry to the processor to make it 64-bit.

Note: The above example is just an example from MONARC, Inc fictional company as a solution. Not all manufactures would take this path to solve.

I guess it can be seen that way.

And that is why, and when, people started to realize that the "bit wars" in early console was meaningless. More bits doesn't mean more performance per se. Now, in our case, Intel/AMD when we did the switch to 64-bit processor, they added a bunch of stuff to make things go faster. New operation code that does thing faster, provide higher precisions, do things

So to recap:

-> "x86" is just a name

-> "16-bit", "32-bit", "64-bit" identifies how many address a processor supports

-> "x86-16", "x86-32", "x86-64", Are just to signify that the processor has an add-on added to the established architecture to support longer instructions. These are Intel names as well. Just to mix things up, you have another name for it: "AMD64" which is the name that AMD came out for x86-64 design, which they invented and patented, and Intel license from AMD. If you look deeply in Windows files you'll see many will "AMD64" inside, despite you having an Intel processor. Intel is pushing the name "x80-64", to hide "AMD64" actual name.

[Litargirio]

- snip -

 

Awesome, thanks!