| Author |
Message |
|
Consolemu
|
Processors of Today
 Posted: Mon May 15, 2000 12:24 am
|
Are there really processors today that have 128-bit address spaces or are they just pulling our legs? If so, wouldn't that be a technical waste? I did some calculations and I found this out...
8-bit = 256 byte memory. (Unknown)
16-bit = 64KB memory (Master System has this much power)
24-bit = 16MB memory (Genesis has this much power)
32-bit = 04GB memory (Playstation has this much power, I think)
64-bit = 18EB memory (N64 has this much power,I think)
64-bit alone can access even more memory that what I think will be needed for the next 10 years. That's roughly 17 or 18 Billon Giga Bytes!
Chris :o)
|
| |
|
- Site Admin
- Joined: 25 Oct 1999
- Posts: 2029
- Location: Monterey, California
|
Posted: Mon May 15, 2000 2:52 am
|
Quote > Are there really processors today that have 128-bit address spaces or are they just pulling our legs? If so, wouldn't that be a technical waste? I did some calculations and I found this out...
What processor manufacturers claim this? I've never heard of that.
Quote > 8-bit = 256 byte memory. (Unknown)
Some microcontrollers.
Quote > 16-bit = 64KB memory (Master System has this much power)
'Power' is a funny word for address space.
Quote > 24-bit = 16MB memory (Genesis has this much power)
Now we run into funny territory. You might say, "The Genesis's CPU supports 24-bit memory addressing" but to my knowledge the data bus does not. What features a chip supports doesn't matter much in the face of what the rest of the hardware allows it.
Quote > 32-bit = 04GB memory (Playstation has this much power, I think)
Again with the 'power'. Guh.
Your desktop PC, or my amiga 4000, or powermac, etc, all support 32-bit addressing. To which I say:
Quote > 64-bit = 18EB memory (N64 has this much power,I think)
Who knows? Even if the CPU uses 64-bit registers as pointers to memory, it doesn't mean that even the CPU can address that much memory, and certainly the rest of the hardware doesn't support it.
Quote > 64-bit alone can access even more memory that what I think will be needed for the next 10 years. That's roughly 17 or 18 Billon Giga Bytes!
Carefuly placing limits.
The way you use the word 'power' it makes me think you're equating claims of '64-bit processors' with '64-bit address space'. Could be, could not be, but that's really not what 'n-bit' CPU's are all about.
Quote > Chris :o)
|
| |
  
|
|
Consolemu
|
Ohh, I see
Posted: Mon May 15, 2000 3:15 am
|
Quote > What processor manufacturers claim this? I've never heard of that.
I'm not going to say that the manufacturer claims this but I think in a magazine or two they said that the Dreamcast was 128-Bit (200Mhz) processor.
Quote > 'Power' is a funny word for address space.
Sorry. Address Space
Quote > > 24-bit = 16MB memory (Genesis has this much power)
> Now we run into funny territory. You might say, "The Genesis's CPU supports 24-bit memory addressing" but to my knowledge the data bus does not. What features a chip supports doesn't matter much in the face of what the rest of the hardware allows it.
This is where I get confused. I thought the data bus was 24-bit wide; giving the system access to 16MB of addressable space.
Quote > > 32-bit = 04GB memory (Playstation has this much power, I think)
> Again with the 'power'. Guh.
Sorry, again :o)
Quote > Your desktop PC, or my amiga 4000, or powermac, etc, all support 32-bit addressing. To which I say:
So Amigas are still around? I didn't know that.
Quote > > 64-bit = 18EB memory (N64 has this much power,I think)
> Who knows? Even if the CPU uses 64-bit registers as pointers to memory, it doesn't mean that even the CPU can address that much memory, and certainly the rest of the hardware doesn't support it.
I guess I'm going to need another lesson in computer archetecture. How does the data bus have any importance with the address space to the CPU? Aren't they independant from each other? Like, no matter how small or large the data bus is the Address space stays the same? But, I kinda know what you mean. Just because a CPU uses 64-bit registers doesn't mean it has a 64-bit address space.
Quote >
> > 64-bit alone can access even more memory that what I think will be needed for the next 10 years. That's roughly 17 or 18 Billon Giga Bytes!
> Carefuly placing limits.
> The way you use the word 'power' it makes me think you're equating claims of '64-bit processors' with '64-bit address space'. Could be, could not be, but that's really not what 'n-bit' CPU's are all about.
Chris :o)
|
| |
|
- Site Admin
- Joined: 25 Oct 1999
- Posts: 2029
- Location: Monterey, California
|
Re: Ohh, I see
Posted: Mon May 15, 2000 6:33 am
|
Quote > > What processor manufacturers claim this? I've never heard of that.
> I'm not going to say that the manufacturer claims this but I think in a magazine or two they said that the Dreamcast was 128-Bit (200Mhz) processor.
It probably is a 128-bit processor, but that doesn't mean the address space is 128-bit (it can be more or less, depending on what's practical, needed, whatever).
The 68000 is a 16 bit chip with a 24 bit address space
the z80 and 6502 are 8-bit chips with 16 bit address space.
Some variations of the 6502 (as used in the Atari VCS) have a smaller address space.
What '128 bit' refers to is the number of bits the CPU can operate on in a single chunk, without splitting it into smaller bit chunks. (like the z80 which can do 16 bit math, but internally breaksit into 8-bit chunks, or the 68000 which does 32 bit math in 16 bit chunks internally)
It can also refer to how data comes in from the data bus. Even with a smaller address space, it can suck in 128-bits of data in a single instruction cycle. The data bus is 128 bit, but not the addressable space (because that would be, as you have noted, damned silly).
Quote > This is where I get confused. I thought the data bus was 24-bit wide; giving the system access to 16MB of addressable space.
Giving the CPU access to 16-mb... whether the system was engineered so that all 16 megs was addressable is another issue.
Quote > > Your desktop PC, or my amiga 4000, or powermac, etc, all support 32-bit addressing. To which I say:
> So Amigas are still around? I didn't know that.
Well I didn't throw mine away when Commodore died, so it is still there.
You can plug a PPC into an amiga, too, but I haven't felt the need to yet.
Quote > I guess I'm going to need another lesson in computer archetecture. How does the data bus have any importance with the address space to the CPU? Aren't they independant from each other? Like, no matter how small or large the data bus is the Address space stays the same? But, I kinda know what you mean. Just because a CPU uses 64-bit registers doesn't mean it has a 64-bit address space.
The address bus is independant of the number of bits that the processor operates on internally. A large address space is a feature, not an integral part of an n-bit cpu.
|
| |
|
|
kmg
|
minor correction
Posted: Mon May 15, 2000 9:05 am
|
Quote > What '128 bit' refers to is the number of bits the CPU can operate on in a single chunk, without splitting it into smaller bit chunks. (like the z80 which can do 16 bit math, but internally breaksit into 8-bit chunks, or the 68000 which does 32 bit math in 16 bit chunks internally)
fwiw, all m68k processors use 32 bit registers internally. the 68000 has a 16 bit data bus, not 16 bit general purpose registers.
|
| |
|
|
Evan Clarke
|
Posted: Mon May 15, 2000 10:13 am
|
Quote > > Are there really processors today that have 128-bit address spaces or are they just pulling our legs? If so, wouldn't that be a technical waste? I did some calculations and I found this out...
> What processor manufacturers claim this? I've never heard of that.
From Intel Architecture Software Developers Manual (page 35):
Table Goes like this:
Model: MainCPURegisterSize: ExterDataBusSize: MaxExternAddSpace
8086: 16: 16: 1MB
286: 16: 16: 16MB
386DX: 32: 32: 4GB
486DX: 32: 32: 4GB
Pentium: 32: 64: 4GB
Pentium Pro: 32: 64: 64GB
Pentium II: 32: 64: 64GB
Pentium III: 32/128 (128 for SIMD-FP Instructions): 64: 64GB
|
| |
|
|
Nyef
|
Re: minor correction
Posted: Mon May 15, 2000 2:47 pm
|
Quote > > What '128 bit' refers to is the number of bits the CPU can operate on in a single chunk, without splitting it into smaller bit chunks. (like the z80 which can do 16 bit math, but internally breaksit into 8-bit chunks, or the 68000 which does 32 bit math in 16 bit chunks internally)
> fwiw, all m68k processors use 32 bit registers internally. the 68000 has a 16 bit data bus, not 16 bit general purpose registers.
True. The 68000 has a 32-bit internal architecture, a 24-bit address bus (linear), and a 16-bit data bus. It is, for some reason a 16-bit chip.
The 8088 has a 16-bit internal architecture, a 20-bit address bus (segmented), and an 8-bit data bus. It is, for some reason, a 16-bit chip.
The 65816 has an 8/16-bit internal architecture (mode switch), a 24-bit address bus (banked), and an 8-bit data bus. It is, for some reason, a 16-bit chip.
Also, the "bit-ness" of a system may have nothing to do with the CPU. The TG-16 uses a 6280, which has an 8-bit internal architecture, a 24-bit address bus (banked), and an 8-bit data bus. The VDP, however, uses 16-bit data paths, thus it's a 16-bit system.
Also, the "bit-ness" of a system may have nothing to do with reality. The Jaguar apparently is a 64-bit system that is billed as a 128-bit system because it has two 64-bit chips. I can neither confirm nor deny this rumor, however.
--Nyef
|
| |
|
- Site Admin
- Joined: 25 Oct 1999
- Posts: 2029
- Location: Monterey, California
|
Re: minor correction
Posted: Mon May 15, 2000 4:34 pm
|
Quote > > > What '128 bit' refers to is the number of bits the CPU can operate on in a single chunk, without splitting it into smaller bit chunks. (like the z80 which can do 16 bit math, but internally breaksit into 8-bit chunks, or the 68000 which does 32 bit math in 16 bit chunks internally)
> > fwiw, all m68k processors use 32 bit registers internally. the 68000 has a 16 bit data bus, not 16 bit general purpose registers.
That's true. But as I said, it operates in 16-bit steps (so a 16 bit to 16 bit add is faster than a 32 bit to 32 bit add, but not faster than an 8 bit add... or does it have an 8 bit add? I don't remember).
Quote > True. The 68000 has a 32-bit internal architecture, a 24-bit address bus (linear), and a 16-bit data bus. It is, for some reason a 16-bit chip.
Starting with the 68020 it went truly 32-bit. That is, 32-bit operations execute as faster (if not faster) than 16 bit operations.
The classic 68000 is a 'kick ass' (official erminology) 16-bitter.
Quote > The 8088 has a 16-bit internal architecture, a 20-bit address bus (segmented), and an 8-bit data bus. It is, for some reason, a 16-bit chip.
Hmm, is it? I thought the 8088 was totally 8-bit inside, with 16-bit instructions available (a la z80). I guess if it's properly 16-bit, then register-to-register 16 bit operations would be as fast or faster than 8 bit operations.
I Don't care to know, really.
Quote > The 65816 has an 8/16-bit internal architecture (mode switch), a 24-bit address bus (banked), and an 8-bit data bus. It is, for some reason, a 16-bit chip.
> Also, the "bit-ness" of a system may have nothing to do with the CPU. The TG-16 uses a 6280, which has an 8-bit internal architecture, a 24-bit address bus (banked), and an 8-bit data bus. The VDP, however, uses 16-bit data paths, thus it's a 16-bit system.
Is it? That's a point of much controversy. I'd call it a hybrid.
They don't call it a TurboGRAFX-16 for nothing.
Quote > Also, the "bit-ness" of a system may have nothing to do with reality. The Jaguar apparently is a 64-bit system that is billed as a 128-bit system because it has two 64-bit chips. I can neither confirm nor deny this rumor, however.
I was under the impression it was two 32-bit chips, and they were going down SNK's path of adding up the bit count of its CPU bits together.
Quote > --Nyef
|
| |
|
|
Limbs a Flyin'
|
Posted: Mon May 15, 2000 5:20 pm
|
i think whats happened here is you have mixed up the address bus with the data bus. a 128bit consumer system is more likely to have a 128bit data bus(moving 16bytes simultanously around the machine) than 128bits address bus( being able to access insanely huge amounts of ram)
however, in the case of the Dreamcast, i think the 128bit refers to the graphic hardware shifting 128bits at a time, not the cpu (which is '64bit'?)
simplistly you could assume that doubling the size of the data bus doubles your performance. well, i suppose that might be true for sending lots of data from chip A to chip B
|
| |
|
|
kmg
|
Re: minor correction
Posted: Mon May 15, 2000 9:54 pm
|
Quote > > > fwiw, all m68k processors use 32 bit registers internally. the 68000 has a 16 bit data bus, not 16 bit general purpose registers.
> That's true. But as I said, it operates in 16-bit steps (so a 16 bit to 16 bit add is faster than a 32 bit to 32 bit add, but not faster than an 8 bit add... or does it have an 8 bit add? I don't remember).
this is really dependant upon the instruction, the processor, and the bus width. the m68k family has a lot of addressing modes and instructions, so you can't really say that adding two 16 bit numbers together is faster than adding two 32 bit numbers. for example, the difference in execution time of adding two registers with add.l (32 bit) and add.w (16 bit) is negligible, while addq.l (32 bit) would be much faster than addi.w (16 bit), at least on the 68000. and yes there is an 8 bit add. 32/16/8 bit operations can be performed on any of the data registers, and the address registers allow 32/16 bit operations. also what do you mean by "it operates in 16-bit steps"? are you refering to how the 68000 reads/writes memory or how the ALU operates?
|
| |
|
