Question about the CPU's in general for next gen

AlNom

Moderator
Moderator
Legend
Does it really matter if the CPU's are 64-bit? Is there anything to gain on a console by going with that?
 
Its the natural progression of things. 64 bit processors can address more memory at once allowing them to do more. A fast 32bit can still beat a slow 64bit. It's like if you owned a company and you had 64 employees and your competition had 32 employees. You could do more work at one time but you would have a more complex office structure. Your 64 employees could be lazy though and your rivals 32 employees could be workaholics. ok bad analogy... It's where the market is going and nobody wants to be left in the past.
 
Simple answer is all things being equal and as long as your console has less than 4GB's of memory the 32 bit implementation will almost certainly be more efficient than the 64 bit version. Just because you have to move less memory around when dealing with pointers.
 
In terms of processors such as Sony & Toshiba's Emotion Engine and IBM's Gekko, which are both 128bit processors, are the 64bit processors, (Cell, Hollywood and the X360 processor) being put into the next gen consoles, in anyway a step backwards? Or are the 128bit processors of this current generation not true 128bit processors?
 
Andy said:
In terms of processors such as Sony & Toshiba's Emotion Engine and IBM's Gekko, which are both 128bit processors, are the 64bit processors, (Cell, Hollywood and the X360 processor) being put into the next gen consoles, in anyway a step backwards? Or are the 128bit processors of this current generation not true 128bit processors?

Ugh, not this again.

"Bitness" has long since become completely useless as a measure of the power of a processor, ever since 32-bits in a GPR/address has become more than enough to do what you want.

The interesting operations that all these processors implement all work on 128-bit packed vectors of numbers.
 
aaaaa00 said:
Andy said:
In terms of processors such as Sony & Toshiba's Emotion Engine and IBM's Gekko, which are both 128bit processors, are the 64bit processors, (Cell, Hollywood and the X360 processor) being put into the next gen consoles, in anyway a step backwards? Or are the 128bit processors of this current generation not true 128bit processors?

Ugh, not this again.

"Bitness" has long since become completely useless as a measure of the power of a processor, ever since 32-bits in a GPR/address has become more than enough to do what you want.

The interesting operations that all these processors implement all work on 128-bit packed vectors of numbers.

That wasn't what I was asking. I never really cared about bit-size, I am just trying to understand for my sake a little more about processors in general. All I wanted to know was what difference does it make from the 128bit processors of todays consoles compared to the new 64bit processors of tomorrows consoles? Are todays consoles truly 128bit or are they say 4x32bit, 2x64bit etc.?
 
Andy said:
aaaaa00 said:
Andy said:
In terms of processors such as Sony & Toshiba's Emotion Engine and IBM's Gekko, which are both 128bit processors, are the 64bit processors, (Cell, Hollywood and the X360 processor) being put into the next gen consoles, in anyway a step backwards? Or are the 128bit processors of this current generation not true 128bit processors?

Ugh, not this again.

"Bitness" has long since become completely useless as a measure of the power of a processor, ever since 32-bits in a GPR/address has become more than enough to do what you want.

The interesting operations that all these processors implement all work on 128-bit packed vectors of numbers.

That wasn't what I was asking. I never really cared about bit-size, I am just trying to understand for my sake a little more about processors in general. All I wanted to know was what difference does it make from the 128bit processors of todays consoles compared to the new 64bit processors of tomorrows consoles? Are todays consoles truly 128bit or are they say 4x32bit, 2x64bit etc.?

But you were asking...

NCPU is 32 bit, XCPU is 32 bit, R5900 (PS2) is 64 bit.

Bitness of a CPU is defined as the size of the integer registers, which usually also defines how much memory can be addresses (2^bitness bytes)

NCPU can operate on 2 x 32 bit FP pairs (64 bit packed SIMD)
XCPU can operate on 4 x 32 bit FP vectors (128 packed SIMD)
VU0/1 can operate on 4 x 32 bit FP vectors (128 packed SIMD)

Marketing departments will define bitness as max(integer reg width, FP/packed SIMD reg width, buswidth).

And to answer your second question: next gen consoles will work on 128 bit packed FP vectors, - only with much higher throughput of course.

Cheers
Gubbi
 
Gubbi said:
Bitness of a CPU is defined as the size of the integer registers
That would make R5900 128bit :p

R59k is probably a case that undermines virtually all accepted definitions though. It uses 32bit pointers, 128bit GPRs, and fully supports only 64bit integer math (actual 128bit arithmetic only supports a few ops).
 
Fafalada said:
Gubbi said:
Bitness of a CPU is defined as the size of the integer registers
That would make R5900 128bit :p

R59k is probably a case that undermines virtually all accepted definitions though. It uses 32bit pointers, 128bit GPRs, and fully supports only 64bit integer math (actual 128bit arithmetic only supports a few ops).

Yeah, just read up on it. A real bastard child of a cpu. Build with a purpose (espcially the thing about only sticking a 32bit MMU in there).

Cheers
Gubbi
 
Back
Top