Dissapointed At The Short Lifecycle Of Original Xbox

[god_at_hell]

mhz in precessors is just a mesurment of how fast the chissel in the processor can turn and with every turn the Cpu loads a instruction in the Xbox theres a 32bit long instruction set and the xbox has a clockspeed of 733mhz

so Xbox 733*32=23456 amounting to 23456 instructions for fullcycle
Ps2 294*128=37632 amounting to 37632 instructions for fullcycle

I don't want to insult you, but this is false.

The Mhz-value represents turns per second ... right. BUT you can not multiply the Mhz-Value with the Bitdepth and say this equals the instruction-value.
The bitdepth of a computer mostly referres to the instructionlenght a CPU can process. It's a quality not a quantity thing. Shure, sometimes you gain speed if you increase the instructionlength which is processable in one turn, but for example if you only have max. 32bit instructions, a 32bit CPU won't be slower than a 64bit CPU .. if they are identical otherwise. The thing is you can not optimise code so much that you only have 64bit instructions ... so you won't get the double performance. also the bithdepth only referres to the smallest bithdepth of register and data-bus together, so you can have a processor which can process 32bit instructions because of 32bit registers, but only is a 16bit CPU because it's data-bus is only 16bit (the Motorola 68000).

There are many more factors which have an impact on the performance of a cpu ... for example:
- the cache-size
- the 2nd level cache-clock (now mostly the same speed as the cpu, but in history the 2. level cache was clocked slower)
- the pipelinelenght (this is the mayor factor why a Athlon was faster than a P4 with the same frequency) ...
small -> good per-mhz-performance, but no high mhz-values
big -> bad per-mhz-performance, but high clockrates (if you don't get a electromigration)
- extra instructions (MMX, SSE, 3DNOW, AltiVEC)
- design (RISC, CISC, Instruction-Emulation)
- number of registers


short: You can not say the PS2 CPU is faster than the X-Box CPU because of the bitdepth.

 

[evol]

mhz in precessors is just a mesurment of how fast the chissel in the processor can turn and with every turn the Cpu loads a instruction in the Xbox theres a 32bit long instruction set and the xbox has a clockspeed of 733mhz

so Xbox 733*32=23456 amounting to 23456 instructions for fullcycle
Ps2 294*128=37632 amounting to 37632 instructions for fullcycle

I don't want to insult you, but this is false.

The Mhz-value represents turns per second ... right. BUT you can not multiply the Mhz-Value with the Bitdepth and say this equals the instruction-value.
The bitdepth of a computer mostly referres to the instructionlenght a CPU can process. It's a quality not a quantity thing. Shure, sometimes you gain speed if you increase the instructionlength which is processable in one turn, but for example if you only have max. 32bit instructions, a 32bit CPU won't be slower than a 64bit CPU .. if they are identical otherwise. The thing is you can not optimise code so much that you only have 64bit instructions ... so you won't get the double performance. also the bithdepth only referres to the smallest bithdepth of register and data-bus together, so you can have a processor which can process 32bit instructions because of 32bit registers, but only is a 16bit CPU because it's data-bus is only 16bit (the Motorola 68000).

There are many more factors which have an impact on the performance of a cpu ... for example:
- the cache-size
- the 2nd level cache-clock (now mostly the same speed as the cpu, but in history the 2. level cache was clocked slower)
- the pipelinelenght (this is the mayor factor why a Athlon was faster than a P4 with the same frequency) ...
small -> good per-mhz-performance, but no high mhz-values
big -> bad per-mhz-performance, but high clockrates (if you don't get a electromigration)
- extra instructions (MMX, SSE, 3DNOW, AltiVEC)
- design (RISC, CISC, Instruction-Emulation)
- number of registers


short: You can not say the PS2 CPU is faster than the X-Box CPU because of the bitdepth.

youre right i must take much more into consideration but i broke it down to the lowest Denominator
and you cant just say that cache-size matters if the cache matches bitdebt or more what matters more is like you said


registers and optimized code
so why did you say i was wrong ? when you should have said that its not the whole story.

I dont want to take the whole song and dance here so read and say once again that 32bitx86 733mhz ,SSE,MMX,CISC PIII is faster than 128bit MIPS 294mhz RISC,AltiVEC,VU0,VU1

and for the analogy of that bitdebt cant be fully utilised that goes also for 32bit code so saying that a 64 bit PPc processor shouldnt perform better than 32bit x86 at same or lower clock speed is fubar

http://arstechnica.com/articles/paedia/cpu/ps2vspc.ars/1