Re: Shit - it's all true
Re: Shit - it's all true
- Subject: Re: Shit - it's all true
- From: Allen Rongone <email@hidden>
- Date: Tue, 7 Jun 2005 00:48:53 -0400
I am by no means an electrical engineer, but have more than 20 years
experience with both CISC and RISC machines and it has been my
experience that RISC processors process instructions more efficiently
and faster due to their design. CISC chips use larger dies to hold all
the complex instructions while RISC chips use the extra space to add
more registers and buffers. By doing this they can execute more
instructions per cycle than their CISC counter parts.
On RISC processors each instruction takes only one cycle, whereas a
CISC processor may require multiple cycles to complete an instruction.
Then take into account pipelining. On a CISC chip, if one instruction
takes one cycle to complete in one pipeline and a second instruction
takes 4 cycles to complete in another pipeline a misalignment could
occur requiring additional resources to monitor the output and
reassemble the instructions in the correct order that they were
inputted. This overhead results in lower efficiency.
I understand that Intel tried to compensate for this with their MMX
technology (which by the way was a big flop) and now use SSE2. However,
the Intel approach to move the instruction set to hardware still
remains and results in more cycles to execute a certain instruction,
whereas the RISC design, eliminate the microcode from the CPU and have
the compilers optimize the code, results in one instruction per cycle
increases performance, not to mention the reduction of productivity
costs since they use less silicon to produce a chip.
I regret that I don't have real-world numbers to prove my point but I
think the philosophy behind both designs speaks for it's self. I'll
agree that the CISC design is better for common integer instructions
(As those used in applications that deal mainly with text and small
numbers) but RISC far outperforms in areas that require complex
calculations (Video, Sound, Complex imaging and scientific or
engineering applications).
So I guess if you just want to write a letter or surf the web, there's
no real difference, but if you really want a computer with "power" I
would still take a RISC based machine over a CISC one. No offense.
Anyway, that's my $.02 worth.
Allen
On Jun 6, 2005, at 22:14, Andrew Oliver wrote:
On 6/6/05 6:46 PM, "Allen Rongone" <email@hidden> wrote:
What I can't understand is why anyone would want to leave the RISC
architecture and switch to the slower CISC?!?! Doesn't sound like a
smart move IMHO.
Please show where RISC and CISC are significantly different in terms of
speed. Please also show, where possible, where RISC architecture in
modern
CPUs outperforms CISC. In both cases I consider differences of more
than 10%
as significant.
The reality is that Intel has been able to push the CISC/x86
architecture
further than almost anyone thought possible 10 years ago when RISC vs
CISC
was a big issue with the first PowerPCs.
Nowadays, top of the line processors from IBM (PowerPC G5/variants),
Intel
(Pentium 4) and AMD (Athlon/Opteron) are remarkably close in terms of
raw
performance, with each of them being at the top of the table in some
benchmarks and bottom of the table in others.
The one missing item from this switch is 64-bit. Intel don't have a
viable
64-bit architecture (and the demo systems at WWDC have regular
Pentium4/3.6GHz processors) and I can't see that changing before
Intel-based
Macs start shipping, so what happens to the work that developers have
put
into place in order to reach 64-bit?
Andrew
:)
"Wasted words from wasted memories
left for those who have wasted, in time
what I could have kept as mine" ... AR
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Applescript-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden