The Official Cpu War Thread !

[Magic Sam]

Hi,

I'd just like to know what kind of CPU architecture do you guys think is the best designed ?

Bye, Magic Sam

 

[hlidegp2x]

Too vague...

for instance, can we reduce CELL to PowerPC ? certainly not.

And what about Alpha ?

 

[Magic Sam]

Too vague...

It could have been worse
QUOTE

for instance, can we reduce CELL to PowerPC ? certainly not.


Mentioning POWER would have been better than PowerPC, as the CELL is derived from POWER4...
QUOTE

And what about Alpha ?


Oops, I forgot that one :huh:

 

[Sphinxter]

Transmeta

 

[Klepto]

Out of the 3 I've written assembly on (x86, ARM, 6502) I much prefer ARM.

IMHO... The dominance of the x86 hasn't done any good for the development of personal computing, it's got a kinda strange architecture and grew from the CISC philosophy while many believe RISC is better for PCs (yes I know that the distinction between the two has become blurred to the point of non-existence). While ARM has it's limitations it is very elegant from a programmer's perspective. Personally I'd have liked to have seen the Transputer have more success as it seems to me to be a more forward looking approach.

 

[A_SN]

From an assembly programming POV (I only know ARM vs x86) I have to go with ARM because it's pretty nice, and x86 hurts my eyes, I mean, shit, it's ugly.

And of course ARM > x86 just for power consumption. The only advantage I know the x86 have a monopoly on desktop machines. The EEE would have been much better off with an ARM if it wasn't for its compatibility with XP and OS X.

 

[AireTamStorm]

I'd say Cell, since it addresses a lot of the bottlenecks you run into from traditional multi core processors... but honestly... all of the CPU architectures you listed are created for a specific purpose and excel in that certain field. If I could vote for worst designed, it'd be x86, for trying to tackle pretty much everything on the planet via additional (nonstandard) instruction sets.

 

[TaG]

The last time I wrote assembly code was on a 6809, and for some odd reason thats been omitted.

 

[hlidegp2x]

I programmed in Z80 (My first computer was a SHARP MZ-80K), x86 (a very long story since I started from 8086), SH (Archos), MIPS (PSP homebrew) and finally ARM (Archos again, GP2X, iPod Touch). I'd like to program in Alpha (but I haven't one), PowerPC and Cell.

Now I really hate x86 whereas ten years ago I could read and write it easily, knowing every bit. Since I discovered SH, MIPS and ARM, I cannot read x86 code back without puking :/

 

[Magic Sam]

I didn't expect so many votes in favour of the ARM architecture...

After a quick google search I found something interesting, the Iyonix PC, which is ARM based

Do you know if ARM is doing well when it comes to desktop computing ?

Thanks for all your interesting replies so far

 

[chad78]

X86, baby. It rules the world. And in a couple of years, Atom will let it take over the handheld market, too.

BTW - I like Walmart, Coca-Cola, McDonald's, and many other winners, too.

 

[senquack]

Intel did do a great job with the FPU instruction set of the 8087. They hired several fantastic engineers and they did a great job, and created an industry standard.

 

[AireTamStorm]

X86, baby. It rules the world. And in a couple of years, Atom will let it take over the handheld market, too.

While Atom is a step in the right direction, Intel has to worry about Via's x86 offerings before they can even think about ARM. The C7 has been a much better CPU in terms of power consumption than the Atom, and ARM CPUs cream -any- x86 CPU in performance per watt.

LONG way to go to get to the handheld market for x86.

 

[Exophase]

Intel did do a great job with the FPU instruction set of the 8087. They hired several fantastic engineers and they did a great job, and created an industry standard.

Are you sure about that... no one really likes the reg stack architecture of 8087 and it has even been superceded on x86 by scalard SIMD, which is much more like other archs have (hence why x86-64 doesn't change the number of FPU registers)

QUOTE
X86, baby. It rules the world. And in a couple of years, Atom will let it take over the handheld market, too.

BTW - I like Walmart, Coca-Cola, McDonald's, and many other winners, too.


lolz at chad78 In before "that was sarcasm, stupid", when that actually isn't suggested at all. You know, just in case.

 

[senquack]

Intel did do a great job with the FPU instruction set of the 8087. They hired several fantastic engineers and they did a great job, and created an industry standard.

Are you sure about that... no one really likes the reg stack architecture of 8087 and it has even been superceded on x86 by scalard SIMD, which is much more like other archs have (hence why x86-64 doesn't change the number of FPU registers)



Well, I guess I should have said they created a good floating point representation standard (essentially designing what would eventually become the industry standard).

As for the instruction set and the 8-register stack: they makes a lot of sense to me, but the only floating point assembly I know is x86-32.

 

[Exophase]

Well, I guess I should have said they created a good floating point representation standard (essentially designing what would eventually become the industry standard).

As for the instruction set and the 8-register stack: they makes a lot of sense to me, but the only floating point assembly I know is x86-32.

Intel didn't create the standard, IEEE did, although 8087 was the first implementation of it (before the standard was finalized). Incidentally it also doesn't conform to it entirely. I think that other archs would have used IEEE 754 regardless of whether or not x86 did.

 

[senquack]

Well, I guess I should have said they created a good floating point representation standard (essentially designing what would eventually become the industry standard).

As for the instruction set and the 8-register stack: they makes a lot of sense to me, but the only floating point assembly I know is x86-32.

Intel didn't create the standard, IEEE did, although 8087 was the first implementation of it (before the standard was finalized). Incidentally it also doesn't conform to it entirely. I think that other archs would have used IEEE 754 regardless of whether or not x86 did.



Hrmm, the book I am reading (Art of Assembly) insists that Intel's engineers created it and the industry standard was based around their work. I have a hard time coming up with something else online that states either way.