Powerpc vs itanium vs x86 performance

  • Thread starter Thread starter ensabah6
  • Start date Start date
  • Tags Tags
    performance x86
Click For Summary

Discussion Overview

The discussion revolves around the performance comparisons between PowerPC, Itanium, and x86 architectures. Participants explore reasons why PowerPC and Itanium have not surpassed x86 performance, touching on historical claims, software compatibility issues, and power consumption considerations. The conversation includes technical evaluations and speculative questions regarding the future of these architectures.

Discussion Character

  • Debate/contested
  • Technical explanation
  • Exploratory

Main Points Raised

  • Some participants suggest that Itanium's performance was hindered by the difficulty of rewriting software for its instruction set, making its speed advantage insufficient to justify the effort.
  • Itanium reportedly outperforms Xeon in floating point operations, but this is seen as irrelevant for most customers.
  • PowerPC is argued to outperform x86 in performance per watt, making it suitable for embedded systems and gaming consoles like Xbox and PS3.
  • There is speculation about whether Intel would have been better off pursuing Alpha RISC instead of Itanium EPIC.
  • Participants question why the Apple G5 was not offered in laptops, with some attributing this to thermal management issues.
  • One participant claims to have conducted performance tests showing x86 outperforms Itanium in integer math without special optimizations.
  • Concerns are raised about how HP continues to sell Itaniums, with some suggesting that customers prioritize OS/software certification and support over CPU specifics.
  • There is a discussion about whether IBM could have engineered a lower power G5 for Apple while maintaining competitive performance against Core 2 Duo.
  • Some participants compare Itanium's floating point unit to hypothetical x86 architectures with integrated GPUs, questioning the legacy-free advantages of RISC over x86.
  • There is speculation about the competitiveness of ARM architectures, like Cortex A9 and Tegra, against x86 in specific markets such as netbooks.

Areas of Agreement / Disagreement

Participants express multiple competing views regarding the performance and viability of PowerPC and Itanium compared to x86. There is no consensus on the reasons for the performance disparities or the future of these architectures.

Contextual Notes

Participants reference historical claims and benchmarks without resolving the underlying assumptions or limitations of their arguments. The discussion reflects varying perspectives on performance metrics and market dynamics.

ensabah6
Messages
691
Reaction score
0
hi,
is there a reason that powerpc and itanium have been unable to outperform the x86?
apple had the g3, g4, g5 and at times claimed powerpc and risc outforms the pentium,
and intel itself bet that itanium epic would outperform the x86, and have lower power consumption and replace the x86.

what went wrong?
 
Computer science news on Phys.org
Itanium might have had a chance if everybody had been willing to rewrite their software for it's novel instruction set. As it turned out writing the compilers was a lot harder than anyone thought and it's speed advantage wasn't enough to warrant the effort. It does outperform a Xeon in floating point, which unfortunately is completely irrelevant for most customers. It's now really an orphan product line, it only has one real customer and is a couple of generations of fab technology behind the cutting edge.

The PowerPC is trickier - it does outperform the x86 in the areas that matter to it's customers = performance/watt. It is the basis for an awful lot of embedded systems as well as being the heart of the Xbox and PS3
 
Last edited:
mgb_phys said:
Itanium might have had a chance if everybody had been willing to rewrite their software for it's novel instruction set. As it turned out writing th ecompilers was a lot harder than anone thought and it's speed advantage wasn't enough to warrant the effort. It does outperform a Xeon in floating point, which unfortunately is completely irrelevant for sost customers. It's now really an orphan product line, it only has one real customer and is a couple of generations of fab technology behind the cutting edge.

The PowerPC is trickier - it does outperform the x86 in the areas that matter to it's customers = performance/watt. It is the basis for an awful lot of embedded systems as well as being the heart of the Xbox and PS3
DO you think Intel would have been better off with Alpha Risc rather than Itanium Epic?

Why couldn't Apple G5 then be offered in a laptop? (the speculation is the reason apple switched to mobile core 2 duo) benchmarks I've seen show G5 lagging core 2 duo.
 
ensabah6 said:
Why couldn't Apple G5 then be offered in a laptop?

I seem to remember it was because they couldn't make one that ran cool enough, but I could easily be wrong because it was some years ago.
 
jtbell said:
I seem to remember it was because they couldn't make one that ran cool enough, but I could easily be wrong because it was some years ago.

I'm aware of that claim, but powerpc RISC was supposed to offer much higher performance per watt than archaic x86
 
mgb_phys said:
Itanium might have had a chance if everybody had been willing to rewrite their software for it's novel instruction set. As it turned out writing the compilers was a lot harder than anyone thought and it's speed advantage wasn't enough to warrant the effort. It does outperform a Xeon in floating point, which unfortunately is completely irrelevant for most customers. It's now really an orphan product line, it only has one real customer and is a couple of generations of fab technology behind the cutting edge.

The PowerPC is trickier - it does outperform the x86 in the areas that matter to it's customers = performance/watt. It is the basis for an awful lot of embedded systems as well as being the heart of the Xbox and PS3

Maybe they should have called it the LOW-PowerPC...
 
I've done performance testing on both x86 & itanium. I wrote integer math algorithms in C & C++ then used native compilers for each platform. The x86 with no special optimizations outperform the Itanium hands down.
 
morris.hoodye said:
I've done performance testing on both x86 & itanium. I wrote integer math algorithms in C & C++ then used native compilers for each platform. The x86 with no special optimizations outperform the Itanium hands down.

So how does HP sell Itaniums? Why does anyone want to buy itanium?

I don't deny that G5 could not fit in a notebook but given it uses legacy free RISC, why couldn't IBM engineer a G5 lower power cpu for Apple?
 
ensabah6 said:
So how does HP sell Itaniums? Why does anyone want to buy itanium?
Nobody cares what CPU is in their server. They care that it is certified to run the OS/Software and that somebody will support it.
HP picked Itanium, because Intel convinced them (possibly with some financial inducement) that it would be the next big thing and they would have a head start over DELL or IBM. HP also had a problem at the time of having three proprietry processor families in house after buying Compaq.

why couldn't IBM engineer a G5 lower power cpu for Apple?
Not while also giving it the same processing power as a Core2duo for the same price for a single customer.
 
  • #10
mgb_phys said:
Nobody cares what CPU is in their server. They care that it is certified to run the OS/Software and that somebody will support it.
HP picked Itanium, because Intel convinced them (possibly with some financial inducement) that it would be the next big thing and they would have a head start over DELL or IBM. HP also had a problem at the time of having three proprietry processor families in house after buying Compaq.


Not while also giving it the same processing power as a Core2duo for the same price for a single customer.

I know that Itanium FPU is world class. How well though does Itanium FPU compare to a hypothetical Sandy Bridge/Fusion x86 + GPU

Wasn't the whole argument for RISC is that it is legacy free unlike x86, and therefore scales better? ARM can offer processor and performance and battery power consumption unmatched by x86 at the low end.

Do you think ARM cortex a9 + tegra is a credible competitor to x86 say in the netbook arena?
 

Similar threads

Comparing RISC vs x86 for Gaming Performance
  • · Replies 2 ·
Replies
2
Views
6K