2.7 GHz G5 versus Intel & AMD systems running Linux

Hank_Reardon

macrumors newbie
Original poster
Apr 25, 2005
23
0
PowerMacs and OSX: Simply a mess

I'm not sure if you guys have seen this yet, but Anandtech.com has finally decided to clear the waters on the G5 vs. x86 argument. Their article available here: http://www.anandtech.com/mac/showdoc.aspx?i=2436&p=1 is an excellent look at both architectures, and provides the most unbiased benchmarks I have seen. I would encourage you all to read it, as you will be quite surprised what you find. Apple comes out looking pretty shabby, not only from a hardware perspective but a software perspective too.

--Hank Reardon

P.S. Think twice before responding without reading the aforementioned article.
 

Applespider

macrumors G4
But they're still only taking one aspect of life using a G5 or Pentium. I'm sure it's very accurate when talking about raw performance but the following quotes from the article need to be taken into account (and probably cover the bits that most Mac users are more concerned with).

This article cares about speed, performance, and nothing else! No comments on how well designed the internals are, no elaborate discussions about user friendliness, out-of-the-box experience and other subjective subjects.
The G5 is probably the fastest CPU when it comes to Adobe After Effects and Final Cut Pro, as this kind of software was made to be run on a PowerMac. Unfortunately, we didn't have access to that kind of software.
Also much of the areas of 'worst' performance are talking about the G5 as a server with more than 10 connections, again something that doesn't affect every Mac user.
 

Fredo Viola

macrumors member
Mar 27, 2003
75
0
OUCH!!!

That really hurt! It looks to me that the g5 did not fare well in any environment, especially the server. But I find those charts a bit hard to understand. Can anyone clarify?
 

Mord

macrumors G4
Aug 24, 2003
10,091
21
UK
it did fine in the workstation tests, the server tests were a tad worrying, i'll have a word with someone i know that works on OS X when i have the chance
 

montex

macrumors regular
Jan 17, 2002
245
0
Seattle, WA
As soon as I read that they were going to use GCC compiler to generate code for their benchmark tests, I knew the tests would show a huge disparity between the Mac and PCs. GCC compiler is notorious for generating inefficient code on PowerMacs, which is the number one reason their benchmarks should be application based rather than raw number crunching.

OTOH, I don't know why the server connection tests did so poorly. And the funny thing is NEITHER DO THEY! Too bad they were too lazy to figure it out. If they could have at least suggested a cause I would have had some respect for them but to just leave things in the "Golly I just don't know" catagory is very amateurish.
 

TylerL

macrumors regular
Jan 2, 2002
166
57
I would've liked to see them do some benchmarking with YellowDog Linux to find out how that performs vs. OS X Server.

...and you'd think that Apple would've put some more weight behind GCC's PPC compilation and optimization in the 5 years they've been using it.
And what about IBM's compiler? Is that ignored by apple because it doesn't create good/any G3 code?
 

brandona788

macrumors newbie
Feb 24, 2005
15
0
They could have at least installed BSD on their PC's to at least have the same type of kernel.

Overall, I think that trying to compare the PowerPC architecture against the x86 architecture is stupid. Each arch. does well on what it is meant to do. So these people need to shut up and write optimized apps for their arch. of choice and let the masses decide on what they want.
 

monkeydo_jb

macrumors 6502
Apr 17, 2002
447
0
Columbia, MO
montex said:
OTOH, I don't know why the server connection tests did so poorly. And the funny thing is NEITHER DO THEY! Too bad they were too lazy to figure it out. If they could have at least suggested a cause I would have had some respect for them but to just leave things in the "Golly I just don't know" catagory is very amateurish.
Did you read the article? They describe exactly why OS X is much slower in

many of the tests - the way Darwin has to create kernel threads.


Also, they admitted that GCC needed to be better optimized to take advantage of AltiVec.
 

nylon

macrumors 65816
Oct 26, 2004
1,092
432
I think we can honestly say that as the Mac platform increases in popularity it will be under greater scrutiny, especially where price/performance ratio in concerned.

There certainly are some instances where the hardware is seriouly lacking given the cost of it. The powerbooks are an ideal example of outdated hardware being sqeezed to their last drop. The centrino machines have certainly got the upper hand here given that you can get a Pentium M based solution for under $1000 these days.

Yes, Apple hardware is beautiful but in the end if the value is not there, no matter how beautiful the hardware is, people will go elsewhere.
 

Westside guy

macrumors 603
Oct 15, 2003
5,384
2,207
The soggy side of the Pacific NW
monkeydo_jb said:
Also, they admitted that GCC needed to be better optimized to take advantage of AltiVec.
Yup - unfortunately on a number of tests the Altivec comments basically amounted to "here's why this test can't really be used to infer anything other than how well or poorly the G5 did on this particular piece of testing software".

I'd like to see benchmarks of Apache 2 performance instead of 1.3 performance. Apache 2 was designed specifically to address linux-centric issues with 1.3. Their reason was to improve Apache's performance on Windows, but it might affect performance on OS X dramatically as well.

I come from a Linux background, and personally I wasn't convinced that those guys really knew what they were doing. The one worthwhile item to take from that article, though, is that currently many/most OSS apps are likely taking a significant hit in performance if you're running them on OS X. Given that they've mostly been developed on x86 that's not too surprising. But I suspect that, over the next few years, the situation will change significantly as more OSS developers become more familiar with OS X in particular and coding for PPC in general.
 

decksnap

macrumors 68040
Apr 11, 2003
3,070
70
I read that article and thought it made the G5 look pretty damn badass. The only problem it had was scaling up on connections in the server area- which looks like a software problem, not a problem with the processor.
 

daveL

macrumors 68020
Jun 18, 2003
2,425
0
Montana
It's not clear to me that they even used the OS X version of the gcc compiler or that they used the G5 compiler optimization flags. They certainly didn't use the gcc 4.0 compiler supplied with Tiger. They also talk about mysql, but it isn't clear whether they were using a generic OS X binary, which would not be G5 optimized, or if they compiled it themselves.
 

patseguin

macrumors 68000
Aug 28, 2003
1,551
446
kkapoor said:
I think we can honestly say that as the Mac platform increases in popularity it will be under greater scrutiny, especially where price/performance ratio in concerned.

There certainly are some instances where the hardware is seriouly lacking given the cost of it. The powerbooks are an ideal example of outdated hardware being sqeezed to their last drop. The centrino machines have certainly got the upper hand here given that you can get a Pentium M based solution for under $1000 these days.

Yes, Apple hardware is beautiful but in the end if the value is not there, no matter how beautiful the hardware is, people will go elsewhere.
I think that may be one one the most intelligent responses I've read. You're right on the money (pun intended) on everything you said.
 

minimax

macrumors 6502
Feb 9, 2005
351
0
patseguin said:
I think that may be one one the most intelligent responses I've read. You're right on the money (pun intended) on everything you said.
I don't agree. The article clearly states the performance of the G5 is comparable to that of the Xeon and Opteron. It's potential is even very impressive compared to the competition, but it has a few achilles heels that undercut it's performance (relatively weak branche prediction, memory latency, etc.) . This article was NOT about the extra qualities of the Mac, which has been stressed by Johan de Gelas from the outset and simply cannot be denied. With apple you pay a premium for design (also of the innerworks) and build quality. So if you don't like the price, buy a pc, but there's no need to drag it into a discussion that is about performance not price. And that performance is actually quite competitive according to this artice.

edit: with 'premium' I am not talking about OSX vs. Windows, which clarly has it's advantages, but, as this article exposes, also has it's limitations.
 

floyde

macrumors 6502a
Apr 7, 2005
808
0
Monterrey, México
Why!!?? Why do you people bother with this? Who cares? ;)
If God came down from heaven and said "x86 is better, faster and cooler than PPC. Mac was created by Satan to sell slow hardware as if it were fast. I coded the universe using a Gentoo box!" I would still use my Powerbook.
Why? Because I like it. And that is ALL that matters to me. Why do you need all that power anyway? To run Doom 3? Mac is a Niche market, its meant for a certain kind of users/tasks. I have yet to hear a single performance complaint (or any kind of complaint, for that matter) from a member of this select group (true mac users). Apple doesn't sell hardware or software, it sells consumer satisfaction. If you like Linux so much, by all means, go use it. It's a great OS (no sarcasm intended) and it obviously fits your needs.

P.S. your article is biased: the good old "focus on my opponent's weaknesses and showcase my strengths". oldest trick in the book. Oh and don't bother with my spelling/grammar, I don't even speak english.
:)
 

minimax

macrumors 6502
Feb 9, 2005
351
0
floyde said:
P.S. your article is biased: the good old "focus on my opponent's weaknesses and showcase my strengths". oldest trick in the book. Oh and don't bother with my spelling/grammar, I don't even speak english.
:)
Biased? perhaps you should remove those blinkers. Just because you like your mac nobody is allowed to put one onder close scrutiny?
 

Flynnstone

macrumors 65816
Feb 25, 2003
1,419
73
Cold beer land
Hank_Reardon said:
Apple comes out looking pretty shabby, not only from a hardware perspective but a software perspective too.
Huh !
Did we read the same article?
Apple G5 came out fine with the exception of server apps.
 

Kelson

macrumors member
Nov 19, 2002
87
11
Dallas, TX
Excellent Work....

Actually, this was very fair. The G5 does very well against top of the line (okay, w/r/t the Opteron 250, there is a 252 out) from Intel and AMD. It has it's areas where it shines, and areas where it is definitely lack.

I found the reasoning to be very solid, and fits w/ my understanding of processor architecture, albeit much of my knowledge was gained via Ars, from Hannibal's excellent writeups.

I think they hit my biggest issue w/ Mac's in their discussion of thread handling. Plain and simple, OS X is terrible at this. In fact, for 10.5, if I had a single wish that would cause me to shell out 129$, it would be for them to bring OS X's thread handling up to the level of Linux 2.6 or Solaris 10. That would be enough for me.

You can argue that this is a server thing, but that is very much bogus. I notice the lousy thread handling on Panther on my 1Ghz TiBook. When I kick open 8 tabs all loading pages, they don't all load, they all block waiting for one page to finish. This also affects switching between applications, and general system response when running many applications, or causes temporary UI freezes for an application when it is performing an action that requires locking certain resources.

In summary, I'm glad that Apple's pathetic thread handling has been brought to light and I hope the light stays on it. Why? Because it is the main aspect of OS X that I want to see fixed, to fully take advantage of dual core architectures.

- Kelson
 

Fukui

macrumors 68000
Jul 19, 2002
1,615
6
That was a pretty stupid article if you ask me.

They said they were comparing performance of HARDWARE when they ended up concluding it was the OSX that was the problem.

DUH! OS X is a slug compared to Linux.
They should have put linux on all machines, then they can compare performance more equitably.
 

floyde

macrumors 6502a
Apr 7, 2005
808
0
Monterrey, México
minimax said:
Biased? perhaps you should remove those blinkers. Just because you like your mac nobody is allowed to put one onder close scrutiny?
lol, I meant that as a joke. I really don't care if it was biased or not, or if it was undeniable proof that x86+linux is faster (I thought the God part would've cleared that up :) ). The point was that there are other reasons why people use/like macs and that since we're not in gradeschool anymore (well at least I'm not), we shouldn't waste our time with "Superman would kick Batman's arse" type of conversations. I mean seriously, unless you plan on starting a Linux company and need to convince people that your product is better, I don't see why you would need to worry about this things.