Aqua-Soft Forumshttp://www.overclockers.com/tips00408/
check it out kids, apple modified systems to make them sh##tier than they really are. once again confirming my belief apple should be a software and gadget company and not a hardware company.
New G5's, Faster Than Nothing
I already knew Apple would mod the results in some way - all computer companies do. But here's an explanation from one member at Spymac. He is ok with PC's AND Macs, so you don't have to worry about him being biased.
Basically, after reading the PDF, this is what I, a chip design layman has walked away with.
They compiled testing software using a stock C compiler and a stock fortran compiler. For the P4s, hyperthreading was disabled, and 5 optimization flags were allowed when they compiled with GCC. For the PowerMac G5, 5 optimization flags were allowed as well.
What intel people are complaining about is that there is a processor specific compiler which highly optimizes the code it compiles for the P4 specifically. In other words, if they compiled it with this other compiler for the P4, the code would run faster. I'm not sure why Hyperthreading was disabled, but I believe it was a bonus for the P4, not a hinderance since this test probably only needed one thread and their "software SMP" would have added overhead. I could be wrong. Don't worry about that if you don't understand it.
What it comes down to is that both processors were allowed 5 optimization flags using an open sourced and highly recognized compiler. In the many years of GCC it has become the standard compiler for Linux, Unix and others. The code has been optimized second only to the compiler created by Intel, as far as I know.
The PowerMac on the other hand, is using a young version of GCC made specifically for Darwin, on a platform and a chip not nearly as old as P4's archatecture and Linux. It was also given the same 5 optimization flags that GCC could produce.
What they did, essencially, was take away the P4's artificial advantage due to processor optimizations in their custom compiler. The Intel crowd is pissed about it, but IBM doesn't have a highly optimized compiler for their 970, and even if they did, it wouldn't be a fair test since the compilers are not the same level of optimization.
What you are left with after all is said and done is that the tests were as fair as they could possibly make them with the youth of the G5 processor and the optimizations of certain compilers. I mean, the G5 DID LOOSE on integer, as it should since the P4 will most likely always win on counting, since they only care about clock speed.
In any case, you should examine real world speed tests, such as the gnome sequencing, FSB data rates, Photoshop renders, or, my personal favorite, how long did it take the average user to figure out how to use it. I think the last benchmark is the most important.
Viper-X-
Everyone modifies stuff to their benefit, and Apple uses their marketing. If you want success you need to be better than the others, and most modify stuff. You cant ocmpare processors that run different stuff. You can compare and AMD and an Intel, but not a G5 and a AMD, or what ever, since they dont run the same OS.
Apple hardware is great, it has quality, great design, and the performance of the machines is great. I saw how iTunes arranged all the songs of a collection folder by folder without affecting the perfomance of the computer, you didnt noticed a slowdown or something like that, but if you move a lot of files on a PC you can see some slow down.
I think the best way to compare two systems is by comparing the performance of Linux on them. Why? Because Linuces are available for both platforms, and we might get a more honest picture..
Originally posted by Taha@Jun 24 2003, 04:33 PMI think the best way to compare two systems is by comparing the performance of Linux on them. Why? Because Linuces are available for both platforms, and we might get a more honest picture..
but linux run's that good on mac? I dont think it has been developed that much as for the x86 plattform...
well, I'm no Contra but I think that it can be compiled for the specific computer, and performance should be the same..
Thats what I was thinking too Taha. Because you can easily optimize it for both systems or run strait ports. Its a great way to test. Because the optimization flags are obvious, no hidden changes.
And it should be CLI only tests. For example, MS windows takes forever to move/copy files (because of the GUI), but if you can move 3GB of files in like 10 seconds through Linux, so its a more efficient, unencumbered test.
Cho008, Most linux code was designed to be portable, so it is easy to get it to run well on PPC.
?)Thats hilarious. 
:lol:
Goku, get a girlfriend, you have too much free time.
AND NOT AN EMOTICON GIRLFRIEND. (they dont count, unfortunatly :-( )
well, the only problem I see here is that they are comparing 2 completely different architectures, Intel based in the x86 and Apple based in RISC, the concept of both architectures in itself is complete different, same concept as comparing CRT to LCD, yeah.. they are both displays.. but completely different. You can not make a fair comparison (sp?) between 2 different architectures,
^^^^ you can still compare scores, even if it is RISC vs. CISC...
i agree with you gents, we need a true system by system comparison.
basically i think the just of the article is that apple modified the systems they were testing to produce unrealistic results, and that, as the editor states at the end of the article, in PC land this would get torn to shreads since they don't say what the systems they are comparing the g5 to have in them.
Ummm, taha and I are idiots. They did use Linux for the comparisions. They used Redhat 9 with GCC 3.3 . I should have read more closely. <_
Originally posted by contrasutra@Jun 24 2003, 01:40 PMUmmm, taha and I are idiots. They did use Linux for the comparisions. They used Redhat 9 with GCC 3.3 . I should have read more closely. <_<
they did?? 
Is there anywhere I can get the benchmark results..?