NO DISINFO after all !!! - G5 vs. Core Duo
It's definately faster than G5. The press shot from the hips when they called Jobs presentation "disinfo". Take a look at this :
http://forums.macrumors.com/showthread.php?t=176161
[quote]MacSpeedZone posts their benchmarks for the Dual Core Intel iMac and claims that Macworld's initial test were misleading:
Sincerely
Zab the Fab
http://forums.macrumors.com/showthread.php?t=176161
[quote]MacSpeedZone posts their benchmarks for the Dual Core Intel iMac and claims that Macworld's initial test were misleading:
Quote:
This is where the Macworld "First Lab Tests" article falls a little flat ... obscuring the processor capacity vs processor usage problem inherent with mutiprocessor machines (or multi-core ... same difference). Using Macworld's logic we could argue, given the data above, the Quad G5 Power Mac is only 14% faster when running some of Apple's own applications. We think that this is misleading, as we pointed out.
They post a comparison chart, taking into account percentage of processor usage as a guide.
This is where the Macworld "First Lab Tests" article falls a little flat ... obscuring the processor capacity vs processor usage problem inherent with mutiprocessor machines (or multi-core ... same difference). Using Macworld's logic we could argue, given the data above, the Quad G5 Power Mac is only 14% faster when running some of Apple's own applications. We think that this is misleading, as we pointed out.
They post a comparison chart, taking into account percentage of processor usage as a guide.
Sincerely
Zab the Fab
Comments
With Dual Procs you don't always up the speed limit but you do widen the lanes of the highway.
It would have been easy for Macworld to run proper tests that actually delivered a bit of value to their readers. Instead they set out to smear Steve Jobs.
I wouldn't wipe my crack with their rag that's how I feel about Macworld Magazine.
For the past few years, I just observe and collect user comments as well as try to get my hands on the hardware or software in question before making any decision. I also use good ol' logic to sort what sounds fishy and what doesn't.
A bunch of user comments is much more useful than a single review. And a hands-on experience wins it all.
Originally posted by kim kap sol
I've stopped reading any tech review...be it software or hardware. Most people don't know what they're talking about. The only reviews I can rely on are Anandtech and Tom's Hardware reviews...but I'm wary of these too.
For the past few years, I just observe and collect user comments as well as try to get my hands on the hardware or software in question before making any decision. I also use good ol' logic to sort what sounds fishy and what doesn't.
A bunch of user comments is much more useful than a single review. And a hands-on experience wins it all.
I couldn't agree more. Have you noticed a lot of poor reporting on Tom's recently as well? The actual hardware testing and reviews seems ok, but the "editorial content" just keeps going down hill.
Anyway, back on topic...I think it's important for everyone to realize that the _potential_ speed of the duo (and more importantly, the entire systems) is huge, and that as more software becomes universal we'll have a better understanding of real-world performance. Also keep in mind that initial universal binaries are not likely to be terribly well optimised, and as we start seeing .2 and .3 releases of universal binaries we should see similar speed increases again.
I can tell you that the Intel iMac feels faster even than my quad powermac. I have yet even ONCE to see the beachball on my Intel mac. Things like browsing 500 files, or right clicking on my applications folder in the dock (200+ apps) snap up immediately, not the small delay that is so frustrating on my powerpc computers.
Everything absolutely flies on the iMac. I can hardly wait for Photoshop and other applications to be native. For day to day operations I use my iMac now, I feel kinda goofy letting my quad g5 sit in the corner room all day..
I do use it for Aperture though.
Originally posted by kim kap sol
I've stopped reading any tech review...be it software or hardware. Most people don't know what they're talking about. The only reviews I can rely on are Anandtech and Tom's Hardware reviews...but I'm wary of these too.
Fortunately, the MacSpeedZone article put things into perspective. At least in this single case. The MacWorld test was just ridiculous.
EDIT: actually not so much the test in itself, as the comments and the conclusions drawn.
Originally posted by concentricity
Anyway, back on topic...I think it's important for everyone to realize that the _potential_ speed of the duo (and more importantly, the entire systems) is huge, and that as more software becomes universal we'll have a better understanding of real-world performance.
So true. The benefits of dual processor systems have been discussed for long and in such detail, that my head hurts when I see comparisons between the two iMacs, the G5 and the Intel one, that do as if there was only one CPU in the later.
Also keep in mind that initial universal binaries are not likely to be terribly well optimised, and as we start seeing .2 and .3 releases of universal binaries we should see similar speed increases again.
This is also true. As with any transition of this kind, you should wait one or two years to see reasonably optimised software for the new systems.
Originally posted by kim kap sol
I've stopped reading any tech review...be it software or hardware. Most people don't know what they're talking about. The only reviews I can rely on are Anandtech and Tom's Hardware reviews...but I'm wary of these too.
True, and soooooo sad:
I like good honest non bias reviews, I dont have the money to "test" tons of configs, new parts, etc, so I up till a year or so ago, took Macworld, PCmag, and so on at their word, boy was I wrong...
Looks like if I want benchmarks, I will take a copy of 3DS MAX to CompUSA with me
I think the only way to even come close to testing anything is to do a specific test of one aspect of the system using a freely multithreaded app. RC5-72 was a good comparison that was fully multithreaded and Altivec-aware that showed the G4s and G5s destroying any x86 chip.
If these guys can't find multithreaded apps, they need to write one. It only takes a few hours. OS X has all the multiprocessing APIs that anyone would need.
Digg this post and get the word out - G5s aren't obsolete by far!
http://digg.com/apple/PowerMac_Quad-...ted_Computing_
As an aside, team Canada was in 20th place when we joined, and now is in 4th... :-) Go Canada!
http://szdg.lpds.sztaki.hu/szdg/team...php?teamid=183
Developers want to add features, so they can market the features.
While they may wish to speed up or multithread or make multiprocessor savvy, it is not something that they can usually market as a feature, so it doesn't get done.
There is so much bloat and inefficiency in modern commercial software that it is ridiculous. Paying an engineer to optimize something costs a lot more than just jacking up the hardware speed, especially if optimizing can't be sold as a new feature.
The IBM 360/67 ran dozens of virtual 360/65s and batch jobs simultaneously, all in 512KB (KILOBYTES) of memory. No room for bloated runtime overhead in that kind of code.
Originally posted by lundy
There is so much bloat and inefficiency in modern commercial software that it is ridiculous. Paying an engineer to optimize something costs a lot more than just jacking up the hardware speed, especially if optimizing can't be sold as a new feature.
The IBM 360/67 ran dozens of virtual 360/65s and batch jobs simultaneously, all in 512KB (KILOBYTES) of memory. No room for bloated runtime overhead in that kind of code.
Yes, and no room for solving problems in a human friendly way. While they ran them simultaneously the typical process was to submit your card deck and expect results the next day. A misplaced card often meant a blown day much less real syntax errors (yes, I've seen card sorters make mistakes).
Which is why some folks thought "Clean Room" software development methodology was a good idea vs just compile the damn thing and see if any syntax errors pop up.
Give me "ridiculous" modern bloat and inefficiency. The old way was necesary because of the limitations of the day...not some virtue. We went for abstractions to solve larger more complex problems than can be solved via batch jobs.
Why don't you go run OS/360 on a linux machine and build yourself some really tight code? Because it isn't overly useful except as an exercise in nostalgia.
Vinea
Used TSO, OS/MVT and OS/360. COBOL and FORTRAN.
Originally posted by vinea
Yes, and no room for solving problems in a human friendly way. While they ran them simultaneously the typical process was to submit your card deck and expect results the next day. A misplaced card often meant a blown day much less real syntax errors (yes, I've seen card sorters make mistakes).
Who said anything about batch jobs? I specifically said virtual 360/65s. From IBM Selectric terminals. Interactive.
Which is why some folks thought "Clean Room" software development methodology was a good idea vs just compile the damn thing and see if any syntax errors pop up.
Absence of syntax errors tells you nothing about the correctness of the code.
Give me "ridiculous" modern bloat and inefficiency. The old way was necesary because of the limitations of the day...not some virtue. We went for abstractions to solve larger more complex problems than can be solved via batch jobs.
Again, I am not talking about batch jobs.
And the bloat is not in just the complexity of the application - the bloat is due to astronomical levels of subroutine calling, message passing, making the integer 5 into an object and passing a message to it asking it what its value is, and so forth.
The original QuickDraw on the Mac 128K was 24K of code. Imagine if we had that kind of efficiency on today's hardware.
Launch MS Excel for Mac and try to move the insertion point in the formula bar. It takes about a half second to move to the next character. It's calling about 30 subroutines to do that, whereas the same thing on Excel for Mac 512K would be maybe 2 assembler instructions.
Yes, the hardware is dual 2 gHz instead of 8 mHz, but the bazillions of calls of repetitive code that hardware has to wade through just to put a character on the screen nullifies a large amount of the gain we got from the hardware increases.
I lament it, but I know it's not going to change. As I said, programmer time is way more expensive than hardware.
That's why I don't expect any effort to be made to remove bloat in Leopard. It isn't cost-effective.
I suggest to copy and paste all important posts into TextEdit until we get the re-indexing done.
Originally posted by lundy
I suggest to copy and paste all important posts into TextEdit until we get the re-indexing done.
Good thing we have automator...