Intel Developer Macs Outperform G5s 829
bonch writes "Developers working with the new Intel-based, developer-only Macs are impressed with the performance. The machines take as little as 10 seconds to boot from Apple logo to desktop, and apparently run Windows XP at 'blazing speeds.' Rosetta tests demonstrate the PowerPC-native build of Firefox running just as fast as it does on a high-end G5."
The real question (Score:2, Interesting)
Re:Dual Boot (Score:5, Interesting)
some apps suffer? (Score:2, Interesting)
Doesn't sound like Rosetta is transparent for everything, then?
Impressed (Score:4, Interesting)
Re:Good news! (Score:3, Interesting)
Comparing FFox under OSX is nothing new. G5 is slower CPU than P4, but at certain jobs, quite a lot faster (with that I mean jobs when PPC functions were actualy used). It would be better to test Photoshop or some video application that was noticeable faster than the same app on Windows, which means that it actualy uses quite a few PPC functions to the fullest.
Re:So the G5 were dog slow after all (Score:3, Interesting)
Re:Dual Boot (Score:4, Interesting)
You would if it was important enough to you. I bought mine so that I could support Apple users. i.e. I saw a very real use for the machine. (Best purchase I ever made, BTW.) With Linux, there's just too much noise and not enough signal to make anyone want to purchase a Linux-built Desktop machine.
Re:Boot times disk/network bound (Score:5, Interesting)
Also, since plugins cannot be emulated, there is no way for anyone to install kernel extensions that slow down the boot times of OS X.
In other words, the speed these people think they're seeing are actually do to a horrific lack of features.
Integer vs. Altivec (Score:5, Interesting)
I think most of us expected the P4 to perform better for Integer like code on applications that don't effectively SMP. So that's not that surprising. I am surprised at the speed of Rosetta, although that will be a mixed bag once again depending upon the application.
What I'm really interested in is speed on stuff that really leveraged Altivec, like A/V programs. I'm curious about Quicktime 7 for instance. Now some of these programs can use some similar functions on the P4. But from what the Altivec folks were telling me some code ought differ by as much as 50%. (i.e. the PPC is twice as fast) A nice simple test is to compare programs like iMovie on both platforms.
Re:Dual Boot (Score:5, Interesting)
But seriously, I hate the "Apple makes money selling machines, not software" myth. Apple also makes a killing on software, and there's higher margins than hardware. Steve Jobs said Tiger had already sold a million boxed copies of Tiger at the WWDC. Multiply by $129? That's a lot of cash coming from just the OS. And don't forget about their stance in the professional media market. How much is Final Cut Pro selling for these days? What does Motion cost?
no (Score:5, Interesting)
Maybe G5s are not so fast. But:
"It's fast," said one developer source of Mac OS X running on Intel's Pentium processors. "Faster than [Mac OS X] on my Dual 2GHz Power Mac G5."
So, uh, a 3.6 Ghz P4 is faster than two 2 Ghz G5 - 4 Ghz? - SMP, but 4 Ghz.
Sorry, I don't buy that. Even more if you take into account that Intel isn't exactly the performance/Hz leader - in fact it's the worst performer these days, Opteron and PM beats the sh*t out of that P4 at much lower speeds from what I've read.
There're lots of factors that can change things - freebsd algorithms, are, for one, optimized for i386 variants. Also, Mac OS X is compiled with -Os - optimized for size, no for speed. (Paranoic mode on=Hey, maybe this switch was planned and it's not a coincidence)
And then there's the Placebo effect [wikipedia.org]. IOW: Show me numbers, don't tell me "it's fast", I don't trust you. In Linus' words: "If we can't measure it, it doesn't exists". Unless someone writes a decent comparative, I'll take this article as Apple Marketing - Apple has been very critized for this change, I wouldn't be suprised that Apple is interested in articles like that, showing how good move has been the switch to intel
Excellent (Score:1, Interesting)
A Pentium 3.6ghz (brand new) is almost twice the clock speed of a single G5 2.0ghz (rather old). On top of the G5's age, multiprocessing isn't a linear performance increase. Two 2.0ghz CPUs are slower than a single CPU at 4ghz. Yes, I know they don't match clock for clock, but Intel keeps optimizing (HT, for example).
What *IS* surprising is the PowerPC emulation. 70% native speed, even 50% is astonishing.
After Apple races out the Intel boxes, they'll be even faster. Remember, Apple is the ultimate modder. The lengths they went in CASE DESIGN and WATER COOLING just to get the G5 to run as fast as it does.
They're going to make some unreal boxes. Yeah, they'll be PC's, but Dell will have nothing on them.
This is encouraging. I wonder how native Photoshop compares. That's all I really care about... Photoshop and vim.
I'm all warm and fuzzy inside. If only they'd use Itanium2 chips also.
Re:Boot times disk/network bound (Score:5, Interesting)
Apple has really been pushing people to Sleep the machines instead of turning them off. Sleep mode uses a very tiny amount of power, and you get your instantaneous boot (with apps open and windows positioned...). I have been doing this with my Macs since OS X appeared and let me tell you it is the only way to go. Especially on laptops. In fact I am still using a CRT on my G5 and the computer 'boots' faster than the monitor (warm up).
DVD leak when? (Score:3, Interesting)
Mod down, bs (Score:3, Interesting)
These Apple boxes will use generic harware found in your standard hp or dell box. Home made pcs will run MacOS just fine. Apple is using a DRM to raise the barrier of entry to nearly infinity for competing hardware companies who want to sell mac clones to maintain a monopoly on selling hardware that runs MacOS. That way they can charge a premium on hardware to subsidize their OS development and make a healthy profit by choosing their own price without worrying about market forces. Apple knows their product is differentiated enough from an MS PC that these machines are not really competing head-to-head, so my monopoly arguement is valid.
So the lesson is (Score:3, Interesting)
It would be really fascinating to see what the sales figures will be like for the next year or two.
Re:Boot times disk/network bound (Score:3, Interesting)
Even if a modern OS has dumped 200 megs of stuff into RAM by the time it's finished booting, that's easily affordable at today's flash RAM prices. You could fit a 256 meg chip to the motherboard (which I'm sure both Apple and MS could mandate if they wanted). If for whatever reason more space was required, the overflow could be put on disc.
And since a boot is a fairly rare event, you don't have to worry about the maximum number of write cycles with flash RAM.
Hey presto - (almost) instant boot!
Re:Integer vs. Altivec (Score:4, Interesting)
Re:Intel Mac's poor hd performance (Score:2, Interesting)
November '98 - My Mac Sucks
http://www.kottke.org/98/11/ [kottke.org]
June 2002 - Switching to the Mac
http://www.kottke.org/02/06/switching-to-the-mac [kottke.org]
Wait a second . . .
Re:WinXP on Mac a fluke, Mac OS X Apple H.W only (Score:3, Interesting)
I rememember what happened to OS/2, with their vow to make "a better Windows than Windows". They did a good job of making sure that Windows apps ran well under OS/2... so good, in fact, that many app vendors stopped developing the OS/2 versions of their software. (After all, why spend money to develop both a Windows and an OS/2 port of your software, when OS/2 customers can just run the Windows version?) The result: less "native" OS/2 software, and the eventual decline and death of OS/2.
Apple may want to watch out for that trap...
Re:So the G5 were dog slow after all (Score:3, Interesting)
Plus with details from the article as technical as "it's fast," and "blazing speeds," you just know that's some newsworthy stuff there.
Also the opposite is also going to be true. An intel compiled binary is going to stomp all over a PPC binary run with Rosetta.
Why even bother making such a comparison, and even more so, why bother making such a claim?
If you absolutely have to run PPC code on Intel you will, and you're not going to be impressed at the speed.
And as for comparing the intel hardware to PPC hardware it's Apples and oranges as far as I'm concerned. It seemed like the article was going as far as to say an Intel osx machine is just as fast when running PPC binaries with an emulator than a top of the line dual processor G5 running native PPC binaries?
Come on! Give us a break. Those would be fine claims to make, as long as they were backed with something a little more concrete than "my web browser seemed to run about the same."
Like I said, I'd love to believe this, I'm no fanboy either way (intel/ppc), and I've never owned a Mac, but this whole article is trite.
If you've got some developers with access to the machines, just spend the 15 minutes and do a couple benchies.
Re:Wrong meaning of "Blazing" (Score:2, Interesting)
Re:Boot times disk/network bound (Score:4, Interesting)
Your point makes sense though about making some sort of assumed last known configuration the default. This would require the user to hit a button if they change their config so a full reboot with full PCI scan could ocurr - probably not too user friendly.
Re:Boot times disk/network bound (Score:1, Interesting)
Holy shit, I would have thought my Mac's battery had died if I had to wait 10 seconds for it to 'resume'.
From opening the lcd latch to putting my fingers on the keyboard, my powerbook is ready to go. 2 seconds tops, and that's if I left a bunch of apps open.
Re:Big-endian vs. little-endian (Score:2, Interesting)
The IO libs just take care of it.
You're kind of missing all of those Mac apps which were written prior to OS X and those afterwards which are based on Carbon and not Cocoa, or which are pure C/C++. The Cocoa IO libs might handle byte swapping ( in fact they do ) but neither the Carbon IO APIs or the C/C++ standard IO APIs do, and a lot of apps use those. MacOS development != NeXT derived.
Jo Meder
Re:Boot times disk/network bound (Score:3, Interesting)
Because flash RAM is slow. Unless I've lost touch with the latest tech, the average HDD provides about 4x the throughput of current flash memory. Next-gen flash is better, but still on the same order of magnitude. A battery-backed (volatile) RAM boot disk could be nifty...
The LinuxBIOS [linuxbios.org] project lets you boot up fast out of flash, but that's mainly because you get the skip all the useless crap that the PC BIOS wastes time on.
Re:Boot times disk/network bound (Score:3, Interesting)
When I upgraded my old 1.7 GB disk to a 13 GB one, bootup got *lots* faster. However, now the CPU was the major stopper, and upgrading from Pentium 166 to Celeron 400 again resulted in a considerable speedup. Upgrading to a 40 GB afterwards disk didn't change that much -- the processor is still the slowest part in the system. After another CPU upgrade it would be different again, of course...
Rosetta only translates G3 code, not G4/G5 (Score:4, Interesting)
Re:Good news! (Score:1, Interesting)
Re:Boot times disk/network bound (Score:5, Interesting)
Take XP, load Office 2003, Norton AV, the standard CD burning and DVD viewing apps, and watch the performance (and boot times) degrade considerably.
Re:Boot times disk/network bound (Score:5, Interesting)
They do, do they? Where was it confirmed? And at what point in the keynote did Jobs ever show off anything but iPhoto and iTunes? http://www.apple.com/quicktime/qtv/wwdc05/ [apple.com] There's the keynote. Go ahead, tell me where.
and FWIW, I was at the keynote. What's in the dock during the keynote is all that comes with the dev kits.
I call shenanigans (Score:3, Interesting)
Re:Impressed (Score:3, Interesting)
QEMU [qemu.org] aims to do the same:
It can run (to some degree; it's still in development) on x86, amd64, PowerPC and a host of other CPUs, and it can run binary code for x86, amd64, PowerPC, SPARC, ARM, and MIPS.
Re:Boot times disk/network bound (Score:3, Interesting)
Some things need to be re-initialised if left for any length of time (Ie, DHCP usually has a timeout)
All open Sockets would probably have to die (But not nesseserilly normal file handles)
There are probably other things as well, but these are the main things that occur to me.
Any processes which rely on hardware parameters would have to be re-initialised, eg PCI bindings, Ram size. Also any hardware state would have to be reset because you can pretty much garantee it wont be in the same state as when you suspended. There are a whole world of things that can go wrong there. Personally would prefer a long boot-up time (Providing the boot-up is reliable) and a long uptime. Then again, i do tend to leave computers on more than is environmentally friendly.
Re:Boot times disk/network bound (Score:4, Interesting)
Re:Boot times disk/network bound (Score:4, Interesting)
Now those of is in the here see that as nonsense, since filling up the hard drive with not-currently-executing code does not have any impact on memory usage or CPU usage.
However, a lot of Windows programs have this tendency to install things that "always run in the background", and that does eat memory and impact CPU load. Back in the day when I actually used Windows a lot (and when RAM was still expensive), it was commonplace for people like us to spend the time digging through the Start Menu and the Registry to disable all those little side-processes.
As a result, people like me had machines that were MUCH faster and more responsive than most normal people with their storebought machines with specs usually much better than mine on-paper.
I suspect the same may be true today, between store pre-loaded crap and resident bits of installed software, even if cheap RAM has averted some of the issue.
Re:Boot times disk/network bound (Score:3, Interesting)
Re:Boot times disk/network bound (Score:3, Interesting)
Re:Integer vs. Altivec (Score:3, Interesting)
Many of the Altivec programmers, however, note key features of Altivec which *aren't* available in SSE2. I've heard mixed things about whether SSE3 will solve these issues. Also (to the other poster) while one can argue that Altivec was simply "strapped on" to the G5 and not as elegant as in the G4, the fact is that the wider bus on the G5 will help programs using Altivec which often are AV programs that really can use the bus increase.
Over at ARS for about a week after the announcement there were many disgruntled Altivec specialists bemoaning the loss of Altivec. While I don't think the SSE* derivitives are as bad as many make out (at least from my limited reading) I think there is some justifiable skepticism over the issue. The main reason G5 Macs could perform many functions as fast as they did and keep up (and sometimes surpass) Intel class chips was Altivec. In general integer code G5s were always behind. (Thus the careful Photoshop bakeoffs utilizing places where Altivec shined)
Since most programs are these general integer programs I think most programs should significantly improve on the Intel systems over more or less equivalent G5s. As I said in the original comment (which many seem to have misread) it will be interesting seeing how AV programs perform on the Intel boxes. I'm getting one soon, although I'll then be under NDA and won't be able to talk about it. (We're just working out the accounting for the rental at the moment)