Slashdot Mirror
Ars Technica Reviews Intel iMacs
Milton Waddams writes "Ars kick off what I'm sure will be a torrent of reviews of the of the new Intel iMac. Overall it looks like it's a bit faster than the iMac G5 and a bit slower than the PowerMac G5 dual core. I'm sure it will surprise many slashdotters to find out that Jobs' statements about the new iMac being twice as fast as the iMac G5 as being slightly over optimistic. AND it doesn't run Windows...yet..." I'm still waiting for the most important benchmark: frames per second in molten core combat.
YouTube.com has a video of both systems booting. So if you're in to computer drag racing here ya go: http://www.youtube.com/?v=zmaAZwkhYeQ
http://religiousfreaks.com/not to mention the intel GMA900 graphics processor isn't exactly the speediest thing around.
1000fps in glxgears? I can beat that by a good 50% with a 4 year old NV GE440 go in my compaq laptop
watch what happens when there are nvidia drivers and ATI drivers available.
P.S. ATI 800 series cards do work (fully accelerated) on the development platform.
I'd post a link but the lawyers are loose.
The truth about Led Zep should never be told on
Look at the history of Apple's processor switches. The first generation PPC machines (6100/7100/8100) were nice, but the second generation PPC machines (7500/8500/9500) were much better. The 2nd gen PPC machines had PCI instead of NuBus, a faster interleaved memory architecture, and a much improved dual-SCSI bus. With the first Macintel, it's obvious that Apple worked very quickly to put Intel Inside and I'm sure that some parts of the design represent a borrowing from PPC designs. I bet that second generation Intel machines are both faster, less likely to have flaws, and more likely to enjoy longer-term OS upgrades.
I know its ungeek of me not to want to be on the bleeding edge, but I'm waiting for the second generation machines.
Two wrongs don't make a right, but three lefts do.
The fat binaries (or Universal) has been around for a very long time. NeXT first introduced them in this exact same format. Knowing that Mac OS X is the current incarnation of NextStep (the OS for NeXT machines), I assumed the Intel move would be relatively easy, if the had been maintaining the code.
Now, in retrospect, it looks like they have for Mac OS X, but maybe not for all the other applications (iLife, FCP, etc).
Now, given that the OS has a long history of multi-platform support, it is only a piece of the puzzle.
Application level changes are a bit harder, especially in relying upon functions specific to a chip. Which, for some applications, is the case. Others should be able to do a direct recompile, if the application is still around in source form, the author is interested, etc.
Back when I had acess to NeXT Cubes, I didn't have to worry about it. However, when I later bought NeXTStep 486, I had to. There were lots of applications for the 680x0 systems, I sometimes had to search for those 486 applications. I assume we are headed back into that world.
So, can it happen? Yes. However, I suspect that Apple will move on with the Intel architecture. I assume the PowerMac G5 will be a well respected machine in the meantime, as it does great for video editing, something Widnows machines still work hard to do poorly.
I suspect it might be like the Amiga. While the Amiga didn't get a lot of respect, those in the video editing world used it much longer than people antipicated.
But, in the end, the new macs will be Intel. As a side note, I just sold my G5 DP to someone looking to do video editing with FCP. Even with them knowing the Intel systems were coming out, they still wanted it.
The fact is that lies and deception are (a) common and (b) shouldn't be. It is completely legitimate to (a) criticise liars and decievers, even if you haven't done so and (b) (as the submitter did) highlight lies and deceptions to reduce their impact. If Joe Blow tells me that Blow Inc's Widgets are "99% plutonium free", by which he means that 1% of each widget is plutonium, rather than that they contain 100 times less plutonium than his competitors as would be the most likely initial reading, I'll tell people. I'll not sit there and say "Hey, there are probably a lot of people who think that his widgets contain less plutonium than everyone else's" simply because of some misplaced "moral code" that says I shouldn't be seen to be apparently critical of someone for doing something that a lot of other people do.
The Intel iMac isn't significantly faster than the iMac G5, despite Job's attempts to imply otherwise. Indeed, if the iMac G5 had undergone the same revisions that the PowerMac line had a few months ago, the chances are they'd be faster than the Pentium equivalents. I've written in my journal that I really don't understand why people reacted with such enthusiasm to the Stevenote this year: read between the lines Jobs spoke and you find that Apple actually was struggling to come up with convincing evidence that justifies the Intel switch. Nonetheless, Jobs used wording that implied massive (2-4X!) improvements in performance, which, surprise surprise, are clearly massive exaggerations and, in the context of comparing this year's Pentiums to last year's G5s, are actually covers for practical failure on their side.
If you're going to use wording that implies massive (and 2-4X really is massive) improvements in performance, don't expect your honesty and virtue to be unquestioned when we subsequently find that no such improvements have actually occurred. If Michael Dell does the same thing, we'll criticise him too. If Intel's CEO does, we'll do the same thing. Today it's Steve Jobs.
You are not alone. This is not normal. None of this is normal.
This comparison really makes me wonder about the Intel-based PowerMac replacement. What kind of processor are they going to put into that? The logical choice would seem to be the Conroe. There are rumors of a 3.3 GHz dual-cores being sent out later this month. Intel claims that Conroe will outperform Core Duo 2-1 on a performance-per-watt basis. So a 3.3 GHz Conroe might be as much as 3.6 times as fast (pure performance, it's ok for a desktop chip to consume twice as much juice as a laptop, right?) as a Core Duo. So if you take the iMac comparisons against the current G5s and extrapolate... well a PowerMac based on a Conroe could be a mighty beast. Give it 4 GB of RAM like the PowerMac in the comparison, and it should easily outperform the PowerMac (at least on non-Altivec tasks, but that's a different story.) Of course it's still going to have the stupid front side bus, albeit running at 1.0+ GHz with 4 MB of L2 cache.
Another possibility would be for Apple to wait for the Extreme Edition of the Conroe, the Kentsfield. That would give them four cores, like the current PowerMacs. It won't be out until 2007, and Apple seems anxious to switch everything over ASAP. So they could go with Woodcrest, basically Conroe for servers. This might let them put together a dual-cpu/dual-core setup like they have with the current PowerMacs. This kind of setup was demonstrated by Intel last fall. There were also rumors last year of Apple pressuring Intel to give them Woodcrest chips ahead of schedule.
And of course there's the more mundane question of what will they call the PowerMac replacement? They seem to want to get away from the Power prefix, while stressing the Pro tie-in to their Pro apps. So maybe Mac Pro? Seems too short. Maybe bring out the whole name, Macintosh Pro. Whatever it is, can it make people as upset as "MacBook" did?
That would depend on the working set size for the given app.
512MB may be slightly cramping the style of the new imac, but it didn't look like any of Ars' benchmarks would need much more than that. Certainly 4.5GB isn't going to make any difference, and if you've been following Ars' articles, you'd know why that particular machine is so loaded. The CPU-bound, disk-bound, or graphics-bound benchmarks aren't going to notice the change in RAM amount. The photoshop test, being done on a fairly large image, might have seen some impact from the difference in available memory.
Given how heterogenous the systems are already, I'm not too concerned with a slight difference in memory size. Given the different instruction sets, execution hardware, cache layout, and memory controller, I think having only 512MB rather than a gig is unlikely to show up in the benchmarks or in most users' usage.
High-speed Road Trip (18.000KPH)