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.
http://arstechnica.com/reviews/hardware/imac-cored uo.ars/7
I tried to boot from a Windows XP installer CD. No dice. I then tried booting from a Vista installer DVD (Build 5270). Again, no dice. When holding down the Option key, the only icon that appeared was for the iMac's internal hard drive. Holding down the D key to try to force booting off of the optical drive failed as well. With the Vista DVD, the optical drive churned a bit and the iMac hesitated as though it were contemplating whether it wanted to boot the foreign OS. Soon afterwards, the familiar gray Apple logo appeared on screen and Mac OS X finished booting.
The new Intel Macs don't have an EFI shell, so there's no way to directly get at the EFI. Someone is ultimately going to have to write and/or use an existing EFI shell to tell the EFI to boot from alternate media to get things going. Naturally, running Windows under virtualization, with technologies like Intel's VT/Vanderpool, which the Core Duo in the new Macs does support, are going to be the way to go for most users anyway.
There's obviously something seriously wrong with that G5 if it is taking that amount of time to boot. My G5 boots in roughly the same amount of time as that Intel iMac.
Also, apparently on the new Intel-based Macs, one holds 'D' instead of 'C' to boot from the optical drive - presumably because Mac OS X ships on DVDs by default.
And tomorrow the stock exchange will be the human race
Actually, the new Yonah processor is significantly faster than the Turion. Check out these benchmarks:
d uo_notebooks_trade_battery_life_for_quicker_respon se/
http://www.tomshardware.com/2006/01/16/will_core_
If you need web hosting, you could do worse than here
By all accounts, the new machine's insides are practically designed by Intel. Intel CPU's, Intel chipsets, probably Intel motherboards. While the SW for Intel will definitely get better over time, I doubt the HW will be less bug-prone (it's already very solid).
The only big change on the horizon is the switch to Merom/Conroe/Woodcrest in the second half. This will bring the eventual switch to a fully 64-bit OS X.
Dan
That was my thought initially too, wait for the second generation. Its probably still good advice, but I think they fact they are using a widely implemented Intel product will increase the reliability of these first generation Intel macs. My first generation pentium-m is still a good performer today. Apple will be using the standard processors, boards, and components of the Centrino duo, which will have a long a lifespan. There is not going to be some earth shattering change in the Macbook components a year from now. Apple will be using the same Intel specs that every other major pc laptop manufacturer will be using, so we can probably through all the first-gen history out the window.
I think the 68k-PowerPC transition is not quite apt. You are correct that the first generation of PPC boxes were not nearly as nice as the second generation boxes, which had much more extensive changes than just the processor. But I think there was a lot more room for improvement in the last 68K boxes versus the PC state of the art at the time than there is in the last PPC boxes versus the PC state of the art. The NuBus expansion card bus versus the PCI bus is just one example. Both the last PPC boxes and the first Intel boxes have the latest and greatest interfaces such as USB2 and PCI Express.
In addition, the first Intel box is not a motherboard that Apple slapped together on its own, like it did for the first PPC boxes. It is a state-of-the-art Intel motherboard with all the latest doodads. Sure, Apple could stick in a fastre graphics card or hard drive, but the motherboard support chips are all modern. I think the next rev of the MacBook Pro will include Firewire 800, which I assume Apple couldn't integrate into the MacBook Pro motherboard in time to meet their ship date, but that's more of an incremental change than was included in the second-generation PPC boxes. (And don't forget the problems with the then-new PCI-bus Macs. Networking was so broken -- remember the Open Transport fiasco? The brand-new networking architecture that wasn't ready at the time the 7200/7500/8500 were released and which those new machines relied on, MacTCP having been deliberately obsoleted? It took several months after the boxes were shipped for *any* PPP dialing software to work at all with the Mac, and it took more than a year after that until most of the more significant networking bugs to be quashed.)
The 64-bit version are out later this year.
Not that many people actually need 64-bit capabilitity, mostly for programs that need very large memory access. iMac users certainly will not, Macbook Pro users is more questionable - my guess is they will upgrade that line with the 64-bit chip at the same time they release the Intel Powermac equivilent.
Hmm, that leads me to wonder what the new name for the Powermac will be... MacMac?
"There is more worth loving than we have strength to love." - Brian Jay Stanley
EvilTwinSkippy is correct. AMD is practicing just as much deception with this scheme as the rest of the industy. I'd remind you that it was AMD who used started the MHz war by consistently clocking their CPUs higher than Intel's. Intel responded with the Pentium IV which was able to clock higher than the AMD chips even if the performance wasn't better. AMD responded with their "speed rating system" which is intended to trick consumers into believing that they are getting a faster clock rate than they are actually getting.
For example, an Athlon XP-M 2400+ is clocked at 1800 MHz, not 2400 as one would expect. The AMD Athlon XP-M 3000+ is clocked at 2200 MHz rather than 3000 MHz. Even more deceivingly, there is another chip called Athlon XP-M 3000+ ("Dublin") which is only clocked at 1600 MHz! How am I, as a consumer, supposed to sort all this nonsense out?
So you can keep your comments about "conservative" ratings. The industry is all about deceptive practices, and (most likely) always will be. Learn to deal with it.
Javascript + Nintendo DSi = DSiCade
Apple practically gives away Xcode to software producers
Practically? Apple literally gives Xcode away. It's free. It even comes with every new Mac, right in the box. And anybody who wants to sign up can get updates and SDKs at no charge.
How many people know the new macs actually ARE pc's?
How do you define a PC? Is it the CPU architecture? Is it the manufacturer of the CPU? Is it the company that wrote the prevalent operating system? Is is the company that first called its product a 'PC', 25 years ago?
Why make things so hard when the answer is right in front of us?
PC = personal computer. All Macs ever produced have been PC's. For that matter, so were the Apple 8-bit computers.
The intel iMac supports spanning! I'm surprised Steve didn't make a big deal about this. There goes one more major reason for people to buy a powermac. Kudos for Ars for mentioning that on the first page.
the more they over-think the plumbing the easier it is to stop up the pipe
glxgears is a benchmark for graphics acceleration. You don't (unless you're insane) sit there and watch it spin at 1908 fps.
What I'm listening to now on Pandora...
That's entirely untrue. You can easily spot the difference in smoothness between video captured in 24FPS and video captured in 60iFPS.
Try playing a video game at 25FPS, and then at 60FPS. Can't tell the difference? If you can't, you've got to be full of it.
Maybe your brain can't absorb all the information on each frame past a certian point, but *anyone* sure as hell can see the difference when it comes to smoothness and fluidness of movement.
And a note about GLXGears - the higher the number, the better chances of getting more complex objects on the screen at a decent frame rate. If you haven't noticed, games are a little more detailed then GLXGears. So while you can spin a few objects at 2000FPS, you might only see 20FPS in the latest game title. But if you get 10,000FPS in GLXGears, you'll probably see much higher performance in the game. It's a BENCHMARK. Seriously.
And what does this mean: "a lot of resources are wasted computing and rendering"? Explain to me what else you want the computer doing when you're running a graphics benchmark? I want mine running the damned benchmark, what else? It's not like everyone's machines are attempting to cure cancer and we should let that happen at all costs. I buy fast computers because I want to use all of the speed, not have an abundance unused in the background.
- It's not the Macs I hate. It's Digg users. -
Jobs cleared that up when he made the announcement last year. XCode can make universal ("fat") binaries outta the box. You don't need two computers to do it, only if you want to run extensive testing.
One note I haven't seen anywhere else that was mentioned in the Ars review is that the new Intel iMac supports monitor spanning, and has a mini-DVI connector. AFAIK, the iMac G5 only supported mini-VGA and mirroring. (Well, there is that hack to enable spanning, but with only an analog video-out, it isn't that that useful for me...) It always seemed like a trivial crippling of iMac to force users who want/need desktop spanning to upgrade to a PowerMac.
Who here actually watched the keynote? Show of hands? I know I did. Let's all go to www.apple.com/quicktime/qtv/mwsf06/ and load the keynote up to 1:07:00. Steve Jobs is completely up front about which testsproduced the numbers (SPECint_rate2000 and SPEC_fp2000) and outright says "Now, everything is not going to run 2-3x, the discs aren't 2-3x faster, etc." He makes it very clear that his numbers are based off of these two benchmarks. He claims they are the most important benchmarks of performance, which is debatable, but they are certainly a fair test of raw cpu power. Other than the chip and motherboard, the only other significant component that has changed is the GPU, going from a Radeon x600 to an x1600. Does anyone disagree that this is in the 2-3x faster range? All in all, people are making a mountain out of a molehill rather than checking the source of the numbers. god bless the internet. -justinb
The "D" key is the new key for booting optical media on Intel-based Macs:
3 081
http://docs.info.apple.com/article.html?artnum=30
Who here actually watched the keynote? Show of hands? I know I did.
Let's all go to www.apple.com/quicktime/qtv/mwsf06/ and load the keynote up to 1:07:00.
Steve Jobs is completely up front about which testsproduced the numbers (SPECint_rate2000 and SPEC_fp2000) and outright says "Now, everything is not going to run 2-3x, the discs aren't 2-3x faster, etc." He makes it very clear that his numbers are based off of these two benchmarks. He claims they are the most important benchmarks of performance, which is debatable, but they are certainly a fair test of raw cpu power. Other than the chip and motherboard, the only other significant component that has changed is the GPU, going from a Radeon x600 to an x1600. Does anyone disagree that this is in the 2-3x faster range?
All in all, people are making a mountain out of a molehill rather than checking the source of the numbers. god bless the internet.
-justinb
While the dev systems uses intel GMA900 integrated graphics, both the MacBook and new iMac use ATI Radeon X1600 chips.
The iBook and mini may use integrated graphics, but they will probably use newer chipsets with graphics faster than the GMA900.
Dan
MacWorld demo people, although trained not to demonstrate or allow rebooting the new Macs on display, did remark that they booted up really fast.
In the video, the G5 likely had more RAM installed, which would make it POST considerably slower. The boot time, however, is probably very representative of how much faster the Intel iMac is at booting. Other reasons the Core Duo may have booted so fast compared to the G5:
- Two processor cores!
- Mac OS X is expressly designed to boot fast by bringing up as much as possible in parallel. That's part of the point of launchd: to identify dependancies and kickstart multiple things at once. This is also why Apple gave up on displaying what was being booted in 10.4, and now just shows a progress bar (which is unrelated to what's actually happening, and only timed to match the previous boot time as a relative indicator). Reporting what servers are being launched would take longer than actually starting them. This parallelism would clearly benefit from multiple processor cores in the Core Duo.
- the G5 may have been booting for the first time, or they may have deleted the cache in an attempt to make the test "fair," not realizing that the cache has a huge impact on boot times. Among other things, Mac OS X caches the kernel extensions so that the next boot only stops to numerate which kexts to load if something in hardware has changed. If you wipe your cache files (/Library/Cache, ~/Library/Cache and System/Library/Cache), the next boot will take a lot longer while boot performance caching is rebuilt.
- other hardware may have been unfairly compared: how fast was the G5's drive? was something wrong with it? was the G5's drive full, and struggling to find space for cache files? was it bound to a directory server, and stalling on boot while looking for the server? was it full of 3rd party software, kexts, startup items, etc?
The video doesn't reveal anything about the demonstrators competence at setting up fair comparison, or their motivation, so we don't know.
Recall the comparison of database servers running on OS X server vs Linux, where they intended to be fair but their assumptions about how to do so were actually really bad?
Or look at the Ars review and benchmarks of the new iMac Core Duo vs the iMac G5. He does an array of benchmarks where the G5 has 1 GB of RAM, and the Intel iMac has 512MB! Sorry Ars, but that's just plain incompetent. Your benchmarks are WORTHLESS to even skim over. How about benchmarking the G5 iMac with 512 and 1 GB installed, and reporting if that makes any difference?!
All of Intel's Core chips are 32-bit. The *next* generation (a.k.a. Merom) will be 64-bits and comes out in September. I hope this helps.
Just "gittin-r-done," day after day.
No, it is not.
Holding the D key is for booting Apple Hardware Test only. You cannot boot from a CD normally in this way.
C is still the correct way of booting from CD on intel.
Right you are! I just verified myself with Apple Enterprise Support:
D - diagnostic partitions only (Apple Hardware Test)
C - optical media
I stand corrected. Thank you.
Consistently, since even the early 486 and Pentium days, AMD (and in fact also Cyrix) CPUs routinely beat Intel CPUs running at somewhat higher clock rates.
Too far back in history, chief. In the 486 days, the AMD and Cyrixes were nothing more than a "cheap" upgrade for a 386. They simply didn't compete. Consumers were thus able to somewhat trust the MHz rating of Intel chips as a general performance indicator.
When the Pentium era arrived, AMD was still not a competitor, but they did manage to produce chips that were "good enough" to be considered cheap alternatives to the Intel processors. As a result, the PC industry did start producing AMD-based machines for their "low cost" product lines.
When the Athlon and PIII came to market, however, is when things got interesting. For the first time, AMD managed to put out an extremely capable chip in comparison to Intel's offerings. But far more interesting was that AMD started ramping their chips to exceptionally high MHz levels to pass up Intel's chips. This practice gave the AMD chips a reputation for high performance, but extreme heat levels and lower reliability. This left Intel with the server market as their chips proved more reliable over the long haul, and performed just as well in most non-gaming situations.
Long story short, Intel and AMD traded various blows in performance, each one gaining a slight lead for a short time only to be quickly shown up by their competitor. AMD, however, managed to keep the MHz trophy the entire time.
Intel got the bright idea to beat AMD at its own game and thus produced the Pentium IV chip. Now the PIV isn't a very good chip, but it can be ramped up in MHz in ways that AMD's Athlons couldn't. AMD struggled for awhile, but quickly realized that they could no longer win the MHz trophy. So they came up with a Jedi Marketing Trick to make consumers think that AMD's chips were running at the same clock levels. That trick was to assign an arbitrarily created "speed rating" that placed the direct competitors to Intel's chips right next to the Intel MHz rating. That way reviewers would pick up the slower AMD chip and compare it with the Intel chip, rather than look at the real MHz.
AMD got a lot of bad press for the decision, but it did eventually pay off. Consumers accepted the PR rating as the "real speed" of the chip, and Intel again started losing market share. AMD finally ripped the market away from Intel with their AMD64 platform, which proved to be much more popular with consumers than Intel's own Itanium line.
So to summarize, AMD started competing with Intel. Knowing that Intel customers used MHz to judge performance, they ramped up their chips to extreme levels. Intel responded with their PIV Northwood core and took the MHz trophy back. AMD got smart and skewed the market by printing a number on the box that wasn't actually a MHz rating, while convincing many consumers that it was.
Clear as mud?
Javascript + Nintendo DSi = DSiCade
The "desk lamp" iMac design hasn't been around for a long while now. Yeah, boy, that flat screen sure is novel; nobody sees the advantages of that...
The G5 model has no tray to catch drool in, even. Slot loading drive, on the side.
"Fundamentalism" isn't about divine morality. It's about human authority.
It has Alivec, using the IBM name: VMX. Same thing though. What the other guy said about it lacking instruction reordering is true though. It also lacks cache locking instructions (that the Xenon has) due to it being a single core. It would be plenty fast enough on its own as long as one optimized for reduced branching though. And, of course, there are those SPEs - which are silly fast as long as you have a parallel single precision FP problem, easily vectorized. Like HD video editing, for example! :)
not to mention the intel GMA900 graphics processor isn't exactly the speediest thing around.
Don't mention it, then. Especially since the iMac Core Duo uses a PCI-Express ATI Radeon X1600.
"Anyone that has ever gotten an idea based on any of my work and done something better with it-good for you."--J.Carmack
I'm afraid that your information is several years out-of-date...
AMD's mobile CPUs are now commonly lower power than even the best of the Pentium-Ms to-date. That's in-addition to being cheaper, and higher performance at the same time.
For example:
mSempron 2800+ (1.6GHz) 25W
mTurion MT-34 (1.8GHz) 25W
mTurion ML-37 (2.0GHz) 35W
mAthlon 64-3000+ (2.0GHz) 35W
vs.
mCel-2.5G 35W
Solo-T1300 (1.66GHz) 37W
Pentium M-780 (2.26GHz) 27W
And that's as fair of a comparison as possible.
Since we've debunked the notebook justification, what alternatives are there?
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
AMD was never really in a position where their chips were the clear MHz winners. Up until the K6/K6-2/K6-3 series, they were always lagging in MHz. The Althon changed that, and for the first time they could out perform the Intel equivalents.
There were brief moments where AMD did have the MHz advantage with the Athlon, but the advantage swapped often between Intel and AMD. Once the P4 was released, that MHz advantage disappeared forever... however the AMD chips could STILL compete in performance, in most cases, even at a lower frequency.
This timeline should clarify this somewhat.
Yonah
Comparable to a Athlon 64 X2 (that's a desktop chip) with way less power draw (both idle and peak load).
Other factors exist too... AMD used to have a reputation for poor QA on the line, and while they seem to have overcome it, hey history is a stinger when you are dealing with companies like Apple.
'' There's not nearly enough evidence to reach a conclusion either way. QuickTime export is one of Altivec's strongest areas, and Xbench scores are notoriously bad at having any relationship to reality. Let's wait and see how they do in real life; perhaps you'll find Apple really does have a clue. ''
One of the XBench tests where the iMac x86 gets "slaughtered" is the "User Interface" test.
It turns out that the iMac x86 runs this test at 67 frames per second. Which is quite consistent with some newer Apple technotes that tell you that screen updates are now coupled with the monitor's refresh rate. If you draw more frames per second then the monitor can display, you are just wasting your time. Seems that the other Macs tested run this test at several hundred frames per second.
Those power figures are from an entire _desktop_ systems, not just the CPUs as you imply. Now, the idle load of a desktop is very different from that of a laptop system --- so those numbers don't support your claim.