Apple Intel Mac prototype specs, performance numbers
updated 10:30 am EDT, Thu June 9, 2005
Intel Mac performance
Apple's Developer Transition Kit features a Pentium 4 660 processor running at 3.6 GHz, according to various online reports. Engineers said the Pentium 4 660 would not be used in any production Mac. Rather, one should look to Intel's roadmap for an idea of what future Macs might include. The Developer machine uses DDR-2 RAM at 533 MHz, SATA-2, and Intel GMA 900 integrated graphics. Production Macs will likely ship with off-the-shelf PC video cards. Manufacturers such as ATI and nVidia must simply provide drivers for the cards. The test machines support 64 bit extensions, but Apple's software does not yet support that technology. The developer units include FireWire 400 and USB 2. USB 2 booting is supported, but FireWire booting is not.
The machines run Microsoft Windows without issue. The Intel Macs use standard Intel chipsets which make running Windows XP possible with just a few driver installations.
The machines also feature a standard Phoenix BIOS rather than Open Firmware. Apple does not specify how to access the system BIOS. However, users at the World Wide Developers conference report gaining access to the BIOS by "mashing keys" at startup. The BIOS is fully functional, featuring the same settings as any standard PC.
Native Performance
Very little information is available on performance of native Mac OS X Intel applications, because few exist. However, those who have used the Developer Transition Kit report performance equal to or exceeding a G5 Mac.
Rosetta Performance
Benchmarks performed on the Developer Transition Kits measure Rosetta performance, not native application performance. Rosetta is Apple's emulation technology for running PowerPC applications on Intel Macs. It emulates a G3 processor and does not support Altivec.
Intel Macs scored between 65 and 70 with Xbench running under Rosetta. A dual-2.5GHz Power Mac G5 scored above 200 on the same test. In the CPU test, where G5 systems score between 100 and 200, the Intel Mac only reached the high teens.
The Intel Mac scored 82 on the Thread test, compared to 225 for the G5 machine. In the Computation Thread test, the Pentium machine scored 110, trailing the G5 by only 45 points.
In both the Lock Contention and Memory tests, the G5 significantly exceeded the performance of the Intel Mac, with scores of 420 to 66, and 378 to 214. The Intel Mac managed to exceed the G5's Stream Memory Test score: 351 to 319. The system memory test, however, is a different story. The G5 beat the Intel machine, scoring a 464 compared to 154.
The Intel Mac scored a 125 on the Interface Test, compared to a 380 for the G5, according to one report. Another report claimed the Intel prototype beat a G5 Mac.
The Intel Mac scored well in both the Quartz graphics and OpenGL graphics tests -- matching or exceeding dual-2.5GHz G5 score.
More information can be found at .











Just keep in mind…
06/09, 11:16am reply
…these are dev boxes and while they are spec'd so, they will most surely not be as such when shipped as commercial units.
These are the G5/Xbox360-like set ups.
Though if i can get all but the logic board at newegg, the would be slick!
T
:dragonflypro:
Senior User
Joined: Sep 2003
Meaningless numbers?
06/09, 11:17am reply
This is definitely useful for knowing how fast Rosetta will feel but it seems a meaningful comparison of the systems will have to wait until the benchmarking application(s) are ported to native MacIntel code.
slipperfrog
Fresh-Faced Recruit
Joined: Nov 2002
But the primary question
06/09, 11:36am reply
is..
Are they snappy?
OtisWild
Junior Member
Joined: Feb 2005
re: snappy
06/09, 11:51am reply
found the sounds of it, they would function fine for about 70% of the Mac market as is. That's scarey. USB2 booting: GOOD, Lack of FW booting is BAD tho. Will Target Disk Mode work on these macs? Do they chime at startup?
ibugv4
Fresh-Faced Recruit
Joined: Jun 2003
Hacked Boxes
06/09, 11:59am reply
Just remember that these are "hacked" boxes designed to allow developers to test their apps, not to do performance testing. The dev box is an intel mother board/Chip set with a bios. Apple develops its own mother boards, which will contain the chip sets to do things like FW booting and FW target mode. We will just have to wait till the production boxes come out have this stuff work.
Apple has 18 months or so to design a world class motherboard/chipset which I am sure they can do. I hope they get rid of the bios, will probably go back to ROM, but that is just a guess.
grossph
Fresh-Faced Recruit
Joined: Apr 2004
what about upgrades?
06/09, 11:59am reply
Will we be able to go out and buy a new faster processor from newegg, for instance, to get our mac up to date once we have bought a new intel mac? Just place the new processor in the slot and away you go. Typically slot form factor sticks around for a while.
olemarc
Fresh-Faced Recruit
Joined: Feb 2000
Re: Meaningless numbers?
06/09, 12:03pm reply
Well, actually the benchmarks aren't much good at all for knowing how fast Rosetta will feel. From what individuals at WWDC have said, Rosetta feels very responsive, in contrast to the impression one would get from the benchmarks.
Since Rosetta is translating instructions and then caching them (as opposed to emulating each instruction), it could be that the benchmarks on their initial run suffered from being translated on-the-fly, and subsequent runs (after the code had been translated and cached) would have very different results. And it's sounding like system API calls -- including Quartz 2D, OpenGL, etc. -- are handled by the processor-native code, so it's really only the CPU-dependent portions of an app that would see any difference, and it's sounding like a pretty safe bet that most major CPU-dependent apps will have universal binaries by the time Intel-based Macs appear for the general public.
Marshall
Fresh-Faced Recruit
Joined: Sep 2000
Fast Quartz and OpenGL
06/09, 12:04pm reply
...the reason why these were fast under Rosetta was because apps that use these frameworks or libraries are using the native (recompiled for x86) ones.
Since the OpenGL and Quartz test tap right into these frameworks, the results will likely yield native speeds. Anything else that gets executed that is contained within the PPC app will have to be emulated. Since Xbench runs all the tests (with the exception of Quartz and OpenGL) with its PPC code, it's only normal that the results will show slow speeds for everything and high speeds for the calls made to outside libraries.
Horsepoo!!!
Banned
Joined: Jun 2003
Marshall beat me to it :P
06/09, 12:05pm reply
:)
Horsepoo!!!
Banned
Joined: Jun 2003
re: video card
06/09, 12:06pm reply
Off-the-shelf PC video cards are mentioned, where they just need new drivers. That sounds like a fantastic boon to Mac gaming. Video cards for the mac will be both diverse *and* cheaper than before.
Aranamac
Junior Member
Joined: Aug 2003