RUMOR: Apple prepping monster eight-core Xeon ‘Clovertown’ Mac Pro

“Apple Computer is prepping a lavish new version of the Mac Pro that will boast nearly twice the brawn of existing models and form the centerpiece of the company’s high-performance professional desktop line,” Kasper Jade reports for AppleInsider.

“People familiar with the Mac maker’s plans say it plans to drop jaws and strike awe with a new king of speed, a super-charged Mac Pro featuring a total of eight cores of processing power. The systems, which resemble the quad-core Mac Pro externally, will house two of Intel’s forthcoming quad-core Xeon 5300 series ‘Clovertown’ chips inside is chassis, those people say,” Jade reports.

Jade reports, “Those familiar with the company’s plans have indicated an introduction could take place any time after mid-Nov.”

Full article here.

[Thanks to MacDailyNews Reader “LinuxGuy and Mac Prodigal Son” for the heads up.]

Related MacDailyNews articles:
AnandTech upgrades and tests Octo-Core ‘Clovertown’ Apple Mac Pro – September 13, 2006
Intel reveals first glimpse of quad-core ‘Clovertown’ chip coming later this year – February 11, 2006

32 Comments

  1. DLMeyer,

    I got my iMac core2duo up to 98% capacity by telling the bioinformatics program I was running that I had 4 processors. Even then it took about 15 minutes to compare all proteins in a bacterium to all proteins in a yeast. Bioinformatics is an inherently parallel process, so it is a natural fit with multi-core processors. I would imagine that other things like video encoding would likewise work well in parallel. You are correct, however, that typical home activities such as Email, web surfing and word processing will not benefit. They no longer benefit from any of the speed increases any more, because those activities are limited by the user rather than the machine. This upcoming quad machine is obviously not designed for the casual home user.

    By the way, I tried to find a simple explanation as to why I had to refer to my machine as having 4 processors for it to run at full capacity. The name core2duo implies something 2×2 is built into the processor, but I would appreciate a simple explanation if anyone has one.

    Ta

  2. By the way, I tried to find a simple explanation as to why I had to refer to my machine as having 4 processors for it to run at full capacity. The name core2duo implies something 2×2 is built into the processor, but I would appreciate a simple explanation if anyone has one.

    How much load did it cause when indicating only two CPUs/cores? If it was much below 200% (or 100% of total load), the application may simply have been buggy.

    Your dual core does indeed offer 200% performance of a single CPU or core (minus overhead as always) and will fully profit from multi-threaded applications.

  3. By the way: “Core 2” indicates version 2 and “Duo” indicates that the physical CPU contains two execution cores which can work in parallel.

    So still “just” two cores in total, but two very powerful ones! ” width=”19″ height=”19″ alt=”grin” style=”border:0;” />

  4. The other piece of the puzzle was talked about awhile back and had to do with a much more efficient thread farming achitecture in Leopard. According to that article Leopard will be able to make very efficient use of many cpu cores even when the applications don’t have any multi core optimizations. This is in contrast to Tiger which makes very poor use of more than 2 cores unless the app itself is hand optimized to do it’s own thread farming accross many cores.

    This is all part of Apples strategy to force HP, Dell and other PC builders to cut their high end profit margins. Without the big margins on the high end kit, most of these guys will have to raise prices on the low end “loss leader” systems to stay in the black.

    Because of the iPod, Intel Macs and a great OS Apple can really start playing hardball with the other PC builders.

  5. HParker, I’m not sure if they are currently designed to take good advantage of Quad Core or if they are merely Quad-enabled, but we can expect the Apple Pro apps to lead the way here. Hardly half a dozen of them, though, even if you count Express and Pro versions individually.

    ping, wrong. BOINC has a “manager” that runs on one core and spawns individual core-bound processes. That’s a “cheat”. They are individuals, each with much of the same overhead. They are not multiple threads working towards the same goal. Sure, there is some savings in that there is but one task manager, but the jobs it spawns are not multi-threading and no linked parallel processing is involved – each task is self-contained and need bear no relation to any other. When you have a proper multi-core, multi-threading job running, it splits off bits of itself that are part of the whole but that can be resolved separately.

    Let’s use iTunes as a descriptive example. You want to rip a CD while playing it. iTunes already has a library manager thread running, and an inactive player thread. It starts a reader thread, which starts and feeds a compression thread, which feeds the library manager thread which activates the player thread. The reader thread quits once it has cached the contents of the disk, the compression thread quits when it has compressed and written the last song, and the music plays on.

    The BOINC software reads in a work unit from an off-site manager and assigns it to a core. That core processes that work unit, spawning no threads to do portions of the work, and reports back for a new assignment when done. While I have medical research processing on both CPUs, I could have that on one and SETI on the other – neither spawning to other cores. Or I could have four projects sharing time on my two CPUs, each taking a configurable amount of BOINC’s ‘share’. While this is a fairly efficient process and makes good use of multiple cores, it’s still multiple single-core processes running on a multi-core system. None of the individual programs running are multi-core in and of themselves.

  6. Paul, that’s strange! When I run Folding@Home – a different core-bound medical research project – it takes every available cycle from the core/CPU it is running on. I have to run a second copy to get it to use the other CPU, then it takes anything available there. As it runs with a +20 priority, this means that anything else gets ‘first dibs’ on the next CPU cycle, but I’m running at over 99% for each core 24/7. That may well be “over 99.99%”, but my stats are not that precise. Anyway, ping explained the naming of your CPU and I can’t explain why you need to lie in your configuration to get it to do the right thing.

    Dan, excellent points, all. Well, I’m not sure how much the iPod figures in the strategy, but Dell and HP are certainly going to have to re-configure their pricing strategy now that Apple is obviously in the same workstation market with them. Apple’s margins are high for the whole industry, but not as high as Dell’s or HP’s in the workstation niche. Both will now have to reduce their losses in the commodity end, making both the iMac and the Mac mini more attractive choices for more people. That or suffer lost sales in the high end, and lost profit there.

  7. ping, wrong. BOINC has a “manager” that runs on one core and spawns individual core-bound processes. That’s a “cheat”.

    I can’t see a real point there.

    You may have personal preferences regarding details of the thread and process organisation of multi-threaded applications (and yes, even multiple threads spanning multiple processes ultimately working for one application qualify as such), but the essential point is whether or not applications make effective use of multiple cores in practice.

    And several applications do in fact do that already – unsurprisingly they’re the ones which actually have a chance of saturating a single core.

  8. How about an ultra-black case that has the ability to suck light into its very being, whilst warping space and time so that you actually finish the computing task before you started it?

    From what I hear, that’s one of the few ways you’ll be able to get hold of a retail copy of Vista before 2008.

  9. Dan is correct — Tiger (and Panther) handle two CPU cores quite well — OSX OS is able to traffic-cop multiple applications, including single-threaded applications, to the “next available” CPU, aiding in reducing the time any particular application request takes to process by managing any CPU request bottlenecks.

    I own a 4+ year old DP G4 867 PowerMac with 2GB of RAM and 4 internal 250 GB 7200 ATA hard drives — per se, it is nearly as fast as a friend’s DP G5 2 GHz, 5 GB RAM PowerMac, when running multiple open applications that are not terribly CPU intensive — browsers, text editors, RSS readers, email clients, Photoshop and Adobe Acrobat Pro — all open simultaneously, humming along nicely, context switching seamlessly, cutting/copying and pasting with impunity — DP G4 867 is very snappy, when compared to DP G5 — OSX’s ability to effectively utilize BOTH CPUs is the key, and allows the older DP G4 hardware to remain competitive in normal usages.

    However, the DP G5 and the newer Intel Macs are able to run MUCH faster than the DP G4 when rendering video or audio — just a fact of raw CPU horsepower.

    Have another friend who has a Quad G5 PowerMac, who does FinalCut Pro stuff for a living — OSX and FCP make VERY GOOD USE of all four G5 CPUs — the difference in speed between a DP G5 and the Quad G5 when rendering the same digital files under FCP is significant — nearly double the speed.

    I expect that Leopard will be tuned to take better advantage of multiple CPUs than Tiger/Panther, even when managing single-threaded applications. Also, I believe that Apple is significantly upgrading their software toolsets, and particularly the CarbonLib capabilities, enabling application developers who still depend heavily upon Carbon software technologies to upgrade their Carbon-based apps with Apple’s newer SW tools and libraries, enabling these older Carbon apps to acquire some of the same tricky features and capabilities as the Cocoa-developed apps.

    Niffy

Reader Feedback

This site uses Akismet to reduce spam. Learn how your comment data is processed.