Macworld, naturally, spends a lot of time focused intently on Apple hardware while I get distracted by beatboxing parrots and modular synthesizers built out of yarn and rubber bands. They have an excellent write-up of the significance of the MacBook Pro Santa Rosa upgrades, with comments on their benchmarks of the equivalent refreshed MacBooks:

MacBook Pro knows the way to Santa Rosa

One thing I was a little unclear on in my previous story is what matters in Santa Rosa, Intel’s latest architecture platform. (They didn’t call it Core 3 Duo, but then, consistent branding and Intel don’t generally go together.) As with Core 2 Duo over Core Duo, we’re getting incremental performance enhancements relative to the previous generation. Each step is relatively small, but they start to add up — hence, Apple quotes 50% gains over the original Core Duo. (And that’s why they dumped PowerPC, which in the mobile space was starting to practically paddle backwards.)

The key differences as far as raw performance: faster front-side bus (800MHz instead of 667), which for audio is a big deal, faster clock speeds on the models themselves at the same price, and fast RAM, plus a faster GPU for GPU-related tasks. (And, um, any day now we’ll start to see audio on the GPU — it’s tough to program, and GPUs are only now becoming the norm, and CPU cycles are getting cheaper, but it will happen.)

Also, none of this was meant to say “eBay your MacBook Pro.” PowerBook G4, maybe, but the first-gen MacBook Pro is still a terrific audio machine, with a GPU that’s no slouch. My main laptop right now is a first-gen MacBook (no Pro), and it blazes through everything I throw at it.

  • Claudio

    Core Duo to Core 2 Duo was an incremental improvement? Considering that the Core 2 Duo is a 64-bit chip, I'd say that's quite a major improvement especially once Apple brings Leopard out. Granted the available music apps won't notice the difference, but the fact that newer versions of these apps will be able to address a lot more memory than their older 32-bit counterparts is a big thing. It might seem incremental now, but as apps evolve to 64-bits the Core 2 Duo begins to show its prowess over the 32-bit Core Duo.

    Still, it's definitely evident that the new 3rd-gen MacBook Pros have improved upon an already improved 2nd-gen MacBook Pro design. I can't wait until this fall when I can afford one. :)

  • nissaneuphoria

    The FSB on the CPU may be up, but the Ram is still stuck at 677. And I don't know too much about computers, but why do P4's have such higher multipliers (and smaller FSB's) than core duos?

  • http://www.createdigitalmusic.com Peter Kirn

    Sorry, you're right. Incremental in terms of 32-bit performance — and there were certainly improvements comparing 32-bit to 32-bit. Of course, remember that the enormous leap from G4 to Intel Centrino sort of dwarfs everything else for the Mac community (64-bit included), but you are certainly right.

  • http://www.andrewswihart.net Andrew Swihart

    Peter, the Core 2 Duo was a revolutionary step beyond the Core Duo, as it was beyond the AMD X2 processor line if you would allow such a comparison. Meanwhile, the CPU is largely untouched with Santa Rosa. It is all about chipset enhancements, things like increased FSB as you mentioned, a 4GB RAM limit, etc., etc. They didn't call it Core 3 Duo because as I said in the other thread, this is an identical CPU core architecture, with minor improvements.

    What I was saying to begin with, which you disagreed with, is that Santa Rosa is not a huge step forward like Core2 was from Core. If you can show me some benchmarks showing otherwise I'd like to see them.

  • http://www.andrewswihart.net Andrew Swihart

    A source of the facts I said:
    http://www.anandtech.com/showdoc.aspx?i=2985&…

    And you should correct your post above. You say this is a new architecture, but it's actually a new "platform", consisting of a CPU and chipset. For this new "platform", the CPU is the same old Core 2 Duo Merom we've had before with the same architecture, just with extended FSB speeds.

    Also, one of the significant, but optional, features of the platform is TurboMemory, which apparently Apple is holding until later or completely bypassing

  • http://www.createdigitalmusic.com Peter Kirn

    Andrew, I agree; you're right. Platform, not architecture. And I'm not trying to draw any larger conclusions about Santa Rosa; maybe I'm inadvertently conflating issues. But on the MacBook Pro lineup itself, I think you have to look at two important issues here: one is, this pushes RAM upgrade capabilities within the realm of what's needed by higher-end samplers (hence the emphasis on this issue), two, the combination of faster FSB with higher CPU clock speeds (not because of Santa Rosa per se but because the new Intel CPUs offer enhanced clock) is going to theoretically add up to improved performance. The proof will be to test what Apple was testing here.

    We're just talking 32-bit to 32-bit benchmarks here, and revolutionary as the architectural changes may be on Core 2 Duo, we still can't look on the Mac at 64-bit performance (and arguably can't on Vista, either, just because the big device compatibility picture isn't there). What you're saying is absolutely right, but I'm just talking about the end-user perspective.

    Also, I'm taking into account the inclusion of the NVIDIA graphics chip, which is a big step forward entirely independent of what Intel is doing. (in fact, by contrast Intel is still, naturally, pushing their integrated graphics architecture, which is still at this point inferior to a dedicated chip)

    Just to go back to Apple's original numbers, they said they had a 39% improvement over the Core Duo with Core 2 Duo (2.33 GHz then compared to 2.16), now they're saying 55% or so 2.4 Core 2 Duo over 2.16 Core Duo, rather than whatever smaller incremental improvement you'd get from Core 2 Duo to newer Core 2 Duo MBP. That's why Macworld said they were interested in doing the tests of the two most recent generations, as I would be.

    So, yes, Santa Rosa is not the step forward from Core2 that Core2 was from Core … what the exact proportion is, you'd have to test. I haven't actually seen any useful real-world benchmarks there yet; I expect Macworld is on the case.

  • http://www.createdigitalmusic.com Peter Kirn

    We haven't touched upon these as much because — well, they're relatively minor issues — but the two other enhancements:

    1. Dynamic threading: One core gets overclocked for single-thread operations … not sure how often this happens, or what the actual performance gain is. (and that's on a per-instruction basis? Here's where I get in over my head.)

    2. TurboMemory … do you mean Robson? I guess the question there is whether Apple is adding OS-level support for this Flash memory stuff. Vista has SuperFetch / ReadyBoost / ReadyDrive, of course, but then, while Intel has some impressive numbers, I've yet to see a compelling real-world case.

    For audio (or video), this stuff to me is less interesting, as you really need a hard disk to get at that data … app loading, etc., is less of a high-priority issue. Maybe I'm missing something there? I've mostly heard about this via Microsoft, and it didn't seem to me a compelling case for overall system performance, once your applications are loaded. :) HDD access there for me is not the big bottleneck … more access to things like video files and large audio samples, etc.

    You may know more about these issues than I do; I'm just hitting on what I, erm, think I know, at least, and what I think of it.

  • http://www.andrewswihart.net Andrew Swihart

    Yeah, Robson is what I'm talking about here: http://www.intel.com/design/flash/nand/turbomemor

    It isn't a big deal to me either, but it was one of the more notable features of the platform, completely new, while we've seen 4GB before. Although I agree the 4GB thing and the higher FSB could mean potential gains for some people who are working on the edge of their systems currently. Myself, I barely touch the CPU power of my X2 and 2GB RAM.

    I just read Anandtech most of the time for this kind of info, they seem to cover everything pretty clearly. I read about Santa Rosa like a year ago, and the only thing that stood out to me at the time was the TurboMemory thing, but even that seems like it's not so important now, and it's not included in the MBP update anyway.

    And lastly, for Apple's benchmarks, I wouldn't trust those for telling anything "real-world", or even meaningful. They were saying their G5's were so much faster than Intel's chips, the they switched to Intel chips and did some other benchmark showing how much better they were than the G5's. I forget exactly what they did and what they compared, but it was pure marketing nonsense. Don't even pay attention to those quotes, except for humor's sake.

  • http://www.createdigitalmusic.com Peter Kirn

    Yeah, I was one of the people who pointed out their original PowerPC Logic benchmark made no sense at the time. I actually found someone from Emagic (this was just after the acquisition) to find out what the heck they were doing. Remember the keynote with audio actually dropping out on the PC? It turns out there were lots of invisible reverbs hiding in the background … far more than you'd ever use. So it wasn't actually false, it was just misleading. Thankfully, they haven't repeated that. And they actually said that they heard our frustration, because they did frustrate the audio people who understood the test didn't make sense.

    In their defense, producing benchmarks of this kind is difficult. For Logic, they just stack Platinum reverbs together and see when they can't any more. Honestly, that's as decent a raw power test as any, and it's easy enough to do. It's just not terribly realistic, because it's sort of one-dimensional, and it's not something you'd do in the real world. (Well, okay, *I* sometimes make projects that involve stacking absurd number of reverbs, but I'm not very realistic, either.)

    It's not a meaningless number — it actually means a lot. It just needs to be balanced against other tests. And, in the end, it's the soft, qualitative experience that matters more than the numbers anyway, because that's what you're actually buying.

    Final Cut Pro turns out to be easier — then they just did a DV encode. Again, it's just one test, but it tells you something — and, in the real world, you might be staring at a progress bar in that case.

    What I found with their Core Duo benchmarks was that, while they were a little exaggerated because they were one-dimensional in that way, they were reflected elsewhere on some level.

    I'll be honest, *personally*, it's the NVIDIA chip that has me excited, and I was waiting for that card to show up in a computer I wanted, and in a machine that runs both Mac and Windows really does fit the bill. There are some cool things I want to try out with NVIDIA's shader extensions.

    And totally, totally trust Anandtech over me. This requires someone who eats and breathes this stuff to really understand the CPU stuff; just reading Intel's white papers can be insufficient. I'll get back to talking about parrots and Monomes. I do have some real machines I can test, and I can comment on that sort of real-world performance. As far as truly understanding the architectural stuff, that's definitely out of my league.

    The other reality is, there's a certain, practical CPU level that's useful for audio. The high-end G5s, AMD X2s, Core Duos for me all make the cut. It's the G4s and, to a lesser extent, older Centrinos where I've found people are actually unable to do the sort of basic music stuff they want to do — talking the CPU only.

  • morgancck

    very helpful advice, thx a lot.

    comparison ipod vs zune
    http://www.freedownloadscenter.com/Multimedia_and

  • Cees Mutsaers

    What will be the next update for the MBP at the end of the year : LED display for the 17", Blue ray, Leopard ? what else?

  • Santa Rosa Owner

    Has anyone who purchased a Santa Rosa experienced this:

    http://www.ableton.com/forum/viewtopic.php?t=7973