サンフランシスコの空の下から~Kyle's monthly topics~(英語)

Vol.1 新しいMac Pro、iMac.そしてWolfram|Alphaについて思うこと

Most of us feel confident that our computers will continue to process faster, display more, and get smaller for years to come, and along with that we also believe that they will decrease in cost, or at least provide significantly greater performance for around the same price. And we know that those changes have so far been almost fantastic, like the difference between a 1GB hard drive costing USD$1,000 in 1991 and a 4GB USB2 thumb drive costing USD$12 in 2009.

How about the new MacPros? They seem cutting edge because once again it seems that Apple has access to the latest and very greatest from Intel, the newest Xeon "Nehalem" processors. The Nehalem architecture, as expected, has better performance than Penryn, something like a 10-25% improvement single-threaded and 20-100% better for multi-threaded, but these estimations are rather subjective. Initial impressions indicate that they will be super fast, but don't know if these new machines offer greater value for money over the older models. Call me skeptical, but I must wait for some solid benchmarks of "real world" applications. Apple lists some on http://www.apple.com/macpro/performance.htm but honestly, can we really trust a marketing page? No, I don't think so either.

About being good value, it does appear that at least the entry level model is a bit more expensive than we might have expected. The base quad-core offers either a 2.66GHz or (for USD$500 more) a 2.93GHz Quad-Core Intel Xeon. The online retailer NewEgg.com, popular amongst US hardware enthusiasts, offers the Intel Core i7 920 Nehalem 2.66GHz quad-core processor for just $288.99. It sells an i7 940 Nehalem 2.93GHz for USD$559.99. These seem lower in price than CPUs were when Apple introduced the early 2008 MacPros, but the overall price of the systems seems higher. Does $2,499.00 for one 2.66GHz CPU, 3GB RAM (three sticks), 640GB HD, 18x double-layer DVD burner, and 512MB NVIDIA GeForce GT 120 seem like a good price to you? Again, I feel that I must wait a bit until people tear the machines apart and count up the cost of the components to determine what Apple's profit margins are before commenting, but it seems on the high side. Businesses, though, even in times of economic hardship like now, will spend a bit more money if they can be assured that the investment will repay itself in a reasonable time.

One nice thing: Apple is at least finally shipping a somewhat reasonable video card as an option, compared to what tended to be rather weary mid-level models in previous machines. For example, the older MacPro typically had a Radeon HD 2600 XT, according to Apple's performance comparison page, but that card was mid-level even back in the summer of 2007 when it was introduced. The new MacPros come with either one or more NVIDIA GeForce GT 120 cards (with 512MB of GDDR3) or, even bettter, a single ATI Radeon HD 4870 512MB for an additional USD$200. Not cutting edge, not a GeForce GTX 295 (not even a GTS 250), but not horrible. I do wish that something faster from NVIDIA had been a build-to-order option though. Also, there is no workstation graphics option yet. Specialized, but required by some people.

Why, though, does Apple still not include a Wi-Fi adapter in every MacPro? Why does this new model support only one monitor out of the box, and Apple requires you to buy an adapter to add a second monitor? I guess Apple assumes that people will attach to one or two wired networks and will buy the USD$50 wireless card if they really want one, and that people will pay either USD$29 if they need to attach a VGA monitor or USD$99 if they need Dual-Link DVI. I can understand how some people feel insulted that they are already buying a premium machine but aren't getting some common functionality unless they pay a few more dollars. Even the Mac mini includes a Mini-DVI to DVI Adapter. Well, at least the keyboard is a full-size model, unlike the tiny one that ship standard with the iMac. I still use my Happy Hacking keyboard from time to time, but my Das Keyboard has supplanted it.

Speaking of iMacs, it certainly looks like the new iMac and Mac mini offer good consumer value, when looking purely at cost vs. features. Certainly, you can buy cheaper and more powerful PCs (I have) and you can stuff many hard drives and lots of memory into a MacPro (but few of us really need to). Yes, maybe ripping DVDs is a lot faster with the Nehalem machines, but the iMac and Mac mini seem to be very good "for the rest of us" machines, assuming you don't need portability. If you already have a monitor, the Mac mini also seems to be a very good value, with NVIDIA 9400M graphics. Still shared memory, but better. I guess USD$200 more for 2GB memory and a 320GB hard drive is only moderately overpriced, considering Apple's typical markups, but I'd probably want to upgrade it myself.

One reason to get a machine with a very modern processor for me would be to maximize the responsiveness of the various Windows virtualization applications. I would like to try Sun's xVM VirtualBox on a Mac, just to compare it to, for example, Parallels Desktop. Another interesting virtualization app is CrossOver Mac, a "commercialized" version of Wine. As you probably know, Wine is a re-implementation of the Win32 API, and a supported application runs as if it were native. Some people might complain about paying for free software (I've complained about the GIMP project not making official binaries available), but for me it depends on whether I am getting good value. This commercial version of Wine costs either USD$39.95 or USD$69.95 depending on the level of support you want. I don't recommend running Windows games (might as well dual boot into Windows if you really, really want that) but it lets you run certain software that you might ordinarily not be able to. They have a screenshot of Microsoft Visio, for example, and even list Ichitaro 2006 amongst its "supported" applications. Of course, their definition of "supported" covers a variety of different scenarios--in the case of Ichitaro 2006, it gets a "Bronze" rating for compatibility, meaning "installs, runs, and can accomplish some portion of its fundamental mission." Not entirely comforting, but perhaps not bad. And some portion of a buyer's money is going to support the Wine project, too. See http://www.codeweavers.com/products/cxmac/ for more information. (You might be wondering why it is called Wine. In the same spirit as GNU's "GNU's Not Unix" name, Wine stands for "Wine Is Not an Emulator." What can I say? Recursive acronyms are so common in the geek world.)

But when it comes right down to it, I am becoming more and more doubtful of the future of our beloved computers, be they Macs or Windows or Linux. It seems to me that Google shows the future of computing, where data and applications live up in the "cloud" and all our OSes and hardware are simply base platforms which give us a web browser. Oh sure, this may not come about for many years, but the building blocks are being put into place now. Yes, gaming technology will continue to push along GPUs and CPUs and RAM and hard drive requirements (I just downloaded a demo version of Unreal Tournament III: 8GB of disk space. Eight! Crazy.) and we don't have nearly the networking infrastructure to deliver that kind of material or run apps anywhere except locally, but when we have terabyte mini-thumb drives we'll be able to carry them around and plug them into any network-connected box which then transforms itself from a dumb terminal into our personal desktop, accessing remote apps and running games and so forth as if it were our machine at home. We will want something beyond Google then, which, despite the accuracy of its search results, is still not much more than just a huge body of indexed data intermixed with related advertising. What might go beyond that? Perhaps Wolfram|Alpha.

Stephen Wolfram sounds like one of those "scary smart" people, someone who can conceptualize an essential theme and then capitalize on it somehow. His Mathematica application has certainly been a significant tool for scientists and mathematicians for many years, and he claims that his discoveries described in his book "A New Kind of Science" means that it "suddenly becomes possible to make progress on a remarkable range of fundamental issues that have never successfully been addressed by any of the existing sciences before." From someone else I would be quite skeptical, but it appears that Wolfram and his team have created something that goes far beyond Google, by using Mathematica as "a symbolic language to represent anything-as well as the algorithmic power to do any kind of computation," and using his New Kind of Science as "a paradigm for understanding how all sorts of complexity could arise from simple rules." It results, he says, in "a computational knowledge engine" whose front end consists of a single text entry field yet "gives access to a huge system, with trillions of pieces of curated data and millions of lines of algorithms."

The idea is to permit anyone to type in a question and have Wolfram|Alpha return an answer. Not an answer based on frequency of words or some unknown "page ranking" system, but a system that draws upon a body of knowledge to actually come up with an answer rather than produce closely matching text. Questions need to be fact based, like "How many states does the Mississippi River pass through?" or "When was the Magna Carta signed?" and not questions of judgment, but even that will be quite an achievement. Sure, searching for those strings in Google could well bring you to the right answer, but what is most interesting is the likelihood that it will be able to produce answers to questions that we cannot imagine now and that would not be answerable via Google.

Wolfram|Alpha will be launching in May, and though I imagine it will be English only for now, Wolfram and company are working on capturing the "shorthand notations that people in every possible field use" and I have no doubt that that will eventually include all known languages. Peek into the future at www.wolframalpha.com/.