Marco.org • About ▾

I’m : a programmer, writer, podcaster, geek, and coffee enthusiast.

Scaling down the Mac Pro

The Mac Pro is a beast.

The case is huge. It hasn’t seen an external redesign since 2003 or a major internal redesign since 2006. It’s nearly 50 pounds loaded. It’s only updated whenever there’s a new generation of Xeon CPUs from Intel, which only happens about every 18 months. This can sometimes make it seem painfully out of date: for instance, today’s Mac Pro still doesn’t have Thunderbolt, 9 months after the port’s introduction, and the base $2499 configuration, which is slower than every 15” MacBook Pro today, ships with only 3 GB of RAM and a 1 TB hard drive.

And it’s very expensive. As Intel has raised Xeon prices and Apple has widened their margins since the line’s introduction in 2006, Mac Pros have become worse deals. A worthwhile configuration usually costs over $3500, often even $5000, and that’s assuming you already have a monitor and you don’t mind buying third-party RAM and hard drives.1

Apple doesn’t sell a lot of Mac Pros. Only 26% of Macs sold today are desktops, and that includes the very popular iMac and the Mac Mini. I don’t think Apple ever breaks out the numbers by product line, but I’d guess that fewer than 5% of Macs sold are Mac Pros. So it’s not very surprising that the Mac Pro may be discontinued soon.

What could Apple change about the Mac Pro to make it attract more buyers?

Smaller?

Probably not significantly.

The top third is optical drive bays and the power supply. They can drop down to just one optical bay, and could even use a slim laptop drive (or omit optical drives entirely), but the huge power supply still needs to go somewhere. The power supply can’t be meaningfully downsized: since Apple uses the same power supply in every Mac Pro, it needs enough capacity to power the fastest CPUs with all RAM, disk, and expansion slots filled with the highest-draw options.

The hard drives are arranged very efficiently. There’s not a lot of savings to be had if the case still needs to hold four 3.5” hard drives. If they got rid of the optical bays, they could put the hard drives in their space, but it wouldn’t be a big reduction overall.

The midsection must be left mostly open so full-length PCI-Express cards can fit.

The bottom third needs to hold the huge Xeon heatsinks and the 8 RAM slots (in dual-CPU models).

If the case is made any narrower, the fans need to be made smaller as well, which would require them to spin faster and produce more noise.

So any meaningful size reduction will make the Mac Pro less desirable to a significant portion of people who actually do buy it.

Cheaper?

Given what the Mac Pro is — a high-end Xeon workstation — it’s actually priced competitively with PC equivalents most of the time.

I just went to Dell’s website and configured a workstation as close as I could get2 to the dual-2.66 GHz Mac Pro with AppleCare:

At this level, and considering that the comparison is only approximate (since it’s impossible to configure exactly matching systems), that difference is negligible. The problem isn’t that Apple’s overcharging for the Mac Pro. The Mac Pro’s specs are just far above what most people want to pay for an expandable tower PC.

The only way to significantly reduce its cost is, again, to remove the reasons why many of its customers buy it:

Consumer-class CPUs?

Apple could make a smaller, much cheaper desktop Mac by abandoning the Xeon CPUs and just using Intel’s mainstream desktop CPUs, such as the quad-core 3.4 GHz Core i7-2600K. They already offer this with the iMac. But if you’re reading about the Mac Pro, you probably want more internal expansion than the iMac will accomodate.

Going to desktop-class CPUs has two big downsides for Mac Pro buyers:

Since so many Mac Pro buyers choose it specifically to have the most CPU power or the highest RAM capacities available, this might not be an acceptable tradeoff to many of them.

As a point of comparison, almost all desktop-class motherboards today are limited to 16–24 GB of RAM, and the top-end 3.4 GHz Core i7 CPU (available already in the iMac) gets a 64-bit GeekBench score of 12,575. The Mac Pro released more than a year ago maxes out fairly affordably at 48–96 GB, and the top-end dual-2.93 GHz Xeons score a 24,159 in Geekbench. And it’s probably going to be updated to even faster CPUs in a few months.

If a 16 GB RAM ceiling and about 12,000 Geekbench points are enough for someone, Apple probably already serves their needs with an iMac or MacBook Pro. Even the Mac Mini maxes out at 16 GB and just under a score of 10,000 today.

And if they need more power or RAM capacity, only the Xeon platform can deliver it.

Less expansion?

Thunderbolt is, essentially, PCI-Express over a cable. It’s fast enough for hard-drive enclosures that let their drives run as fast as if they were internal. It’s even fast enough for some external GPUs. Many roles of the Mac Pro’s PCI-Express slots could be replaced with Thunderbolt peripherals.3

I wonder how many Mac Pro owners use all four hard-drive bays. I bet the number has decreased over time as individual hard drives have grown so large and SSDs have become so fast.4

If card slots and internal drive bays aren’t as necessary as they used to be for a Mac Pro customer, it’s much more likely that a Mac Mini, iMac, or even a MacBook Pro might solve their needs sufficiently for less money.

But for the buyers who do need card slots, internal 3.5” drive bays, or some edge-case Mac Pro capabilities such as the ability to drive lots of monitors, no other Mac can fill the Mac Pro’s role. They’re the ones still buying them, after all — any reduction in expansion potential would leave them stranded.

A matter of time

It’s impossible to significantly change the Mac Pro without removing most of its need to exist.

But I think it’s clear, especially looking at Thunderbolt’s development recently, that Apple is in the middle of a transition away from needing the Mac Pro.

Fewer customers will choose Mac Pros as time goes on. Once that level drops below Apple’s threshold for viability or needing to care, the line will be discontinued.

I bet that time will be about two years from now: enough time for Apple to release one more generation with Thunderbolt and the new Sandy Bridge-based Xeon E5 CPUs in early 2012, giving the Mac Pro a full lifecycle to become even more irrelevant before they’re quietly removed from sale.

A few power users will complain, but most won’t care: by that time, most former Mac Pro customers will have already switched away.


  1. Configuring the single-socket 2010 Mac Pro with 16 GB of RAM, a reasonable amount for power users, costs $775 from Apple or $145 from OWC

  2. I configured the Dell Precision T5500 workstation with Dual Xeon X5650 (2.66 GHz) CPUs, 6 GB (6x1) RAM, 3 Year ProSupport, Mini-Tower with 1394 Card, 256MB ATI FireMV 2260, 1 TB 7200 RPM hard drive, 16X DVD+/-RW, and Internal Chassis Speaker ($5). I couldn’t find options for Wi-Fi, Bluetooth, FireWire 800, or optical audio ports, but their warranty includes on-site service, so there are a few mismatches but I think this is a good approximation. 

  3. Very few Thunderbolt peripherals exist yet. One interesting theory to consider: if Apple did discontinue the Mac Pro, it might force peripheral makers to adopt Thunderbolt more aggressively for devices such as professional video-capture cards. This may be more valuable to Apple than the Mac Pro. 

  4. If so, Apple probably doesn’t sell many RAID cards anymore. Maybe this is also why they discontinued the Xserve RAID in 2008. Also worth noting: Fibre Channel is no longer exclusive to the Mac Pro. 

Consumer Reports

Consumer Reports announced today that they approve of the iPhone 4S, after famously refusing to recommend the iPhone 4 because of that antenna issue that turned out to affect almost none of the millions of people who bought it.

But the iPhone 4S was defeated by three Android phones in CR’s scoring criteria:

These pluses were not enough, however, to allow the iPhone 4S to outscore the best new Android-based phones in our Ratings. Those top scorers included the Samsung Galaxy S II phones, the Motorola Droid Bionic, and several other phones that boast larger displays than the iPhone 4S and run on faster 4G networks. […]

Other phones that topped the iPhone 4S include the LG Thrill ($100 on AT&T), which has the ability to capture stills and videos in 3D, as well as display them on its 4.3-inch 3D display, and the Motorola Droid Bionic ($300 on Verizon), which also has a superb 4.3-inch, high-resolution (540 x 960) display, with excellent keypad readability under most lighting conditions, even in bright light.

I’m looking at their full test results (I’ve been a CR website subscriber for six years), and I’m really not confident in the metrics and priorities that they seem to be using. Even some of the measurements seem suspicious to me:

Consumer Reports is a good place to get an overview of objective specs and measurements for a product category. I’ve often consulted their ratings to guide many purchases, mostly appliances, without any major regrets.

For their ratings to be useful to my purchase, their priorities and criteria need to approximately match mine. This is easy for most of the products they review: most people want a dishwasher to be able to quietly, effectively, and reliably wash dishes. A dishwasher that’s quieter is objectively better than a louder one. An air conditioner that uses less energy for the same cooling is better than a less efficient model. You can assign numbers and scores to factors like these.

Smartphones have too many subjective criteria, and even the measurable stats don’t always yield a definite answer on what’s better. If you want a huge screen, you’ll get a huge phone, so is a larger screen size a good thing or not? Fast 4G network access kills battery life, so is 4G a good feature for you? Do you want the best normal camera, or a lower-quality 3D camera? Do you want any particular apps or games that are only available on one platform? Do you need a kickstand? (Do they still make those?) These all depend on your priorities.

A product as complex and multifaceted as a modern smartphone is beyond Consumer Reports’ ability to rate in a way that’s useful to most buyers.

The Asus MacBook Air, sort of

Casey Johnston at Ars Technica reviews the Asus Zenbook, one of Intel’s “ultrabooks”:

While we found the performance of the Zenbook to be unfaltering, it does fall short in a few areas like screen and sound quality (the 11-inch Core i7 we tested is also cheaper than the MacBook Air). But in one key respect it absolutely fall down hard: the trackpad is fickle and barely functional, to the point that using the Zenbook as a primary traveling work machine caused us a good deal of frustration.

I think we’re going to look back and laugh at ultrabooks in a few years. It’s a laptop category arranged and marketed by Intel that basically means “PCs that copy the MacBook Air as much as possible”, and Intel needed to start a $300 million fund to help PC manufacturers accomplish this.

The Asus Zenbook is one of the first “ultrabooks”:

Looks familiar, right?

The copying runs deep with the Zenbook. As shameless as the external styling is, here’s its internal layout:


Image from AnandTech

Here’s the 11” MacBook Air:


Image from iFixit

It’s sad, really, that the state-of-the-art in the PC world is attempting to copy Apple. Why isn’t Asus trying to blow the MacBook Air out of the water with something radically better?

I don’t know how people who just copy others’ work can take pride in what they do.

Maybe they don’t. Because despite being handed Apple’s design to copy and presumably having a lot of help from Intel, Asus still didn’t do it right. It’s close in many ways, but it fails on so many details (and critical functionality, like the trackpad), according to Johnston.

More than anything, this shows how competitive Apple has become in the PC market: they have such great manufacturing expertise, component deals, and supply chain management that other PC makers can barely remain competitive in the market segments that Apple competes in, even with outside help and possible subsidies.

It looks like people who want to run Windows on something a lot like a MacBook Air are better served by just running Windows on a MacBook Air.

The relationship between Readability and Instapaper

I know it’s bad form to mention your competitors, but I’ve been asked about Readability’s announcement enough today that it’s more of a charade not to talk about it, especially since a lot of people are under the wrong impression.

Here’s my relationship with Readability:

The Readability founders came to me in 2010, shared their idea of paying publishers for what people read with their text-view bookmarklet, and wanted to explore whether we could work together. We had very different priorities at the time: they wanted their service to focus on the publisher-payment system, and I wanted to focus on my iPhone and iPad apps.

We figured out a way to work together: I’d build a white-label version of the Instapaper app that worked with the Readability service instead of mine, with no source-code sharing, and I’d get a royalty for each copy sold. That way, I could keep my efforts focused on what I care most about, the iOS app, and they could have a full-featured iOS reading app from day one without having to build it themselves. I would also advise the company, promote the service on my blog and on Instapaper, and allow people to link Readability to their Instapaper accounts.

In February of this year, the app was finished and ready to launch, but it was rejected by Apple for the in-app-purchase subscription-matching rule, which had just gone into effect. Readability decided that they didn’t want to give Apple the 30%, so the app was put on hold.

By May, Readability told me that they were not going to ship an iOS app for the foreseeable future, and were deciding how to pivot their business into other areas instead. We decided to end our development contract, since there was no reason for me to invest any more work into an app that wasn’t going to ship. Our business relationship ended and we remained acquaintances, but they stopped inviting me to advisory meetings.

In June, Readability hired another developer to make an app that didn’t involve Instapaper at all. They pivoted their business away from the publisher-payment focus and into a direct Instapaper competitor. Over the next few months, they continued adding mostly Instapaper-like features to their service.

So, while they are certainly a competitor to Instapaper now, it didn’t catch me by surprise.

This is a very big and increasingly crowded market, and there’s no reason why we can’t respectfully share it.

A human review of the Kindle Fire

I expected the Kindle Fire to be good for books, great for magazines and newspapers, great for video, and good for apps and games.

In practice, it’s none of these. Granted, I’ve only spent two days with it, so I can’t share any long-term impressions. But I’m honestly unlikely to have any, because this isn’t a device that makes me want to use it more. And that’s fatal.

A tablet is a tough sell. It’s too big for your pocket, so you won’t always have it available like a phone. It’s too small to have rich and precise input methods like keyboards and mice, and its power and size constraints prevent it from using advanced PC-class hardware, so it’s probably not going to replace your laptop. It’s just one more gadget to charge, encase, carry (sometimes), care for, and update. And it’s one more expenditure that can easily be cut and done without, especially in an economic depression.

“Tablets” weren’t a category that anyone needed to give a damn about until the iPad. It was a massive hit not because it managed to remove any of the problems inherent to tablets, but because it was so delightful, fun, and pleasant to use that anyone who tried their friend’s iPad for a few minutes needed to have one of their own.

I expected the Kindle Fire to be a compelling iPad alternative, but I can’t call it delightful, fun, or pleasant to use. Quite the opposite, actually: using the Fire is frustrating and unpleasant, and it feels like work.

For most people, every other computer in their life feels like work, and they don’t need another one.

It’s not an iPad competitor or alternative. It’s not the same kind of device at all. And, whatever it is, it’s a bad version of it.

That’s probably all you need to know about the Kindle Fire. Below is a detailed account of the issues I ran into, but I won’t take offense if you’re burnt out on long Kindle Fire reviews and stop here.


The big, boring part

This is about my experience using the Kindle Fire. And I’m going to try my best to make this about the Fire on its own merits, not about how it compares to the iPad, for the most part.

I’ve read part of a book, three magazines, and a newspaper. I’ve played two games and watched four TV shows from two sources. I’ve also taken far too long to set up my email, failed to find a good RSS reader, turned a lot of pages accidentally, repeated taps that did nothing the first time, and crashed a few apps and the Fire itself.

I’ve run into a lot of problems, actually:

Interface

Hardware

Reading

Watching

Listening

Playing

App…ing

And finally, I don’t like the “carousel” flip-card-style home screen interface. Like the Xbox 360 interface makeover from a couple of years ago, it makes browsing through more than a few items difficult and slow because you can’t really see the items if you skim through quickly. It’s a poor, unusable interface metaphor that our industry should retire.

The Fire is an Android version, sort of, of the iPod Touch. It’s the first device available that’s inexpensive and offers Android in a somewhat reasonable package without a cellular contract.

But that’s just about all I can say for it. It’s a bad game player, a bad app platform, a bad web browser, a bad video player, and, most disappointingly, a bad Kindle.

If I didn’t need the Fire for Instapaper testing, I’d return it.

Amazon will take over Android app distribution

Assume the Kindle Fire will sell very well, even though it needs a lot of help.

The only way most Kindle Fire owners are going to be installing apps is from the Amazon Appstore for Android. The Fire doesn’t ship with Google Market and most buyers won’t be savvy or motivated enough to hack it or side-load anything.

It’s similar to Apple’s App Store lockdown on iOS devices: the Amazon Appstore is the only game in town on this probably-soon-to-be-very-popular Android device. Most Android developers, therefore, need to ensure that their app is available in the Amazon Appstore.

So far, Amazon has not been great to developers. (Or book publishers, for that matter.) By most accounts, dealing with Amazon is actually much worse for developers than dealing with Apple. By putting your app in the Amazon Appstore, you’re giving up a lot more control than Apple asks of us: you’re giving up the ability to set your own price and control your app’s description, among many other restrictions. By comparison, it makes Apple look almost… open.

One of the biggest draws to the Android platform, the “open” Android Market, has just been sidestepped and made largely irrelevant for tablets. If the Fire sells anywhere near its target volumes, Amazon has hijacked the Android app retail channel for the long term: most sales of Android tablet software will be through the Amazon Appstore, and if your app isn’t there, it’s effectively invisible to the Android tablet userbase.

How long will it be before this effect spreads to the much larger Android-phone market? All it would take is a deal between Amazon and one of the big handset manufacturers to preload the Amazon Appstore, placed more prominently than Google’s Android Market, on all of their phones for a little while. Amazon knows how to play the retail game — it’s their business, and they’re incredibly good at it.

A truly open facet of Android — the open-source codebase, minus Google’s apps — has enabled one company with a strong market position to step in, effectively close it, and make themselves the gatekeeper. And as gatekeepers go, Apple looks quite benevolent by comparison.

Whatever works for you

In my earlier 20s when I knew everything, I was a much bigger evangelist for my technology choices. I’m accused of fanboyism a lot more these days, but only because Hacker News keeps sending huge waves of people here who tell me I’m an idiot. But I used to be much more annoying with pushing my choices onto others.

I’ve naturally reduced such evangelism as I approach 30 and realize I don’t know anything, but I’m now making a much more conscious effort to stop it.

I spent Thanksgiving weekend in my hometown and visited my friend’s parents. They used to generously pay me to fix their computer problems, but it wasn’t always productive: everything took far longer than I thought it would, and my efforts to fix one problem often created others. It was inevitable: they’re an architect and a graphic designer, and I was a computer nerd with very little professional IT experience, so I never fully appreciated the complexity of their software setups or their priorities for getting their jobs done.

My friend’s father spent this weekend battling similar issues. Having failed to set up a suitable sync system with an Android phone, he had exchanged it for an iPhone and was trying to set up iCloud and whatever software (iTunes?) syncs calendars and contacts with Outlook on Windows. It wasn’t working properly.

He’s probably not going to get it to work, and he’ll probably return the iPhone and just tolerate a lack of a synced smartphone for a few more years until he tries again with whatever shoddy pile of hacks we’ve cooked up by then to (not) sync contacts and calendars with our other piles of hacks, a simple problem that we’ve been (not) solving for decades that still isn’t reliable for everyone.

My friend told his father to dump Outlook and go all-Apple, which of course isn’t going to happen. Previously, I’d try to convince him, too. But not this time. His father has many good reasons not to switch, and I don’t understand any of them.

I said I couldn’t help him.

The iPhone and many of Apple’s products work very well for me. But for him, they don’t. It doesn’t really matter whether it’s Microsoft’s fault or Apple’s fault or iCloud flaking out or some other third-party software interacting with something somewhere. To him, the iPhone doesn’t work with his setup.

I bet very few other phones, if any, would work exactly the way he wants with no other modifications to his setup. But that also doesn’t really matter. He got a product that claimed to work in his setup, but when he tried it, it didn’t.

I choose to fit myself into most of Apple’s intended-use constraints because their products tend to work better that way, which makes my life easier. But that requires trade-offs that many people can’t or won’t make.

Previous-me tried to persuade everyone to switch to my setup, but I now know that it’s not worth the effort. I’ll never know someone else’s requirements, environment, or priorities as well as they do. I don’t know shit about Windows or Outlook or architecture.

You should use whatever works for you. And I no longer have the patience or hubris to convince you what that should be. All I can offer is one data point: what I use, and how it works for me.

More iOS device and OS version stats from Instapaper

I get a lot of good feedback from developers who say these posts are helpful, so I don’t feel bad spamming everyone with updates every few months.

Here’s my most recent data from people using the Instapaper app:

Compared to August’s data, you can see a few unsurprising trends:

We can also extract a bit of insight into Apple’s device sales:


  1. Instapaper started requiring iOS 4.2 recently. I’ve heard from only one person who was upset that it stopped working on his original iPhone, but he also mentioned that lots of his other apps were recently updated to require iOS 4 as well. 

Ads via The Deck