Jon Alper's probably insufficiently humble opinion

At WWDC 2019 Keynote, Tim Cook telegraphed an intro video for the new Mac Pro so many of us had been waiting for with reference to  “a product missing from our  lineup.” The video ran and there it was, black, curved and reminding everyone who’s been looking forward to a new Mac Pro of the first thing you saw before scrolling down the old product page for the 2013 Mac Pro. A reminder of just why we ended up so unhappy with how long it had been since Apple had, as they said, painted themselves into a thermal corner.

Apple trolled us all with those first few frames of video.

 

 

Gloriously!

Minimizing the compromises while maximizing flexibility is a the core design theme of this Mac. It’s a profoundly different direction from the 2013 Mac Pro in all the best and most unexpected ways. Arguably, even if you never buy one, it’s an expensive (for Apple to make, not just for customers to buy) message from Apple practically shouting: ‘’We’re committed to the Mac, we’re listening to our customers and we understand some of them need something wholly unlike an iMac let alone a laptop.”

 

It’s not a sports car. This is realization of the 2010 Steve Jobs’ quote: “When we were an agrarian nation, all cars were trucks. But as people moved more towards urban centers, people started to get into cars. I think PCs are going to be like trucks. Less people will need them. And this is going to make some people uneasy.” This truck is a thundering, earth mover of a truck.

instead of another curvaceous exercise in impressive but ultimately misguided innovation like the 2013 ‘Trash Can’ Mac Pro (the 2013 Mac Pro got more of a bad rap than it really deserved, I manage a bunch of them pounding away at multi-camera ProRes 422 HD in Final Cut X, occasionally 4k, 8 hours a day for years with very, very few problems), what we got as the video played on was what most of us had never dreamt we would and many of us said we wanted. A cheese grater! As David Byrne said, “it’s cool, it’s a multipurpose shape. A box”. A monument to form following function. A full capitulation by Apple to the realities of physics purely in service to the user instead of a flexing of aesthetic design muscle. A new enclosure design that put functionality first for the first time since the PowerMac G5.

This design is near perfect given the goals and constraints outside of Apple’s control. Near, but more on that later, The entirety of the machines internals are accessible with a turn of a rotating latch on the top of the Mac that lets the user lift a machined aluminum sleeve off the top and expose all four sides of the internals. The two stainless steel handles are integral to the structure of the machine. When you lift by these handles, you’re lifting the entire machine by its very, very sturdy looking stainless steel skeleton and these handles, unlike the hand-goring handles of the G3 or the knife’s edge finger-severing handles of the original cheese grater case introduced with the PowerMac G5 and evolved through the 2012 Mac Pro, these are smooth, round and oh so very shiny. In order to present the balance one must to these sorts of articles I must point out a flaw or two with this enclosure.

It’s not black. It should be black. All the things should be black.You should look at them and ask yourself how much more black could they get and realize “None. None more black.”  It’s a rule.

On a more serious note, there has been concern expressed (First by Dr. Ian Cutress of AnandTech) that a lack of dust filtration behind the copious number of intricately machined speed holes (Yes it has speed holes, these help it go faster almost as much as a Type R badge would have) will lead to components overheating. While that’s true, dust accumulation does reduce a machine’s ability dissipate heat and crashy machines often do return to good behavior after a nice going over with compressed air, the same need arises with filters getting clogged. Given that, is dust ingress that big a concern given how easy it is open this Mac? Is the up to 300 cubic feet per minute of airflow this Macs fans move too windy for the dust bunnies to do anything but run scared?

The handles and feet, also mean it takes up more vertical space than housing its innards demand though the handles act as a roll bar of sorts to protect connectors at the top.

Why worry about that extra space taken up? The case doesn’t lend itself to being rack-mountable and that’s a deployment approach many of those who’ll need this machine will really want to take. Not just as servers in data centers but in equipment racks both built into studios and installed in portable shock protected road cases.  Would I have loved to see this machine with the handles on the front and rear and removable top and bottom plates to expose screw holes for mounting hardware? Damned skippy! Would there be some remarkably effective but space-wasteful third party solution for this like there has been every time Apple’s ignored this use case before? Sure!

But…

 

The 2019 Mac Pro will be made available in an alternative rack mountable chassis, rails and all by the look of it 4U tall. While the dream of a single enclosure readily adaptable to rack, tower or desktop uses didn’t happen, the availability of an Apple option for rack mounting at all was truly shocking news and an indication of the profound influence of the Pro Workflow Team at Apple (The link previous goes to a great @panzer article on Tech Crunch).

The Mac Pro 2019 seems to have delivered exactly as we’d never have dared hope for; an extremely flexible, fast, customizable machine remarkably free of compromises in any axis you care to measure, including price.

 

 

Those of us who’ve relied on Apple’s ‘big iron’ since the beginning had likely given up on ever seeing something like this. Not seeing something like this again, seeing something like this ever. This isn’t just another example of Apple saying ”It is the most powerful Mac we have ever created” and it being ‘true’ because it includes the latest chip that’s faster than the last and some wonderful new feature like a Retina display or controversial experiment like the TouchBar. This is the fastest, the most configurable, flexible, accessible, repairable and upgradable Mac ever made. And, it’s not, by far, the most expensive Mac they’ve ever sold, either. The Mac IIfx cost between ten and twelve thousand dollars, (1990 dollars. According to the US Department.of Labor, that top end Mac IIfx cost $39,416.14 in April 2019 dollars.)

So what do we get for a starting price of $5.999?

Power, Cooling and design theme

 

The power supply, sitting at the bottom of the machine is spec’d at 1.4kw peak power and sustained power of 1.28kw. It’s not completely clear if these wattage numbers reflect maximum load or maximum current draw but, either way, this machine offers, probably almost down to the tens of watts, every bit of electricity it can to power itself and whatever customizations you choose. Any beefier a power supply and it would push a 15A wall outlet far too close to its limit to be safely plugged in anywhere but the newest home construction and office spaces. While some servers and machines intended for more industrial applications can include beefier supplies, to push this Mac’s PSU that far would have made an already niche product simply unavailable to many users even if they had the money to buy it.

We only know some of how this power can be distributed and, if you make a choice that exceeds the limits, how the Mac will respond. Apple states “over 300 watts” is available to the CPU but the load each of the various CPU options will actually draw is unknown. Intel rates the TDP (Thermal Design Power) in watts but as a measure of heat not electricity. How much electrical load will be variable both by the workload on the CPU and from whatever decisions Apple makes in the way they set the machine up to manage it so what the actual power demands the CPU options make is still unclear.

In all aspects of the machine’s design from power, cooling, CPU options, storage, graphics, RAM and other cards, Apple seems to have pushed to limits imposed by reality, not design aesthetic akin to the way the PSU is limited to the power limited you can get out of the wall.

Starting with the fans, really, the whole cooling architecture, Apple made a point of stating that that there is more cooling available than necessary and that this cooling is quiet enough to be as quiet as the iMac Pro when sitting under a desk “in typical load conditions.” There’s some hedging there with both the ‘under a desk’ and ‘typical’ caveats, but mention was made of 12db in one interview. Three, front mounted fans draw in cool air and blow hot air out the back across the CPU, slots, the whole front side of the board. The back side of the board is half obscured by a large blower and ducting which would appear to cool the RAM and SSD modules. The cooling system has been said to be able to move 300 cubic feet of air per minute. How it sounds doing that is yet another open question.

Storage

 

The internal SSD modules are some form of NVMe sticks, presumably configured in a RAID 0-like manner for maximum throughput like the iMac Pro with a metal cover more similar in appearance to the SSDs used in the 2013 Mac Pro which are upgradable with third party solutions.

 

 

Marco Arment (who was present at the hands-on area at WWDC and apparently, asked) said on the Accidental Tech Podcast 329: Mac Pro Day that these are upgradable in the 2919 Mac Pro but there has been some concern he may have been misinformed and the T2 chip may preclude upgrades.

The good news is, this uncertainty doesn’t apply to the overall question of aftermarket internal storage in the 2019 Mac Pro. Two options have been announced and at least a third is almost certainly coming. Promise Technology has announced two products to add spinning disk but one can easily imagine 2.5” SSDs being an option in place of the platters. The Pegasus R41 uses an MPX Module Slot (more on those later) to PCI attach a four disk hardware RAID and the Pegasus J2i offers two bays (one comes loaded) SATA drive bracket which connects to the two internal, presumably, SATA III ports and power inside the Mac Pro enclosure. A tab relief and two mounting screw holes to match the tab in Promise’s J2i marketing image are visible in the top plate of the Mac Pro frame above the SATA ports on the logic board. Beyond what we know, there is at least one more likely approach to adding DAS (Direct Attached Storage) beyond the obvious Thunderbolt and USB options; PCIe cards with nVME slots.

My hunch is, Marco was told correctly but that Apple won’t sell them after market so it’ll come down to if/when third parties step up.

A USB-A port sits inside the machine on the same panel as the SATA ports and (again presumably) SATA power outlet. In a testament to the presence and thoughtfulness of Apple’s Pro Workflows Team, the stated purpose of this port is to plug in a copy protection dongle inside the machine where it’s less likely to grow legs or be snapped off which are two of the almost endless joys provided for the pleasure of paying customers to copy protect their high end software.

As a side note, as annoying as dongles are, they are in almost every way preferable to other copy protections involving online activation and node locking which are abusive in the extreme and create administrative agony for anyone trying to manage these machines in a production environment.

Adjacent to this USB-A port is a mysterious latch. I initially speculated that this was to lock the copy protection dongle in place when dragging (Or rolling! It’s the third Apple computer [but the first Mac] made to offer wheels. https://en.wikipedia.org/wiki/Apple_Network_Server) the machine around to keep it from rattling out of its connector. Matt Christensen (@_mchristensen) pointed out on Twitter that this latch likely moves the long black rail in front of the PCI slots to lock and unlock cards that provide a hook for this purpose on their board edges. Which or both functions, (something else?) this latch serves is yet another item for the TBD list.

 

PCI Slots

 

This Mac has one more slot than the “slottiest” Mac in Apple History, the Mac IIfx. (6 NuBus Slots and 1 Processor Direct slot). The 2019 Mac Pro has 8 PCI slots. Not one, or two or three.. four stones!.. Err.. Eight slots! PCIe gen 3 slots. 

One half length 4 Lane Slot is used for the Apple i/o card (presumably standard on the base config). This card provides two Thunderbolt 3 and two USB-A ports which appear to be USB 3.0 given Apple stipulating 5gbs. Because it’s a 4 lane slot, it’s likely that the Thunderbolt 3 ports share a bus and share bandwidth with the two USB-A ports as well. The Thunderbolt 3 ports carry 10 Gb/s USB-C as well. DisplayPort for these Thunderbolt ports is routed through the system from any of the three available graphics card options. In addition, this card provides a 3.5mm headphone jack with headset support so a mic input as well. None of the combination digital optical and analog audio magic we’ve seen in the past are here and it’s not clear how port will negotiate line levels, in or out.

This all being on a card seems (again according to Marco) to be another example of Apple leaving the decisions about where to compromise up to the user as much as possible. If you don’t need these ports (there are two more Thunderbolt 3/USB-C/Display port connectors permanently installed into the top of the case), apparently, you can pull (or not spec?) the Apple i/o card and use the slot for something else. If you’re running a recording studio, this audio jack will not put your oars in the water. You’ll have some other audio (and MIDI) interface instead. The ability for the end user to forego this card to put those four lanes to other use sacrifices back panel Thunderbolt but at the very least, Apple’s preserved flexibility for their own future designs (and repairs) by putting this i/o in a card. It’s unknown whether you could, for reasons unfathomable, buy a second one of of these Apple i/o boards. (even if so, are there limitations to display port availability in other slots?)

Assuming you choose the single slot AMD Radeon Pro 580X GPU card option and leave the Apple i/o card in place If, for example, you’re more focused on audio or crunching big data than you are graphics, that’s six slots left! 

The whole “Six Slots” thing was a big deal in 1999, Avid showed Apple the door for Media Composer support on Mac because Apple wouldn’t ship a six slot Mac, The last and only ’six slot (PCI) Mac’, the 9600 retired in November 1997.  These were dark times for the Mac video professional and Apple as a whole. A former employer was not one bit pleased and MacWeek was a buzz with doom editorial. (Any possibility that the impending release of Final Cut Pro in 1999 might have contributed to AVID’s six slotted temper tantrum is pure speculation)

Recalling that bit of history, there is a delicious irony in that if you’re more focused on audio or crunching big data than you are video and choose the AMD Radeon Pro 580X and leave in the Apple i/o card, this Mac is the first computer to support 6 Avid ProTools HDX DSP cards. (No word from AVID yet as to the impact of this machine on their other product plans.) 

This scenario, the decision to use six slots for AVID HDX boards is an example of where, because of the truly remarkable flexibility of this design, the decisions you make when configuring this machine will be highly dependent on the compromises you need to make to have it serve your intended purposes. We’re back to compromises again, but thanks to Apple’s choices, those compromises are overwhelmingly left to the user rather than predetermined by the design of the Mac itself. 

Card counts and lanes. The processor (assuming I’ve guessed correctly) supports a total of 64 PCI lanes. So, while the total number of slots and the marked number of lanes totals 92 the CPU only offers 64. That’s 28 lanes you can’t use. (Installing a 4 lane card in a 16 lane slot works the reverse sometimes works but slower)) The supporting Intel Chipset (C620?) provides 20 more PCI lanes but of those are likely consumed by the two network interfaces, built in Thunderbolt and the two built in SSDs and, perhaps WiFi. The Bluetooth (via USB) and SATA (and the chipset supports SATA III) and some of the other stock USB is provided from the support chip directly without using PCI lanes.

The reason the listed numbers of lanes exceeds the total is so you’re not stuck wanting three 16 two cards and only having two16 lane slots but if your card choices need all the lanes in their slot you’re ‘borrowing them’ from elsewhere.  Don’t let this freak you out though. Those AVID HDX cards are each 4 lanes. An AJA KONA 5 needs 8 lanes. (And by way of putting the Mac Pro’s price in perspective costs $2,995). All this perhaps a  backward way of suggesting many of the cards you may end up sticking in a 16 lane slot won’t use all 16 lanes.

Graphics

 

The base configuration’s graphics card is the aforementioned double-wide card that consumes one of the 8 PCI slots. This “great all around graphics card”, the Radeon Pro 580x, routes four display port signals to the internal Thunderbolt ports, has two HDMI 2.0 ports and 8GB GDDR5 video memory and provides 5.6 teraflops single precision, 36 compute units and 2304 stream processors. It should be assumed that no matter what other graphics card you choose to install (and Apple has explicitly said “the slot maintains compatibility with standard cards” and there are two 300w 8pin aux power connectors available and what looks to be some other unspecified smaller aux power port toward the front edge of the logic board) in the three remaining double wide, full length 16 lane slots, you will likely need to, and almost certainly want to leave one of Apple’s cards in place.

It should (perhaps must?) remain installed in order to provide display port to the Thunderbolt 3 ports on the Apple i/o card and the built in top mounted Thunderbolt 3 ports. How does it do that? The bottom and third from the bottom PCI slots are meant for what Apple calls an MPX module and the base Radeon Pro 580x is packaged as a ‘half height’ MPX Module. Each of the two MPX module bays (a full height MPX module is quadruple wide) has an extra connector on the card edge that fits into a second Apple specific slot in line with the PCI slot but further forward that carries display port, PCIe and delivers up to 500w of power. This card drives up to six 4K displays, two 5K displays, or two Pro Display XDRs.

Before you embark on the likely futile effort to figure out all of the many NVIDIA cards Apple didn’t choose to supply with this machine you believe are better and cheaper than Apple’s two higher-end MPX offerings, so you can complain about this Mac, you do, theoretically have the option to supply your own.Unless there’s a sudden detente between Apple and NVIDIA (seems highly unlikely) and/or that Apple opens the spec for the MPX Module, you’d likely leave the Apple card in place add a standard PCI NVIDIA card and rely on NVIDIA to build a Mac-supporting card that navigates this cooling airflow and maintain usable drivers. Absent any official detente, installing your own NVIDIA card with NVIDIA drivers seems like a recipe for all the fun Windows users enjoy dancing around with OS updates, GPU drivers and the uncertainties about which apps get worse and which get better depending on which GPU you choose.

The lower end of the other two MPX Module offerings at launch is an AMD Radeon Pro Vega II with 32GB of HBM2 Memory and 1TB/s (that’s big B Byte according to Apple’s specs) bandwidth, 64 Compute Units, 4096 stream processors, 14.1 teraflops of single precision or, if about half vague is cool with you, 28.2 teraflops of half precision. This is a ‘full wide’ MPX module so it will obscure one additional 16 Lane full length slot. This MPX Module features four Thunderbolt 3 Ports and one HDMI 2.0 port. Support for up to six 4K displays, two 5K displays, or two Pro Display XDRs. Apple notes that this “Full-height MPX Module fills an MPX bay and uses extra power and PCIe bandwidth” and this may help explain some of the additional mystery of that ‘other PCI slot with the extra pins” that, in part, defines the MPX module. It would seem that what Apple is saying with that caveat is that while it the full height module obscures one of the ‘normal’ PCI slots, you get some or all of the missing lanes back in the form of those four Thunderbolt 3 ports. If you get all 16 lanes you’ve covered up back? That’s up to four more Thunderbolt buses. Apple taketh away and Apple giveth back bigtime. A means to use a slot obscured by cooling for anything at all isn’t something I’ve seen on any other machine from anyone.

The top end graphics offering is an AMD Radeon Pro Vega II Duo. That would be two AMD Radeon Pro Vega II GPUs stacked up together in a quadruple wide trailer. And shacked up they are. The two GPU chips live side by each like two eggs on a pair of toast on the same board and are connected via what Apple (or is it AMD?) calls “Infinity Fabric Link” which they say connects the two GPUs at 84GB (big B bytes again) per second. “That’s five times faster than the PCI bus” Because the two of these chips writhing around in concert (or is it congress)  are under the same fabric sheets, they double the maximum pixels they can push to up to six 4K displays, three 5K displays, or two Pro Display XDRs. You can install two of them. How much power will be left for anything else adds to our list of open questions.

Regarding that NIVIDIA thing, in claiming ‘the worlds most powerful graphics card’ Apple said,Maxon tells Apple that Cinema 4D is “twenty percent faster  “than a Windows PC maxed out with three of the latest NVIDIA Quadra cards”. Not mentioned is whether this comparison needed two AMD Radeon Pro Vega II Duo cards and, yes the Mac Pro sports two.

Another enigma wrapped up in this box alongside an ambiguity or two is; what’s that connector on the other card edge of both of the two higher end MPX offerings? When describing Infinity Fabric Link during the WWDC  Keynote, John Ternus doubled the specs of the Duo module when he pointed out the Mac supported two of them. Is it possible this top connector allows the linking of two MPX Modules via Infinity Fabric? If so, do they both need to be the Duo? Can they be two of the single chip? One of each? People with soon-to-be-depleted-bank-accounts will want to know!

AfterBurner

 

Apple also announced another PCI card, the single slot AfterBurner, an FPGA (Field-programmable gate array) tuned for Pro Res and Pro Res RAW decode. For video folks on high end jobs, this could be a really huge deal.

If you’ve never explored Pro Res before and you’re a video editor, now is the time to go read up. Aside from Pro Res, Avid’s DNX is the only editing codec you should tolerate in your life unless you your camera both forces and justifies it’s native format. (There are some really good acquisition formats but not on cameras that don’t cost more-than-new-Mac-Pro money. (BMCC perhaps aside.)) Apple states the Mac Pro with Afterburner can play back 3 streams of 8k Pro Res RAW or 12 streams of 4k. AVID editors notwithstanding, If your work demands this, the Mac Pro with AfterBurner is a no brainer. What’s still to be determined is the value if you’re not pushing that hard. A 2013 Mac Pro can happily chew up and spit out more than 3 streams of HD Pro Res 422 all day with fast enough storage. Does AfterBurner meaningfully raise the HD stream count? Does it accelerate ProRes Encode? (AppleInsider says yes  to encode but every person Apple to mention it that I’ve heard has only said playback (decode)). Will your workflow and storage infrastructure mean you’re more bound by file i/o performance than Pro Res decode? What are the AfterBurner benefits relative to the benefits of raising the core count of the CPU you select? To put it another way, just because you’re a Final Cut Pro X shop with the wisdom to shoot, edit and master in Pro Res doesn’t mean your content and workflow won’t get all the benefit you can actually use from CPU and GPU without buying AfterBurner. Then again, if it’s twenty bucks, why the heck not? Pro Tip: Don’t expect it to cost $20 though it it’s not likely in the same cost brackets as the higher end MPX modules.

All these questions will need to wait until it’s benchmarked. But.. while it would be unwise to base purchasing decisions on this, Greg Joswiak and Craig Federighi had some tantalizing things to say on the live WWDC episode of The Talk Show with John Gruber. Specifically they said “it can be reprogrammed.” Then Craig, “There’s more to come….I mean, I’m not announcing anything but.. one could imagine.” Oooh let’s imagine! Reprogrammed for audio DSP doing for Logic what AVID’s HDX cards do for ProTools? Or, completely out of the media universe, a developer API to set it work on tasks like CoreML training? But hey, since we’re imagining, Joz also said, “in milliseconds.” Could one hope, one day, to buy an AfterBurner, or three, and have them dynamically re-assigned to accelerate different tasks? More from Joz: “It can literally process six billion pixels a second” and Craig, “virtually like dedicated hardware speed.” one can imagine a lot indeed.

I’ll take this whole imagining thing too many steps too far. Afterburner is shown slotting into what’s labeled as slot 5 16x. Does it need all 16 lanes? What, perhaps, could one do if these were installed in a Thunderbolt external PCI chassis?

Never buy any tool that can’t, out of the box, solve a problem for you but, what was said on the Talk Show points to some interesting possibilities. This, of course, assuming that any of these Mac Pro components, the graphics cards MPX modules, the i/o boards, or AfterBurner, are available after market or only at CTO time when buying the system. That uncertainty is substantial.

The potential though, if any of the above comes to pass starts looking quite interesting and this may be the sleeper bit of hardware news.

CPU

 

“To get the last ounce of performance out of this processor, we’re giving it over 300 watts of power and a massive heatsink for cooling. So this means it can run full unconstrained, All the time.” – John Ternus

 

By “that processor” we can assume Mr. Ternus was referring to the 28 Core model. How much CPU core count will matter for what you want to do with your Mac is an open question. Actually, it’s a whole bunch of them. Cost being the first. We don’t yet know which specific of the latest Xeon CPUs Apple has chosen but it seems likely to be from this bunch and the retail prices, for the CPU alone, run from the $749 Xeon W-3223 with 8 Cores to the $4,449 Xeon W-3275. There had been some very smart speculation from @octothorpe and @tankgrrl on Twitter that may indicate I have the wrong guess as to which chips Apple has chosen and, if they’re right, as they point out, it may be that the there are also motherboard variants depending on CPU chosen at config time. The question hangs on the number of memory channels being six for all chips offered or only four for one version of the 8 core chip people have speculated Apple may have chosen. The specifics are yet more questions we’ll need to wait for more answers before we start falling in love with a particular configuration. One hint that @octothorpe may be onto something is that the 32GB config in Apple’s specs page is set up as four 8GB DIMMs while the remaining memory options are groups of six sticks. It’s possible that even a six channel CPU may not require that all six channels have RAM available and only that balancing the channels equally optimizes RAM performance, The specifics of how best to populate DIMM slots is yet another open question.

 

 

Networking and More

 

 

Beyond the WiFi, two 10Gbps RJ45 ports beckon at the back of the machine adjacent to the power inlet port near the bottom. The flexibility you get with two ports is substantial. Local shared storage like a https://lumaforge.com/jellyfish on it’s own LAN and you can use the other port connected to a LAN segregated from your storage network to connect with collaborators from other departments (save every bit of 10Gbps Ethernet bandwidth for the shared storage, really, it’s a good idea). Of course, if you’re a Chrome user, you may find you need the 28 Core CPU option if you use more than one browser tab at a time. (An exaggeration of course but for lots of good reasons. don’t use Chrome)

While we’re on that side of the board, we have another few mysteries to ponder. For example, in the cooling discussion above we learned of three fans and one blower and a front to back airflow. The fans, perpendicular to the mother board cool CPU, cards and MPX modules and any of those additional storage options discussed above. The blower, ducted, blowing air in the same front to back direction but mounted parallel to the mother board on the RAM side. It’s clearly shown during the Keynote drawing air over the RAM slots but it also showcases two more mysteries. First, what is the hotspot shown near the center of the board? Seems likely to be Intel support chips for the CPU. Second, there appears to be a door. A rounded rectangle with two latches above the blower. One might speculate that this hides the WiFi 802.11 a/c (a/b/g and n compatible) and Bluetooth 5.0 hardware but the door, if it is a door, looks big for the job. It also raises yet another question. Where are the antennae?

It appears this Mac comes with yet another ‘sure to sell for insane amounts of money on eBay unless and until Apple makes them available for purchase separately” set of input devices. A full sized keyboard with correctly colored (black) keys and silver frame, a Magic Mouse 2 in silvered black, or a Magic Trackpad 2 also in silver and black. Being called a Magic Mouse 2 just as the current model is doesn’t paint a rosy picture for Apple finally putting the charging port in a saner location, where a mouses tail belongs).

Configuring yours

 

When it comes time to buy a few of these machines for the house, the yacht and the bunkhouse, the answers to the questions above (summarized below) should be matched against some very careful reflection about what work you’re planning to do with this beast of a machine.

 

Power Questions:
– How much power does the power supply deliver? Is 1.4Kw what it pulls from the wall or is it what you can expect to get out of it? (No power supply is 100% efficient.)
How much does the Mac pull from the wall? Yes really, you need to know. don’t bring it home to discover the outlet you have in mind shares a circuit breaker, or a fuse, with another room.
How big a UPS do you need?
– How much power do the various options actually consume? Just because an MPX Module has 500w available to it doesn’t mean the current ones on offer all pull 500w.
– What will the Mac do if you put too much load on the Power Supply. Did Apple get all Apple-y and implement a mechanism for you to troubleshoot it with a software integration?

RAM Questions:
– How must the RAM be installed to optimize performance? Is it so finicky it needs matched chips or not. If matched, how matched and all of them or just pairs in a channel?

AfterBurner Questions:
– If you’re considering AfterBurner, how much will it currently improve your workflow? Will your workflow evolve in the near term to benefit from AfterBurner? A shift in acquisition format to ProRes RAW could be a justification even at HD or 4k. This is a very different question from “How will AfterBurner evolve to support your workflows.” It’s desirable to assume it will but it doesn’t make fiscal sense to buy on hope, unless it’s surprisingly cheap and not available after market.

MPX Module (GPU) questions:
– Are you planning a graphics intensive use case or not so much? If so, how much? It seems likely that a AMD Radeon Pro Vega II is a sweet spot for general purpose media use. But again, the questions of what options will be available after market from Apple should impact your thinking and you should know what they are before you buy. If these cards will be available after market then maybe start as cheap as you can if you’re coming from older hardware. If there is an Infinity Fabric Link with some connector between AMD Radeon Pro Vega II modules, then start with one and add a second later as needed. If there’s no clear information from Apple about availability of components at retail and you can possibly afford it, spec the AMD Radeon Pro Vega II Duo because it may be the only chance you get to choose it.

CPU Questions:
– What are trusted benchmark sites saying after review models go out? By trusted I mean Ian Cutrress at AnandTech https://www.anandtech.com/Author/140 or the gang at ArsTechnica more than the sort of site that used to try and pitch you on the innovation of Power Computing Macs back in the day. What are the sweet spots for your use cases? Is an extra grand (or more) better spent on more cores, more GPU or an AfterBurner, or, fo that matter Apple SSD?

Enclosure Questions:
– If you’re not going into a datacenter, what is the rack mountable option like? Is it as quiet? Do your workloads and environment mean it would go into a rack behind your mix desk or in a road case with rails? Do you need the wheels? Bear in mind that while Doug Brooks, the Apple Product Manager for the new Mac Pro made a believable case on Mac Power Users case for the ruggedness of the design and value and use of the wheels, sliding the Mac across your flat and level workplace floor on the Apple wheels is something very, very, different from pushing it onto the lift gate of a truck. The latter, you want it in a proper road case.

The Random Head Scratchers:
– What’s the mystery latch do? The card lock? Sliding bar to catch the hook on many PCI cards? Lock in the USB-A stick? Both? Something else?
– Where are the WiFi and BlueTooth antennas?
– What’s the deal with those two 8 pin Power Connectors described and what’s up with what looks like two others, one adjacent to the SATA connectors? The other labeled 5-6?
– What, other than RAM does that blower keep cool?

And, maybe the least obvious and in some applications most important questions:How loud does it get when you make it beg for mercy?

Who is this Mac for and why all the hoopla?

 

So, who needs this Mac? As with any Apple product announcement the reactions span the gamut from giddy excitement to bitter resentment. A lot of why this range of reactions is in play with this product is that it’s called “Pro”. It’s that damned word, “pro”.  I’m a Pro so if it’s not for me, I’m personally insulted. I’m a pro in a field other than the ones this machine is for. The work I do on my computer is a sideline or hobby but I am a pro because I get paid to do it. To understand the Mac Pro 2019, we need to define what the “Pro” needs and that’s no easy trick but it boils down to this: A Pro, as defined by this product, is somebody who needs to do an absurd amount of math on a ridiculous amount of data as fast as possible at sustained levels day in and day out.

A brain surgeon is a pro, she doesn’t need a Mac Pro. A photographer can be a pro and much as we may feel like the thousand 40MB RAW stills we shot last weekend are massive amounts of data but in the ballpark of one hour of ProRes422 HD video. A writer can be a pro but even if they’re writing The Iliad II: Homer Goes To Springfield, they’re not dealing with that much data or doing much math on those data. If the Mac Pro is too expensive for you, it doesn’t make you any less brilliant, capable, worthy or less ‘professional’. Saying it’s not the machine for you is not the equivalent of saying “you’re not a good enough driver to own a Ferrari, go buy a Prius.”

This computer is for (aside those lucky enough to be able to afford it who just plain want it, which is fine too) anyone who either couldn’t do their work without it or who, because of this machine can earn more, or achieve more with it. If you’re making a movie in 8k? Well here’s hoping it’s going to play in IMAX theaters because otherwise, it’s just a bit silly to work in 8k at this point. But if you’re making that movie, you need this Mac;  if you’re working with some monstrous, multi-hundred gigabyte dataset and you need to park it all in RAM to analyze it and you’re a Swift programmer, you need this Mac; if you’re operating a studio with clients like Trent Reznor or recording the score for “Star Wars: Jar Jar’s Revenge” for Disney, you need this Mac; if you’re just churning out hour after hour of corporate or educational video with too small a staff budget and too little time, you may need this Mac; and you need this Mac if you’re doing a whole season of a reality show with ten 4k cameras.

So that’s need. But who should buy it if they can, with a hard stretch, safely afford it but can’t say they need it? That may be the hardest question to answer. One way people seem to ask this question, “is this a ten-year Mac?” was raised on ATPFM and thoughtfully responded to (as always) by Jon Siracusa in 329: Mac Pro Day. (Though I don’t think they really predicted this in the prior episode. ;-)) You definitely need to listen to his take on it but here’s mine: there’s no such thing as a ten year Mac but as the Macs in my house, and those at work have reached six and eight years old. I’d have preferred they hadn’t gone that ling and I aim to follow the three-year-lifespan theory when budgeting for my actual work, I can say there’s sometimes, often, such a thing as a five-year Mac. Even if Apple were to announce ARM Macs, it’s unlikely to happen before this time next year. When they are announced, if they are, they’re likely, at first, to be released in the categories where they’d have the most advantage over Intel: specificall,. performance-per-unit-heat and battery life, as in MacBook Adorable-class laptops. To replace Intel in “Pro” Macs, Apple would need to prove the viability of the A series chips under sustained load and support features like PCI and Thunderbolt. They’d also need to have bridging technology to manage the transition from Intel to ARM instruction sets in all cases. 

Catalina supports Macs dating back to 2012 and only now ends the long process of deprecating 32bit software. Generally speaking, macOS gets security updates for two years after it’s been replaced. Apple publishes their policy on vintage and obsolete products as five years. So, this, unless your need for performance evolve past it, is comfortably, a five year Mac. If it gets the kinds of speed bumps we’ve so happily been seeing Apple get back to providing lately, then can you hope to upgrade it to stay current? Maybe. The CPU is socketed but nothing ensures the next generation Xeon chips will have the same package.

Then there’s PCI. if the processors are the ones linked to above, then they’re the reason this Mac is PCI Gen 3 rather than 4 but the next speed bump may also go there and make a new logic board necessary. With Apple’s decision to put i/o on a daughter card, it may well be that USB-4 to the extent it’s meaningfully different in the end from Thunderbolt 3 (doesn’t look it) may be something on an upgrade path.

The thorny question that seems, as always to be provoking a fair amount of whining in the internet, “is it worth it?”

A lot of very smart people didn’t seem to want this Mac. They’re pros. Just not the same kinds of pros this machine was built and therefore priced, to serve. For many of the users I know who feel this Mac overshot their cost and spec marks, what they seem to want is a PowerMac 7600 class of Mac (damn what a deal that machine was!) . The Xeon CPU and ECC RAM perhaps, A slot or three and a replaceable GPU. This is a perfectly reasonable desire and I’d love to see their needs met. I think it’s yet another segment for the ‘PC as Truck’ user segment. I sincerely feel their pain and I sincerely hope Apple decides they’re a market. A price gap between Mac Pro and iMac Pro it could slot into nicely. The investment in this Mac Pro, I think, raises the odds for users like these. That said, what if Apple had built that less over-the-top Mac Pro and not the one we got? That would have been short of the mark for the target market of this machine and of the value a ‘halo’ Mac brings to Apple. If they can only do one, or do one of the two first, this is the one Apple needed to do. That’s not to suggest the pros who want more than a mini and don’t want all the internals glued in behind a built in display aren’t right to want what they want. They absolutely are and those with this need aren’t whining. They are, I hope, making their wishes clear enough to Apple that Apple builds what they need.

What actual “whining” there is comes from a couple of much less reasonable mindsets: “This is a ripoff I can build a Windows PC” or, “A Hackintosh is cheaper!” Maybe you can. Maybe you should but if that’s the case, that doesn’t make this Mac a ripoff. Not by a long shot. The 2013 Mac Pro sitting on my desk has been anything but and as I said above, the dozen or so I support have more than justified their price in the six years they’ve served. The basis for the ‘ripoff’ argument comes not from the ‘can he machine earn more than it costs’ assessment but rather from pulling apart the bill or materials on the Mac Pro and ignoring the value of integration, quiet cooling, a system warranty and, frankly, the aesthetics including how loud it gets under load which do matter for quality of life with the machine. The BOM doesn’t tell the whole story.

One good and thoughtful take in the question is here.  Even Linus of “Linus Tech Tips” YouTube channel has a mostly reasonable take. Linus is technically very well versed, though not necessarily in professional (as opposed to YouTube professional) media production [See above re: the problems with the word “pro”]. He’s just got a ‘thing’ about Apple the way many PC enthusiasts do manifest at its most extreme with words like, “extortionate.”  In addition to a few factual mistakes (minor like top mounted Thunderbolt, not, as he says, front mounted), he comes up with a parts list (missing cooling and not specifying which case, which motherboard and the wireless bits or duel 10G NIC) that adds up to $3,160. He asks “Is a system bristling with Thunderbolt 3 Ports worth nearly a $3,000 markup?” Ignoring aesthetics, quiet, warranty, integration, time saved not sourcing and building the machine and wrangling any incompatibilities you find along the way? If it’s not worth it to you, then you’re not the customer for this machine. The customer for this machine is a professiona needing an integrated and supported solution. They’d buy a Windows workstation and pay about the same price the Mac Pro will cost (RAM and storage CTO markups notwithstanding, Apple legitimately loses in comparisons there.).

Now the unpopular position. If you’re comparing price to a Hackintosh, you’re not in the target market either. This is a product for professionals and professionals don’t stake the reliable performance of their work for clients or employers on license-violating bodge jobs. Sorry, not sorry. But…as a hobbyist is it probably a good thing for the Hackintosh users that there’s a Mac with support for a broader PCI card ecosystem? Hellz to the yeah!

This article will be updated as we learn more, and my inevitable errors are called out. Meaningful editorial change will be noted, typos and awkward wording, just fixed on the sneak.

For more about The 2019 Mac Pro check out @RichzardTaylorTV on Twitter and at fcpx.tv where he’s maintaining an ever growing list of Mac Pro articles, videos and information

[Posted and edited with MarsEdit ]

 

UPDATE 6.12.19: A variety of typos fixed and some small tweaks of language to clarify where folks had asked questions or commented on ambiguity. Then there’s this piece from Dr. Ian Cutress at AnandTech explicitly discussing the CPU choice in Mac Pro. Apparently confirmed my guess broadly. (specific chips from the line still uncertain).

It appears there’s a bug in a certain podcast client that’s really serious. I’ll let the real journalist who’s been working on it keep the scoop he more than deserves. I am looking forward to his work on the issue. If he does publish something, I’ll update to link to it here.

Talking about it with him on Twitter as he chased it down reminded me of this old piece from 2006.

I published the post below pseudonymously at the time because my job precluded publishing pieces that could be misconstrued to be the official position of my then employer.

The blog it appeared on is now defunct. There was only this post and one other I didn’t write alone and didn’t completely agree with in tone so it’s not like there’s a lurking archive of old verbosity you can fear being shoveled at you as re-posts here. [Edit 6.9.19: It’s not as obvious as it should be, the old post below was written sometime around 2006 during the first Podcast Gold Rush and the point wasn’t “Podcasts are pointless” it was “If your business is making television programs, investing a $100 in a podcast to promote one wasn’t an effective strategy. It was ‘chasing the new shiny’ without understanding the medium or the audience.” A lot of the underlying ideas below are still valid but it took iPhone (and much cheaper bandwidth) to make podcasting mainstream. This was an argument against a multi-million dollar media entity dealing with a lot of other pressures on its business model making an investment in podcasts as a promotional vehicle for its core business.]

 

———————————-

The podcast question

A question I’m all too often asked is: “How do I make/promote/submit a podcast?”. That, for the overwhelming majority of people who’ve asked it of me, is the wrong question…

As promised, the topic for this post is podcasting. Were I to succomb to my more primal nature, I’d begin with a stream of invective usually reserved for the guy who finished the last beer before the the pregame show ended. In an effort to walk the delicate line between fully conveying my sentiments on the subject and not allowing my feelings to overshadow the real issues, I’ll try to restrain myself.

This is long. About four thousand words worth of long. Executive summaries are for people who don’t have time to digest a complicated problem before making a snap judgement that will affect countless lives and move millions of dollars. You, dear reader, are the reflective type who really wants to explore a topic and get to the meat of the matter.

For the moment, we’ll restrict the discussion to audio-only podcats rather than video podcasts or (gagging sounds reminiscent of Bill The Cat) “vodcasts” the inanity of which should be apparent by the time we get there.

I’m not going to cite particularly bad podcasts or podcasters because, well, it would be like shooting fish in a barrel and there’s no need to be mean. More important, if I cite them, you, dear reader, would be encouraged to go subscribe to one to confirm or disupte my assessment and the last thing I want is to send them more traffic.

You can pretty much randomly sample podcasts produced by individuals and get a clear sense of the worthlessness of most of them to a general audience. They’re not worthless to their makers and even a close circle of people with a personal connection to their makers. They’re worthless as media. If that distinction isn’t clear or is a distinction you find offensive, the rest of this piece isn’t worth your time to read. Stop now… enjoy yourself elsewhere… it’s fine. Really. Personal publishing is a good thing… it’s just not what this piece is about.

I have no complaint that people use the internet to publish their highly personal opinions and anything-but-journalistically-informed commentaries. That the internet fascilitates this is a good thing. My complaint about podcasts has nothing to do with the volume of crud out there but rather how the buzz about podcasting is making otherwise highly capable, informed and content-driven producers get themselves all tied in knots trying to be players in a “new genre”, or worse yet, making Phred the Venture Capitalist throw good money at doomed business endeavors in this new “space”. Why do I care if VCs waste money? Well, money invested in the pointless is unavailable to the worthwhile. [2012 Note: How many Podcasting startups from 2006 are booming concerns now or got sold for big bucks? I count, let’s see, none.  There are many extant podcasts that are terrifically good but that’s surviving content, not this business:  http://en.wikipedia.org/wiki/Podzinger which was being very aggressively pitched to me at the time  and which is now moved on to broader and hopefully more fruitful pastures. The underlying tech was interesting. Hitching their business model to podcasting was unwise and I told them so at the time.]

Let’s begin with what podcasting can be when it’s good. A podcast, when it’s good, is like a recording of radio. That it’s a recording of radio rather than live radio is an important distinction and one that further reduces the general utility of podcasts as we’ll discuss later.

Podcasts usually fall into one of a narrow set of genres: audible op-ed, radio play, confessional, act of audio exhibitionism or interviews. The overwhelming majority of podcasts seek to be in the audible op-ed category. Frighteningly few podcasts strike a useful balance between the Op and the Ed. They frequently dregrade into the mire of pure Op (Opposite). This is the kindest possible description. In reality most are rant-du-jour or autobiographical. They’re trivially easy to make, have a way of rolling off the tongues of their creators and they are, of course, cathartic for the maker.

Another way podcasting can be good media is as story-telling. There was a great tradition of radio play but the pre-television glory days of radio dramas are, sadly long behind us. (I know I picked a post-television example. Do you think maybe I did that by way of not-so-subtly pointing out that the medium shouldn’t be as dead as it is?)

Even at its best, however, podcasting is far less flexible than radio for the content producer because it’s pre-recorded and the overwhelming majority of radio is live. Talk radio remains a wildly popular genre whether you like Rush Limbaugh or Howard Stern. Some radio is pre-recorded but much of the best of pre-recorded radio retains much of the intimacy and immediacy of live radio because it was recently or topically recorded live (a news dispatch from a BBC reporter in the field for example) or is delicately produced or, is music programming with a live DJ.

Podcasting is inherently time-shifted and this has significant content implications. You can’t realistically have a call-in podcast. Well, you could but the request for a listener call-in and the actual comment being heard in a podcast would be temporily disconnected in a way that’s difficult (at best) to produce around. Creating the sense of immediacy and engagement that drives call in show popularity without being live is so difficult as to approach impossible for all but the most adept producers. There are exceptions like NPR’s Car Talk becuase it’s extremely well produced and the artifice of “call-in” is delicately maintained and not over-played while the hosts are uniquely engaging. To expect a podcaster to hit this bar would be unreasonable on many levels.

The flexibility that podcasts offer the user to download and forward actually creates constraints for the creator. These constraints include rights clearance for downloadability, the need to create in-band context, the latency between creation and audience reaction, cost per user to distribute etc. These are all ways podcasts differ from radio.

So, if you have an ongoing and rights-cleared source of essentially ephemeral content that is particularly well suited to a radio-like user experience, and if you want to fascilitate the automated offline use of this content for your users, then podcasting makes sense. A collection of (not breaking) news stories or profiles a commuter will want to listen to on long subway ride to work is poster-child content for podcasting and this use case is what’s best supported by the delivery system invented by Kevin Marks and then riffed on by any of the various folks who make much more noise about their contributions. Unless you (or, more precisely, your content) meets this pretty narrow set of defining criteria you’re wasting a lot of time, effort, and bandwidth podcasting. If you’re thinking a podcast is a great way to promote your website, you’ve got the whole een-tar-web thing turned upside down. If you think a podcast is a great way to promote your television show, well…

The audience measurement models of podcasting are broken.

The pathetic ratio of podcasts to subscribers is a function of the early adopter appeal of both creation and consumption (More recent articles don’t present it so clearly as that piece from 6.6.05 but look at your own logs. If they’re anything like what I and colleagues are seeing the weirdness is still there only the gross numbers are larger). The idea that such a personal creative work could get a mass audience drives creation among the nominal digerati. The white knuckle terror that somebody else will beat them to market in the next big thing drives Big Media to disgorge all manner of inane podcastery. The development of products like PODTRAC is an attempt to make sane the irrational optimism that drives fruitless investment in podcasting. These guys seem to be trying to find the right answers and I haven’t the time or space to deconstruct their approach completely. The problem is, the premises of PODTRAC’s methods start seming thin as soon as you begin examine their claims. To choose one example, they claim they can report feed subscriptions and differentiate partial rather than complete downloads. Sounds good except basic http log analysis would tell you the same things. Not necessarily a flaw in PODTRAC per se but careful engineering of a web site coupled with log analysis will tell you a lot more about user behavior than even the best of what PODTRAC claims.

As people talk about podcast usage there’s a lot of confusion. The conflation of terms and the muddy thinking that comes from it is evident here. Read the sidebar. They mix terms all over the place. If no deliberate obfuscation is intended (and I’d bet it’s not), then they just aren’t really thinking about the data. If you really want to know about actual usage of your podcasts and don’t want to give away a Porsche by mentioning it’s free to the first person to hear the offer twenty minutes into your podcast and call you at your mom’s house, there’s Audible who’d like to convince you to use their proprietary format and pay for the tracking. Om Malik discusses the Audible offering in his blog but the response from Audible’s blog consultant does more to make me doubt the idea.

As the medium matures, one would think the audience to creator ratio would rationalize and competition and appeal would create a balance similar to the web as whole: An enormous volume of crap amateur content with the cream rising to the top alongside “professionally” produced content. Issues intrinsic to the nature of podcasting preclude this outcome for some very concrete reasons. Changes in the way podcasting is implemented may improve this but, those technological improvements are really all about making podcasting more like downloadable files made available from websites and less like long-dead push technologies like Pointcast. To fix podcasting, one must innovate backward.

The discovery/promotion model of podcasts is broken.

Users subscribe to podcasts via tools like iTunes and “discover them” via aggregation sites like Odeo or Podcast Alley (you’ll note there are no links here…that stuff is bad for you…go eat a carrot stick). Distributors create a light weight metadata structure to describe the podcast content (rss) which, by design, is limited in its ability to describe the content. Based on this cursory description of the content and, in some cases, on a graphic image, users are supposed to decide a particular podcast is of interest to them. Some sites let users sample a podcast. The enormous volume of noise (see above) means that finding subscribers for your podcast requires an investment in promotion that actually exceeds that required to promote a web site. Why? Because podcasts can’t be found by search. There are people attempting to remediate this problem like Podzinger but that, in and of itself, is broken because to search for a podcast on podzinger you have to know you want one and… what exactly is a podcast? Try explaining podcasting to your mother and ask her if she’s interested enough that, after you’d subscribed her to NPR, The NYT or another podcast of particular interest to mumsy, she’d go looking for more. If she is, see if she can do it. To find a podcast, you have to go looking for a podcast. The “phenomenon” of podcasting is irrepairably self referential. It’s like blogs. You know when a blog becomes good? When it becomes a website. Funny thing about that innovating backward business. Ole Tim had him one honkin’ good idea when he invented the web.

The playback model of podcasts is broken.

The podcast client downloads the episodes linked in the feeds the user has subscribed to and the user either syncs the audio files to a portable device or plays them on their desktop pc. Assuming the user gets a single audio file per pordcast, they queue up and play the day’s haul. Each file has needs its contextualization in-line in the audio file. If I collect six episodes of “Little Timmy’s Daily Podcast About World of Warcraft” each episode will start with “Hi, I’m little Timmy. This is my podcast about World of Warcraft.” If it doesn’t start with at least that, the user who receives a single file or who subscribes to the feed after an introuctory file has “scrolled off” the feed doesn’t know what their listening to. Even Little Timmy knows enough to know that’s important. (Yes there is ID3 metadata available but this is pretty shallow. Playback environments are inconsistent in how much they show and how they deal with prioritizing RSS metadata vs. ID3 metadata.)

In point of fact most podcasts waste a lot more than a sentence setting context and branding at the begining and end of each segment. This is actually worse for many video podcasts because despite the opportunity video affords to use a simple title card and/or radio-like voiceover to achieve this goal, an enormous number of video podcasts spend too much time (and bandwidth) setting context for each episode (TikiBar TV burned about 10% of their total runtime on intro and outro).

Podcast producers have to set context for each episode, they often do it to excess but they must do it becuase each episode is severable from the context of the “feed”. One can never know reliably which podcast is the first one a user hears and one can’t even be sure that the user got the file via podcast where the RSS metadata could set context. This “setup” is all wasted bandwidth and wasted time for the regular user but all users get the padding.

This padding is the tip of the wasted effort iceberg. Because podcasts are delivered via rss with attachements, a user may subscribe to a feed and listen to only one file. Once subscribed, the user receives each new episode whether they listen to the last one or not. Eventually, iTunes throws a warning and asks whether the user wants to remain subscribed to an unlistened to podcast but how much waste is there?

What percentage of the delivered files are completely unused by the recipient? Since download counts can’t be trusted as anything but an indtcator of mounting bandwidth bills, what, other than new subscription rankings does a podcast publisher have to judge their success or tune their offerings to better meet their audiences’ needs? Producers need feedback both in the form of user comments and metrics. The delivery system of podcasts makes metrics meaningless. iTunes rankings are based on recent subscription activity not active subscriptions or downloads so they too are misleading. When aggregation sites offer previews or, worse yet, re-constitute a feed or the content the creator/server of the feed can’t even wildly optimisitically interpret their download count.

The delivery model is broken.

The delivery model is broken because the content model is almost universally abused. If you’ve ever been responsible for managing server infrastructure, all of the above has you thinking about this already but, in case it’s not obvious, podcasting is a wretchedly wasteful and hard to manage way to deliver media. The files are enormous. Not enormous compared to other web audio and video but positively massive compared to the text, graphics and/or Flash the content most podcasts really should be delivered in. The “subscribe” model means users pull down enourmous wads of data they can’t possibly be actually using. The fact that the pulling of files is done by automated clients means traffic can be oddly spiky (though well designed clients mitgate this). This is fine if you’re a broadcaster and your users are using a DVR to record every episode of The Brady Bunch off air or cable but on the internet (absent workable I.P. multicast [Hah!]) it costs money to move bits. There’s no method for meta-file handling and file types are, for reasons of guarantied compatibility with themost common client applications and the iPod, restricted to .mp3 and .m4v. With steaming (including progressive or “fast start”) media, you can use metafiles to arbitrate different versions of a given piece of content tuned both for your client’s available bandwidth and to manage your own bandwidth usage. Browsers have been matured enough to, usually, leave useful referrer data, user agent strings and handle cookies. Podcast clients don’t do these things to even the fragmented and broken extent that browsers do and even if they did, the screwiness created by all the various aggrators and “portals” replicating the feeds confounds attempt to use the servers to tune bandwidth utilization or optimize for user experience. If a user was offered a download of an audio file from a web site, the producer could offer MP3, AAC, Ogg and WMA versions to the preference of the user. With a podcast client, you’d better have hit a common denominator format and bit rate.

There’s no gold in them thar hills.

All of this would be fine if there was any possibility of a growing value proposition but there just isn’t. Once a content producer has decided to deliver to the technology instead of the user, they’ve foregone opportunities to maximize value. An individual or small organization faces an uphill battle from a content perspective because podcasting has all the requirements of radio but fewer of the benefits. To compete in a marketplace being crowded with recognized brands and high production values the podcaster must reach a high bar. The nature of the medium escalates the cost of delivering advertising but the size of the audience precludes charging more for it. The size of the audience for any given podcast is capped. There are only 24 hours in a day. Assuming users engage with some form of media in some way for 8 hours a day and typical podcasts are 7 minutes long, users who dedicate all 8 hours to consuming podcasts (absurd I know but stay with me here) they can listen to a total of about 68 episodes a day. With any experience in media at all, a producer will recognize that sports, news, sex and pop culture will lead the pack. The wasted downloads we discussed earlier will swamp the net and the disk drives of users. Big Media will displace small players. What chance does any single producer have to reach critical mass in these circumstances? How can small producers get visibility for their podcasts when the space is crowded by Big Media and niche audiences can’t search usefully? From a bang for buck (or bit) analysis does reason support significant investment in this “medium”?

A quick scan of the top 25 podcasts today on iTunes, 3.11.06, shows that less than half are audio. Roughly half of the pure audio podcasts on the top 25 are old media shoveling offerings into the podcast-o-sphere including an audio summary of the day’s top headlines from the New York Times, a couple of different NPR programs, a political chat show from KCRW and the ESPN radio podcast. The Onion Radio News, and a Simpsons audio podcast round out most of the remaining most popular audio podcasts. Given the rise in popularity of video podcasts relative to audio, the lesson of television’s rise over radio is playing itself out again. This is a bit bizzare since traditional radio continues to thrive delivering to use cases where portability (cars, walkpersons, the workplace) and, theoretically, podcasts could be used in the same contexts as radio. Yet, video, once again, seems to have greater mass appeal. Why is this the case? It’s likely that podcasts continue to be consumed by the early adopter seeking novelty and haven’t permeated into the general culture. Were your average mom and pop internet users podcast consumers, audio only content that favored portability or or multi-tasking (listen while you work/drive) use would, likely, have more traction.

Now that Big Media has entered the podcast-o-sphere with a vengeance, the recognizable brands and high production value content is winning the popularity contest. Even the outliers at the top of the popularity pile have significant brands in the alt-internet universe like Homestar Runner and Joe Cartoon. This is inevitble and wouldn’t be a problem except for the ultimate limiting factor of podcasts which is that they have no long tail and they aren’t searchable. This is the showstopper flaw in podcasting as a “medium”.

So… the right question…

Remember… we’re talking about/to professional content producers here… if you want to make a podcast about your summer at band camp, have at it! Your friends and family will love it and, if you discover you’re a talented producer in the process then so much the better. But, no matter what, you and those close to you will enjoy it!

“Should I make a podcast at all and if so, how much should I invest in it?”

That’s the right question and like all right questions there’s no single right answer. There is something of a decision tree process to getting the right answer for a particular producer. What follows is, by no means, an exhaustive exploration but it should be a good foundation.

Start with content. Assuming you have the rights to distribute downloadables ask yourself; “Is my content conducive to a podcast use-case?” If you make a news or public affairs radio show and your ad sales folks think they can sell additional spots for the podcast, making a podcast is likely a no-brainer. If you make a cooking show and you can create short segments that teach useful things. Put the clips on your web site. Build an accessible index page of all the clips. Include transcripts, clear descriptive titles and descriptions on your site for every clip. Offer the clips in formats and contexts that work well in Windows Media Player and QuickTime. When you do the encoding for WiMP and QuickTime, also encode an MP3 if your clips are audio, and an M4V if the clips are video. Package the clips so they stand alone but do it efficiently. A bit of branding at the front, a bit of branding at the back and end on a slate with with URL for your web site. Now you have all the metadata you need for an RSS feed and all the file formats you need so the effort of offering as a podcast is trivial. What they heck? Incremental cost. Why not? If your desire to be in podcast space motivates you to package the clips, great! Don’t let the desire to be in podcast space make you forget the “traditional” web. Puting the clips on your site in a useful way will generate more traffic, be more usable to a broader user base and will allow you to maintain a long tail archive that will make you a destination for content in your genre.

Consider your goals. Are you hoping to use a podcast to promote something else or is the podcast an end in and of itself? If you’re planning to use the podcast to promote something else, stop. Immediately. Think. How much investment will I need to make to produce and promote the podcast so it can promote the thing I really care about? Is it branding? Getting your brand visible is important but what’s the ROI of a podcast relative to increased investment elsewhere? Consider the signal to noise problem. Most podcasts are crap. Will your brand rise above the crap or be burried under it? Can you make a podcast that appeals to the demographic that seems to like the crap? Can you climb up out of the pile of crap and break through or embed within the top crust of “Big Media”?

Consider your costs. To break this down, I’m going to pick on a podcast I actually like. I’m going to pick on them because I think they could have been so much smarter than they are. Why do I care that they could have been smarter? Because if they’re smart, they make money. If they make money, they make more content and I want more content. A 48 second audio podcast from the Onion uses 279k on disk (a wee bit more over the wire due to overhead). The audio file uses eight seconds for intro and branding. The last six seconds are used for an audio advertisement for Chilli’s restaurants. The ad plays after content has very obviously ended. The URL for The Onion is never mentioned nor is a URL for Chilli’s. There’s no way to know how much Chilli’s paid for the spot. If they paid a lot more than they do for a banner ad, they’ve been bushwacked by the podcast hype and more power to the Onion for bilking them. If they paid less than they do for a banner ad, the Onion got hornswaggled because making that podcast cost them a lot more than updating their homepage.

Now let’s look at the Onion home page. As of this writing the homepage downloads 328kb. Before it loads, an interstitial Flash ad is displayed (a subject for another post) and the user can skip (if you must use interstitials, make them skipable, way to go Onion!) or watch the ad and load the Onion home page. The home page loads a banner ad for a web based employment agency, a weird masthead ad for a content site, a Flash ad for a movie, a sidebar ad for the New York Times, another ad at the footer for the same content site plugged in the masthead, approximately 24 sponsored text links, a link to a sample personal ad (a personals service the Onion syndicates) a graphic and link to the Onion store where they sell stuff and a spot in the sidebar to promote the podcast. The page weight for the home page includes data served at the expense of others (ads) and is roughly comparable in size (read bandwidth cost) to 48 seconds of worth podcast.

The podcast has one advertiser and no option for the user to take any action (click a link, dig deeper for more content, click to an affiliate or buy Onion swag). Some significant percentage of the users who download the podcast will never hear it. Every user who loads the Onion’s homepage will see some fraction of it. Spiders will load the homepage and index it. If the Onion’s CMS isn’t too stupid to maintain persistant links and the Onion maintains a content archive, every page on their site will be indexed by search engines and exposed to any user who searches terms relevant to those pages. Users who like the articles will link to them in their emails and blogs. The long tail builds traffic, increases impressions and increases revenue.

Users who like the podcast will email their friends a particularly funny eposide. When this happens, the Onion dodges 279k of bandwidth bill and this is a good thing but, absent context and metadata contained in the rss xml or on a page containing the link to download (get the hint?), there’s very little payoff for the Onion from the secondary recipient of the file. Some users may forward a link to the feed so this ups the subscription rate (and the bandwidth bill) but it is also likely to be unsatisfying for the the second layer user because they may not get the episode their friend was raving about because if the first user didn’t get a chance to listen for a day or two, the recpient dawdles following the link to the rss the episode described by the first user because it may have “scrolled off the feed” Spiders won’t index the content and, unless the Onion maintains an index (they do but it seems only the most recent episodes and the information about each episode is shallow and not especially spider friendly).

So… why exactly is the Onion getting into the audio business? Why, since they decided to do it, are they doing it in such a short sighted and podcast-centric way? What’s the business model?. Most important of all, are they actually delivering value to the user as efficiently as possible?

How do these economics work when instead of 48 seconds and audio-only at 292k we get three and a half minutes and 16 MB of iPod-friendly video?

It’s time to take podcasting out to the back forty and put it out of my misery. Downlodable audio and video are wonderful! One can do amazingly powerful things with them and there are ways to make money doing it. Using rss with attachments to shovel bits down the pipe to users who won’t see/hear them and can’t do much with them when they do is just wasteful except in very special cases. In those special cases, strategic thinking about maintaining an online archive, placing interstial advertisements and being tidy about mentioning URLs in-band are the starting points for not being a clueless n00b.

Wanna see what you should be looking to as the future of internet rich media content models? The Crystal Method: London [9.27.2012 Note: The link preceeding is now dead but the piece had been a really nice ‘Wired QuickTime” interactive piece to promote the album London by “The Crystal Method”. Almost all (aside the media skin) functionality of that piece could be replicated today in HTML 5 or, heaven forfend, a Flash ‘projector’ if one wanted it to be a downloadable piece.]

**** UPDATES: See Bottom of Post *******

Facebook just bought Instagram for one billion dollars. What did they buy? An iOS/Android app that lets you apply ‘very quickly get tacky’ photo filters to your cell phone snaps?

Nah…ugly, hopelessly hipster photo-effects does not a billion dollar business make.

Did they buy a tool that lets users upload images to a free service and share those pics on Facebook and Twitter from their cell phones?

Nah… it’s not that hard to do.  Facebook and Twitter phone apps already support this functionality and that’s not worth one beeeeelyon dollars.

Well if they didn’t buy these two defining features of Instragram what did they pay a billion dollars for? Facebook paid a billion dollars to own more of you. Well, dear readers, I hope not literally you because I hope you, like me, were smart enough not to sign up for Instagram in the first place but a lot of other people did. A lot of people I sincerely respect as technologists did. Even a lot of real photographers I know did and, from day one, it baffled me.

Well, now the bill for the ‘free’ they enjoyed comes due.

Now, Facebook, who many ‘digerati‘  have managed to completely avoid (or, who, like me, regret joining and now try to manage more closely) owns all the information you uploaded to and shared on Instagram. No, they don’t own the copyright to your photos. They don’t even really own the metadata but they own you in the  l337 sense of the word ‘own’ or should I say p0wn you.

They’ve now dominated, defeated, fragged you and made you their… Well you get the idea.

Cameras, your phone cameras included, store  time, technical and often (usually on a phone) location metadata. Metadata is data about data. In this case, data about your pictures. Not just data that could be extrapolated with facial recognition or some other high tech fun, it’s simple and highly revealing data attached to the digital photo. It’s metadata uploaded right alongside the retro-sepia-lomo-shot of  your latest achievement in home canning. The metadata can often be extracted, aggregated and analyzed and, if it included GPS data (and again, on a cell phone camera, it usually does), you’re ‘checking in’ every time time you upload a picture.

All the metadata from GPS data you may have allowed to have stored and uploaded with your photos to whatever contexts you’ve shared them in, to the kinds of content you’re keen to photograph, to when you tend to take pictures to share. All of it. Owned.

The metadata Instagram have uploaded from your phone with your photos. The choices you’ve made about content. The pictures you took of every craft-brewed beer you’ve drunk. All of that is now in Facebook’s hands.

If you use Facebook, if you’ve shared your Instragram pics on Twitter, that’s all correlatable into one more-disturbing-than-you-can-likely-imagine profile of who you are, who you interact with, where you go, when you go there and, given the proclivities people have for the content of cell phone photos, what you eat, drink, smoke or otherwise ingest.

Now all that information is right there in the same mine-able cache of data along with everything you told Facebook about yourself, your friends and your family. And, worse yet, right in the mix along with everything your less than cautious friends might have decided they thought was ok to share about you.

“Big deal Jon you paranoid recluse, get a life!”

Oh yeah? Read this: This Creepy App Isn’t Just Stalking Women Without Their Knowledge, It’s A Wake-Up Call About Facebook Privacy and when you’re done and you say “But Jon, I don’t check in with Foursquare and besides, they took down that nasty stalker app.”. Bzzt… wrong answer. No Winnebago for you.  No autographed picture of Randy Mantooth either. They know more, not less than you think.

A decision or an accident by Facebook that shows this data to anyone able to access your page and a web scrape will make a “Girls Around Me” level of resolution and tracking  easy work for for any serious but average developer. By serious but average I mean the guys who work in your employer’s IT department. The PI your soon to be ex-husband’s lawyer hires. The PI his lawyer uses to build his case in effort to use custody of the kids as a cudgel. The kid who gets mad at you, teacher, for failing his plagiarized term paper. The stalker with resources. Your political opponent. Anyone with, frankly not especially hard to come by, resources who wants to do you harm or who wants to look in a general geographical area for somebody to do harm.

Never forget dear readers, we’d all be much better off if we started thinking of our personal information as currency and our opinions as monetizable content. Even if you’re not worried about any of the above, and the truth is, really most of shouldn’t. All of this makes you more the target for advertisers including charities and political campaigns. It’s more information about you that search engines can use to skew what results you get to help support your preferemces, or preconceptions.

The real lesson here is, a free service to help you share your content isn’t free. It’s costing you every time you use it. Start choosing more carefully. Start taking more control of your data.

 

***UPDATES 4.11.12***

Check out Andy Ihnatko’s piece in the Chicago Sun Times

I’m told by a reliable source that Instagram defaults location data to “off”. I didn’t remember. While I do actually think that’s good behavior, I’d guess many users turn it on because they like the ability to define place as part of what they post. I know there’s a lot of EXIF and other photo metadata to be mined all over the web.  It’s also not an Instagram-only issue. The point above is that you should know what you reveal where.  See this post on Aperture’s lookups though that post isn’t about what you post but rather how Apple looks up location info from GPS coordinates.