Slashdot Mirror
What Would It Cost To Build a Windows Version of the Pricey New Mac Pro?
zacharye writes "The new Mac Pro is the most powerful and flexible computer Apple has ever created, and it's also extremely expensive — or is it? With a price tag that can climb up around $10,000, Apple's latest enterprise workhorse clearly isn't cheap. For businesses with a need for all that muscle, however, is that steep price justifiable or is there a premium 'Apple tax' that companies will have to pay? Shortly after the new Mac Pro was finally made available for purchase last week, one PC enthusiast set out to answer that question and in order to do so, he asked another one: How much would it cost to build a comparable Windows 8 machine?"
The real comparison comes in how good the machine is at doing what you need it to do. If you're making a movie or doing serious sound editing, video editing, or modeling, this machine and the accompanying software is clearly top-tier, compared to trying to assemble a full workflow yourself that includes the hardware, software, and infrastructure integration. And the fact that you just order it off the shelf and it comes with everything and integrates with everything isn't really priced into this comparison.
E pluribus unum
Unfortunately Apple has a tendency to do weird, non-standard, undocumented things with their hardware configuration, or else I'd be using an Apple laptop myself (without OSX).
See the stuff surrounding the Thunderbolt connector under Linux for an example -- despite, ostensibly, being a standard Thunderbolt port, the Linux implementation doesn't quite work properly with Apple's hardware (hotplug doesn't work, and the OS doesn't even see the Thunderbolt port unless something was plugged in at boot), but works perfectly with the reference Intel hardware. Not to mention their exclusive use of Broadcom wireless cards, the most difficult cards to work with in general (no supported open source drivers unlike the other big two, Atheros and Intel).
Why do I think they ordered those parts from the most expensive sources possible?
Or it could just be the riced up hipster case.
... $9,599 which includes 64GBs of ECC DDR3 memory, a 1TB PCIe SSD, two AMD D700 (W9000) GPUs, and a twelve core Intel Xeon 2.7GHz processor.
While there is nothing really remarkable about this list of parts, it’s the way that they are integrated that provides both pros and cons. On the pro side, you have all this workstation grade hardware in a cylinder that is less than 10 inches tall and under 7 inches wide, with the power supply inside. This makes it very easy to take it on site or pack with you.
Pack with you? Because that's a concern with desktop workstations? I guess you can discount the dual monitor setup if portability is the key? Oh, right, OSX, so you basically have to bring it with you because everyone else is running a different OS and your programs aren't compatible. I don't give half a crap about the story, or I'd go to build the thing online in a tower configuration. Maybe throw in some LEDs, black-light ground effects, a custom body job with clear side panel and glitter+glue monogram too -- You know, really rice it to the next level.
I'll bet this thing just smokes. I've always aimed high when redoing my desktop, back in January I loaded up 32GB of DDR3 RAM, 6 TB RAID V, 250GB SSD for boot and OS space and a 6 core AMD CPU, which is fairly adequate. It has to be as I'll expect it to run for 5 or 6 years before I upgrade again. I built and even beastlier machine for a friend who's doing a lot of media work. It's an absolute screamer, but again, he is expecting it to be competent for the next 5 or 6 years. I build his last one and it motored along well until he decided it was time to upgrade, too. When you spend money, you don't want to do it often.
A feeling of having made the same mistake before: Deja Foobar
This is a business level product.
While you can build one cheaper using DYI parts, however the time spent in wages, for souring the hardware, software and doing the software can add up very quickly
.
Once you've got Windows and drivers installed, you're at a relatively even playing field. Whether you're installing Premiere or Final Cut, you're still stuck doing software installations no matter what you buy.
Then there is also support and maintenance - will having a custom built machine cost more in the long run?
The more you spent on the machine - the bigger the margin for the DYI version - however at the end of the day - is the cost worth it for business?
The crux of the difference - and why the comparison is all but impossible to make - is the fact that you get to truly choose your parts, based on exactly what you need. Entry level Quadro card? $600 or so for one of them. Uncle-Sam-is-picking-up-the-tab model? $5,000 each, I think they support triple SLI.
64GB of ECC RAM? For a handful of use cases, sure. for the vast majority of workstation work? 16 or 32GB can usually suffice, and saves a whole lot of coin.
1TB of SSD? There's that...and then there's a quartet of 256GB SSDs with a spanned partition or RAID-0, possibly with another quartet of 3TByte SATA drives in a RAID5, the latter of which is possible with either no expenditure (depending on the motherboard), or limited expenditure (anywhere from an inexpensive host bus adapter to an IBM or Adaptec RAID controller), which still ends up being less expensive than having to get one of those Thunderbolt drive bay towers that cost twice the price of a half decent SATA RAID controller. Even without that, Thunderbolt drives made by LaCie are nearly double the price of internal Western Digital drives, and you'll still need to shell out $40-$60 for cables.
Super skinny case? Yeah, that's Apple's thing. Cases of every possible shape and size, anywhere from cheap, flimsy aluminum, to completely transparent plexiglass to neon lights to almost fully soundproofed to half a dozen case fans to having room for 13 hard drives or half a dozen Blu-Ray burners? Apple will never have that number of options.
The question of whether it's worth the cost really depends on what the business need is. If the business need is for cubic inches, then the Mac Pro is about the best desktop computing experience you're going to get per square inch. If any higher amount of storage is necessary, the pendulum quickly swings in favor of the PC route. If an optical drive is necessary (yes kids, there are video producers who still give DVDs or Blu-Ray discs to their clients), external drives are invariably more costly and slower than internal drives. If you've got something like a Presonus Firepod or any number of other Firewire peripherals (remember, Firewire was Apple's darling before Thunderbolt, so there's plenty of very expensive add-on gear that uses it), you're adding adapters for those on the Mac side, while plenty of PC motherboards still support it - and if they don't, a PCI(e) card that can support several pieces of hardware costs about the same as a single adapter from Apple.
The way I ultimately figure it is this: If Apple's product, as it ships, fits the bill, get it. No sense in spending time and money for redundant work. If you're looking for even the slightest amount of hardware variation, or you need any meaningful amount of onboard storage, or you can part with just a little bit of performance or the ECCness of its RAM or a nice GeForce card will fit your needs...it's incredibly trivial to avoid parting with that kind of money.
The last time I did the comparison, it came down to deciding which specs were really important to match and which weren't. There are many different versions of various 3ghz processors out there with vastly differing prices.
Some have more cores, some run cooler, some have more on-die memory, or more more threads, or some instruction that may or may not be useful to you. More banks of FPUs. Smaller process size. Bigger process size.
Other components are similarly difficult to compare. RAM has it's speed, bus size, timing - should you get a balanced set? Is that a thing still? Was it ever? ECC?
Anyway, I found when I gathered all the "hidden" specs and priced things out (several years ago) that the macs were actually competitive for the hardware. However, the hidden specs are hidden because the market doesn't really respond to them when they're revealed. Are they irrelevant? If you ignore the hidden specs, you can select hardware that is vastly cheaper, which although isn't the same, maybe is close enough.
Objective comparison is hard to find, and I think part of it is that there are plenty of sites doing hardware comparisons and presenting them in ways that really obscure the difference between the hardware, and there is really no consumer friendly software profiler available on the market.
Doubling your RAM isn't going to help if the programs you run are bottlenecked on loading data off the disk. Adding more L3 cache isn't going to help if your program already fits in what you have or if it spends most of its time waiting on user input. Better sleep/downclock modes would help there. The won't help for high-performance gaming.
How do you really know what you need? Which specs are really relevant?
Can you be Even More Awesome?!
You don't even have to make $200+ an hour. I'm a researcher. Divide my annual salary up by hour and I barely make $50 an hour. But the actual research I do gets funded by grants, which run at about $250k per project year. In those grant applications, I allocate money for computers, and those computers are chosen 100% on their fit for the job, and $10k for a single laptop in the context of a three year $1.5 million grant isn't even an individual line item. 'Fit for the job' is basically 'does it run the software I need/write' followed by 'lowest downtime' followed by 'make my staff the happiest'. Which five years ago meant macs for data collection in the field, and linux on whatever hardware was most appropriate for everything else. Now it means android tablets for the field, and linux on whatever hardware is most appropriate for everything else. If the software I needed only ran on windows I'd buy that too, but the handful of times I've used windows-only software in the last decade the tech support issues have tripled so I avoid it where possible because you lose so much time, and in my field it's become extremely rare to find the only software that does the job you need only runs on windows.
I want to play star trek Armada on the computer that can run Final Cut Pro and Lightroom.
If it can't then it's useless to me.
(yeah, i am comparing platform dependent items ( and hard drive partitioning storage types ) here. and I have no use for Final Cut Pro and Lightroom In fact, I have no idea what Final Cut Pro and Lightroom are..)
_ _ _ Go for the eyes Boo! GO FOR THE EYES!
Weird.
The PSU is only rated for 450W.
Is it powered by magical unicorns on the inside or something? Or is there some kind of lame throttling where you're either using the CPU, the CPU and a single GPU, or both GPUs, but not everything at once?
I don't see how you can implement a lower-level protocol (eg: raw thunderbolt DMA) using a higher-level abstraction of that protocol (eg: pci-e traffic). That's like saying you'll implement Internet-layer frames only using TCP. Similarly, I don't see how you can expose something that doesn't conform to anything remotely like pci-e as a hot plug pci-e device - the latency tolerances to remain in spec are way different for a start.
I too have implemented a driver, from a high-end FPGA to the Mac, and the OSX kernel does not get involved unless you're traversing controllers within that Mac, or the route cannot be expressed within a single transaction, or if the destination is local. It just doesn't. These are to my knowledge the only 3 reasons for the local CPU to get involved:
[1] If you have a machine with devices (1,2,..) on multiple thunderbolt controllers (say A and B), it's possible to have a route like A2 -> A1-> A0 -/-> B0 -> B1, and of course the kernel is involved then because the individual controller chips A and B are not bridged together in any other way. The kernel has to route between A0 (local) and B0 (also local).
[2] The initial spec for thunderbolt allowed a lot of flexibility with source-defined routing tables, but it wasn't taken advantage of, and the later chips from Intel removed some of that functionality (or, more likely, just reassigned the chip real-estate to something more useful). There are now potentially valid routes that can't be expressed within a single frame, and the kernel has to be involved at that point as well, to make sure packets get to their correct destination. It is, however, unlikely that users will see these routing issues in real-world scenarios, you have to have a lot of devices on multiple busses before it's an issue.
[3] The destination is the local machine. Of course, the kernel has to get involved then.
I have a lot of diagnostic code that monitors bandwidth, packet lifetime and routing, and latencies. I've run massive stress tests on multiple machines and devices connected via thunderbolt, and so far, the above 3 reasons are the only ones that an OSX machine enters the kernel for any thunderbolt-related cause. It is quite clear when the kernel does get involved compared to when it doesn't, so I'm confident that if it doesn't have to get involved, there is no interaction.
Physicists get Hadrons!
Yeah, I'd like to spend $3k telling Adobe to use a second GPU. Oh, without actually telling them.
The only thing the Mac Pro is ahead of the curve on is form factor, and they've clearly made serious compromises to achieve that.
450W power supply? That's not a lot - it doesn't sound like you'll be getting maximum performance out of that hardware.
And this is a hardware discussion. Apple's hardware has always been competitively priced - good quality components, decent build quality, generally a good design - but only for the first 2-3 months of its life.
The next 2-3 years it's horribly overpriced, and especially if you live in the UK where Apple charge a significantly higher price than in the US.
Audio editing as well. full 7.1 surround mixes can consume as much ram as a video editing station can.
Then when you need to do Sync you still need the video loaded in memory as well so suddenly you start needing more than the video editors did.
Do not look at laser with remaining good eye.
Remember the applications are going to be OSX applications. So by late 2014 or so the applications will be targeted for your particular hardware configuration and tested against it.
As far as the more general issue, Apple machines tend to be balanced for general use cases in a way that PCs aren't. The result is you often get features you wouldn't have paid for but really love. For example I bought the rMBP for the SSD and the quadcore. The retina screen, which on a PC I wouldn't have gotten however has been by far my favorite feature.