A look at the 2014 Mac mini
blog.macminicolo.net
blog.macminicolo.net
In this case, the Mac mini is the same size as the previous generation, which had had user-replaceable RAM. Soldering it on the RAM in this new generation is just a dick move on Apple's part. It's hard to think of a legitimate reason for them to do this except to make more money from you.
In comparison, the rMBP has a perfect display, but I will need to get the most expensive model. I'd also lose on ethernet support, and need to get an adapter. I am not sure why rMBP lacks an Ethernet port in any case, this is supposed to be a laptop for professionals. Sometimes my job requires me to have Ethernet access.
I honestly don't care about making my devices thinner and thinner. My iPhone 4 is thin enough. My laptop may be heavy but it's more of a desktop replacement anyway.
> I honestly don't care about making my devices thinner
> and thinner. My iPhone 4 is thin enough. My laptop may
> be heavy but it's more of a desktop replacement anyway.
Building a Hackintosh has been hassle free for a good while now. I've been running mine since Lion (2011).
No crashes. Noise and power consumption are dependent on what hardware you pick. I was waiting for the new Minis but seeing them I think I'll just keep my i3, RAM, SDD and get a smaller form factor motherboard and keep using the hackintosh.
But anyway, I don't care about the "increasingly large segment of the market". I care about me. So I don't know why you're fluttering about.
It looks like I am going look for small form factor cases and stick Linux on them and probably get a Mac book Pro when I save up enough money.
I think I prefer the slightly larger mini-PCs from Zotec. If I build from scratch, the AMD-based AM1 platforms are superb. $30 socketed miniITX motherboards and $20 chips are great.
I'm not defending this stance, btw, just stating what I've observed. Our household recently experienced a HDD failure in an iMac purchased late 2010. At 4 years old, the HDD was due for a failure (based on my experience with spinning magnetic disks). No problem, I thought. I'm an experienced hacker; taking this thing apart will be no problem at all.
I was right in that taking the iMac apart would be little challenge for me. Honestly, the iMac is a miracle of serviceability for such a compact computer. All of the connectors unsnapped as expected while using the correct tools. Nothing was terribly inaccessible, with the exception of the LCD driver cable. That connector terrifies me, but everything else was cake.
What killed me was the HDD thermal monitoring system. Apple uses the internal HDD thermal sensing equipment. Ostensibly (and probably true), monitoring via SMART requires an interrupt, which momentarily halts HDD communication. I/O on a HDD is bad enough without these interruptions, so hats off to Apple for prioritizing performance.
The down side of this is that there is no standard for the thermal sensor pins in the connector provided on the HDD. You have to 1) make sure to buy a HDD of the same make as the one you're replacing, and 2) cross your fingers that it isn't a WD HDD, which apparently went through a running change, which A) flipped the thermal sensing pins, and B) changed the connector subtly.
For some context, have a look at this article:
http://blog.macsales.com/19617-diagnosing-2009-2010-imac-fan...
I experienced two challenges. First, my connector had a "nub" on one side, which fit in to a slot in the port on the OEM HDD. The new WD drive had no such slot. Second, my HDD temperature sensor cable had an 8-pin connector. In the article above, you can see that the connector is a 6-pin connector in an 8-pin hole. Because my connector was 8-pins, I couldn't simply flip it, because that shifted the cables by 1 column. I ended up hand filing the nub off the connector, then literally grinding and filing one column of the pins off the connector, turning it in to a 6-pin connector, so that I could flip it.
The moral of the story is that Apple makes zero considerations for end-user serviceability. None whatsoever. I wouldn't go so far as to say that they intentionally make it difficult, just that when faced with any decision where the choices are A) benefit the design or profitability of the device in some way or B) retain end-user serviceability, factor B never even gets considered.
Specially since I value technical features and flexibility to change hardware over design.
This article read really positive despite basically saying the Mini is a terrible purchase. If you want a toy Mac with no concern for performance, most models from the past few years will suit you if you can find a good deal on Craigslist.
I was on the fence because I really wanted a Mini and the new $700 version didn't seem terrible, but now with the knowledge that they put tamper-proof screws on it and bundled it with a 5400rpm HDD ($200 minimum to get an upgrade), I feel like they're not interested in customers like me.
I found an Asus form factor prefab (http://www.amazon.com/Asus-VivoPC-VM40B-02-ASUS-Desktop/dp/B...) that should handle the HTPC/Plex/quiet/price/value minimum requirements I have in case anyone else was looking for alternatives.
About the only remotely positive spin I can put on it is that, relative to the 2012 Mac Mini, the "Apple tax" you pay for 16GB of RAM on the 2014 Mac Mini is somewhat offset by the reduction in the price of the base hardware.
The Mac Mini matters to me, because they make nice servers - companies like MacMiniColo and MacMiniVault will host your Mac Mini for $50 a month or less.
From a pure "computing horsepower per dollar" perspective, it's superior to virtual hosting - try hosting a 200GB database on a virtual host with 16GB of RAM and virtual CPUs anywhere near the performance of those in the Mac Mini. It gets real expensive real fast.
Obviously, virtual hosting has a boatload of other advantages, and it's the right solution for most situations. But Mac Mini hosting fills a niche for small companies with data < 1TB.
Now if you said 1U, 1amp, and 10TB of metered bandwidth, then $50/mo sounds more like it.
OVH pricing gets you pretty close.
If there isn't one, I'll take two! Cheaper than Azure!
Limited upgrade possibilities, tiny storage, poor airflow, not ECC RAM, no redundancy in network interfaces or power supplies, no remote console and you have to bastardise the things to get two disks in.
You can do a lot better for $50/month without the initial up front cost of the machine!
http://www.techradar.com/us/reviews/pc-mac/peripherals/serve...
As for size, unless you're living in an RV, put it under a desk. Just Say No to being hypnotized by shiny rectangles.
Consumers and small businesses can have an affordable server with incremental upgrade to many TBs of storage and 32GB RAM, plus OSS software to take low-latency, surveillance-free advantage of all that computing power and local storage/cache. HP and Lenovo sell similar models. With virtualization, it can run more than one operating system, replacing several boxes.
Why settle for a neutered and overpriced shiny rectangle?
Clearly the Dell is more expandable and potentially more powerful depending on config. 32GB RAM ceiling is fantastic too, especially for virtualization.
Physical size is an issue for some applications but not others. Either one of these would be fine as a workstation, but I know which one of these boxes I'd want in my living room. Size also matters if you're racking these things up, or having one colo'd somewhere.
The Mac Mini kills the Dell in terms of performance/watt, especially when you move past the base CPU models of either the Mini or the T20. That alone can erase the price advantage of the Dell (over several years of ownership) if you're using them as always-on servers and you factor in the cost of providing cooling for them as well. Again, this is an issue in some scenarios but not others.
The next version of the NUc will support Broadwell (Core) and Braswell (Atom) with a much-reduced power envelope that beats the Haswell processors in this "new" Mac Mini. And it will support 4K graphics with integrated GPU.
http://www.extremetech.com/computing/188533-intels-nuc-2-0-l...
I love my apple hardware and depend on my mbp as my primary computing device, but don't see the appeal of mac os X as a server environment. The things that make the mac great (for me) don't add value -- and in fact get in the way. I'm much much happier using Linux on my servers and much much happier NOT using it on my day-to-day machine.
I can imagine there could be some specialized cases in which running services on the mac make sense but I am astonished there are enough that macminicolo can make a business of it (and congrats to them for this by the way!). I assume this is a failure of imagination on my part, thus my question
what does running Mac OS as a server buy you?
Some people run big images/video batch jobs on OSX machines. You can create pretty nice workflows for that stuff with Automator/Applescript. That may grow in popularity a bit now that Yosemite can be scripted with Javascript. I don't know how common that is.
Mac Mini /hardware/ is popular with some people who use them as servers for other operating systems. This is attractive because some companies (MacMiniVault, MacMiniColo) will host Mac Minis /really cheaply/ ($35/month) since they sip power and they can cram 96 Mac Minis into a single server cabinet.
Theoretically this could also be accomplished with compact boxes from Dell or anybody else. There's nothing innately special about the Mac Mini's hardware. But since the form factor of the Mac Mini doesn't change too frequently, a small cottage industry has formed around hosting Mac Minis.
Virtual hosting is great and is probably the right answer most of the time, but Mac Mini hosting fills/filled a nice niche -- virtual hosting gets expensive fast if you need a fast CPU, 16GB of memory, and hundreds of GB of storage.
http://arstechnica.com/apple/2014/07/1099-imac-review-lose-5...
I have a somewhat weird feeling that this is going to happen in a couple of years.
I base that on MacSales.com offering storage upgrades for all 5 generations of MBA.
I wish I could have bought the 2.6GHz quad-core as opposed to the 2.3GHz, but that was a build-to-order option only. Still, the 2.3GHz is much faster than the top end 2014 Mac mini when it comes to multithreaded performance:
http://www.macrumors.com/2014/10/19/mac-mini-2014-benchmark/
You're just paying for the brand and the convenience of not having to put it in yourself. RAM from Apple has to have one of the worst price/performance ratios.
It's one of the reasons I do a lot of my work on a proper machine with ECC and SAS disks.
Due to some duff RAM a few years ago I lost a week of work. It silently corrupted filesystem buffers.
Let's say you're building the next Twitter. It just doesn't matter that much if one out of every million tweets gets a bit flipped now and then. A lot of (most? nearly all?) online services are like that.
Would you notice if Twitter flipped a bit in one or two of your tweets a year? Or if YouTube flipped a bit in a few of the videos you watch every year?
I know, I know, it's gross. We're engineers. Thinking about bits getting flipped on us gives us the chills.
Anyway, that's not really how memory errors work. In aggregate, if you consider all of the memory on the planet as a whole, the bit error rate is extremely low. However, any given chip that's prone to bit errors will have them very frequently. Without ECC you have no way to detect this, and with OEM-provided, soldered-to-motherboard memory chips you also have no way to qualify your memory, either. You'd have to qualify and reject the entire box using something like a periodic memtest86+.
However, any given chip that's prone to bit errors will have them very frequently.
That's not the purpose of ECC. It's not "RAID for memory"; it's not designed to keep you up and running when your memory is physically failing.
ECC "can detect and correct errors of a single bit per 64-bit 'word' (the unit of bus transfer), and detect (but not correct) errors of two bits per 64-bit word." ref: http://en.wikipedia.org/wiki/ECC_memory#Pros_and_cons_of_ECC (yeah, I know... Wikipedia)
A failing ECC memory chip's errors are going to overwhelm the ability of ECC to correct it - they're almost always going to be more than 2 bits per 64 bit word.
Where ECC shines is, as I said, correcting the occasional flipped bit. Which happens way more than any of us would really like to think about - cosmic rays, electrical interference... it happens. Soft errors.
with OEM-provided, soldered-to-motherboard memory chips you also have no way to qualify your memory, either. You'd have to qualify and reject the entire box using something like a periodic memtest86+.