Indeed. Without ECC, you just built a fast machine. Which is fine, but it's not what I'd consider a workstation.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Mac Pro vs. Hackintosh
- Thread starter adddictedtomac
- Start date
- Sort by reaction score
Indeed. Without ECC, you just built a fast machine. Which is fine, but it's not what I'd consider a workstation.
I couldn't care less about ECC RAM. It's no mystery to those of us that use both. I use a MacPro at work, and a Hackintosh at home. The Hackintosh is faster, gets projects done quicker, therefore makes me more money. ECC RAM on the machine at work doesn't mean buttkiss in the above equation.
There's probably some scientific application where it may make a big difference, but with desktop apps doing most tasks you're going to be doing on a Mac (video editing, photo editing, etc.) there's nothing magical about ECC memory. People make it sound like normal fast RAM will corrupt all your data and crash your machine and erase what you're working on every 5 minutes, but of course we all know that's not the case at all.
There's probably some scientific application where it may make a big difference, but with desktop apps doing most tasks you're going to be doing on a Mac (video editing, photo editing, etc.) there's nothing magical about ECC memory. People make it sound like normal fast RAM will corrupt all your data and crash your machine and erase what you're working on every 5 minutes, but of course we all know that's not the case at all.
I use and have used both for as long as ECC has existed.
Presently ECC is only about 2% slower in the worst case. That's from both hard-core benchmarking and manufacturer blurbs.
While it's not as bad as your bottom paragraph describes it actually can be. Where I'm located I probably see a data glitch on average, every month or two. No one can determine for certain if the occasional system crash is due to bit-flipping errors or not tho so I won't even venture to guess - and even the ECC systems do that. I see corruption in data-streams tho and the longer it's in RAM the more apt I am to encounter it. Once glitched it will save out that way too - or crash trying. Also when DL'ing large files via torrent or more direct protocols on a non-ECC machine every so often I have to redownload it because it's corrupt. I never ever see these kinds of errors on my ECC systems. A very long running coincidence? Possibly, I kinda doubt it tho. Especially when industry experts and computer scientists say essentially the same thing regarding ECC vs. Non-ECC system integrity.
Where I lived in the USA previously, it was less often but densities were a lot lower back then too and dem der science types say that matters quite a lot. I guess more than half of noticing this difference is just bothering to notice. I mean if a data glitch is seen and ignored or repaired instead of reloaded (or whatever) then the user might not even think to consider it was due to the lack of ECC. Ya know...?
Presently ECC is only about 2% slower in the worst case. That's from both hard-core benchmarking and manufacturer blurbs.
While it's not as bad as your bottom paragraph describes it actually can be. Where I'm located I probably see a data glitch on average, every month or two. No one can determine for certain if the occasional system crash is due to bit-flipping errors or not tho so I won't even venture to guess - and even the ECC systems do that. I see corruption in data-streams tho and the longer it's in RAM the more apt I am to encounter it. Once glitched it will save out that way too - or crash trying. Also when DL'ing large files via torrent or more direct protocols on a non-ECC machine every so often I have to redownload it because it's corrupt. I never ever see these kinds of errors on my ECC systems. A very long running coincidence? Possibly, I kinda doubt it tho. Especially when industry experts and computer scientists say essentially the same thing regarding ECC vs. Non-ECC system integrity.
Where I lived in the USA previously, it was less often but densities were a lot lower back then too and dem der science types say that matters quite a lot. I guess more than half of noticing this difference is just bothering to notice. I mean if a data glitch is seen and ignored or repaired instead of reloaded (or whatever) then the user might not even think to consider it was due to the lack of ECC. Ya know...?
It does for some of us, because speed isn't everything. Sometimes, getting the right answer matters, and flipped bits are A Bad Thing.
Macs are not specialized video and photo editing machines. Nor are "workstations" in general the sole domain of a particular breed of creative professional. There are myriad scientific, mathematical and financial uses of Macs, uses that take advantage of the combination of a friendly UI, UNIX underpinnings and a long history of (sadly waning) support from Apple where ECC RAM absolutely does matter.
That's great, but doesn't make any difference to my uses.
Mine are, since that's what I do with them. ECC memory isn't a requirement for me to do those tasks. I can't recall the last time I had a system crash or other error, and Final Cut, Photoshop etc. virtually *NEVER* crash, or if they do, it's not necessarily a matter of ECC RAM or not, since I've seen Final Cut crash on my work MacPro a few times over the years.
I said, there are probably scientific uses where it matters. But I (and millions of others) aren't using our machines for that purpose.
I think people are taking things to mean "ECC memory sucks!" I'm not saying it that at all. It's fine for what it is. There are probably uses for it where it matters.
But there's an attempt to try and act as if it's some special requirement for those of us using OSX on hardware without it (there isn't) and that there's something wrong with 'just a fast machine' that doesn't have ECC memory. (There's absolutely NOTHING wrong with 'just a fast machine' that doesn't have ECC memory that gets people's tasks done for them.) Labeling it a machine a 'workstation' or not doesn't matter a hill of beans either.
If ECC RAM was a requirement, I still wouldn't buy a current MacPro, I'd build a Hackintosh with a server motherboard and use ECC memory. That's easily do-able and still in many cases more cost-effective. It simply isn't necessary in many cases.
True, but I think scaling "up" is one of the earmarks of building a workstation machine in contrast to say, a fast gaming machine. Things like ECC memory - you may not needed it, but there's someone who does, and our general "getting **** done" machine includes it. The same goes for PCI slots, lots of RAM, multiple processors/high core counts, etc. No one user is probably going to need all of those, but the point of a workstation is that nearly any serious workload *can* be done.
There's no "Well crap, no Mac Pro for me, because it's lacking X thing".
Hackintosh machines are, of course, another kettle of fish, because they're inherently bespoke machines.
You seem to be dismissing everything that's not a MacPro as some 'gaming machine' and that's where we disagree. The MacPro is great hardware, but not everyone is curing cancer or modeling DNA with theirs. Many people are doing things they could easily do on a much more cost-effective machine- not an iMac or a Mini, but they mythical in-between model that Apple doesn't make.
Those of us building and using Hackintoshes for work are quite happy not to pay a premium for 'workstation' hardware we don't actually need, as well as not paying a premium for 2010 era workstation hardware vs. 2013 era hardware. The pretense that something I don't actually need is somehow a negative in this equation is just a false goalpost in a game us Hack users are not even participating in.
As I've said, I could buy a dual LGA2011 board of newer vintage than Apple's 2010 era LGA 1366 hardware, put two Xeons in it, 512GB of ECC memory (gee, what was the MacPro's max again? 64?) dual GTX680's, install OSX on it and that machine still wouldn't in and of itself get my work done any faster or better. (Although it'd be a hell of a machine. I've built similar for clients with LGA 1366 hardware and 128GB of ECC RAM.)
The reason is, the hardware is not the main bottleneck, it's simply the fact it takes user-time to make video edit-decisions, edit photos, etc.
First, I'm not dismissing. I'm saying I don't consider it a workstation (as a species of computer, not as a 'computer with which work is done'). That's not dismissive, any more than saying a Volkswagen GTI isn't a Ford F-150. I love purely fast machines - if I end up having separate "Work" and "Home" computers, my home computer would likely be something I wouldn't call a workstation, because as you've noted, it's not necessarily cost effective to cover all your bases.
But from my perspective, a "Workstation" needs to be able to handle edge use cases.
I don't disagree with this. What I disagree with is the occasional representation of Hackintosh machines that cut some corners and use consumer-level parts as 'workstations', when they are really fast, rather nice machines that are still utterly unsuited for some tasks.
It's the difference between "My Hackintosh is just as good as a Mac Pro" and "My Hackintosh is just as good as a Mac Pro for my purposes". You can call that a semantic difference, but this forum tends to somewhat marginalize all professional uses for a Mac Pro that aren't audio/video/graphics, so I push back a bit at times. Especially when the initial comment I replied to suggested that because something isn't going to be used as a server, it doesn't need ECC. It's that it *might* not need ECC.
It would get my work done vastly faster, so if you've got a spare, send it my way
Servers and Workstations and Desktops, Oh my!
Well, these definitions are a bizitch - loosely defined and morphing over time.
We have to deal with three general class terms which overlap to a confusing degree: Server, Workstation, and Desktop. Search those terms and you will find as many different definitions as there are authors of them - and they have changed radically over time with the advent of differing form-factors, technologies, and the general march of higher and higher spec. components.
Today I think we have to lose some or most of the performance distinctions which in the very recent past were a big part of the class distinctions. Right, today we have Desktop classed systems with 6-cores running at 4 and 4.5GHz while just a couple of years ago were limited to Workstation Grade systems. Heck, even Hand-Held devices (another class distinction) these days have 4 processors and are running at very high clock speeds.
Currently we seem to be generally limited to just system purpose/deployment and system integrity as the two central distinguishing variants.
System purpose or deployment, is of course subject to implementation. I can use a Server blade as a Desktop for example, just by placing it in the home and installing a wide range of applications commonly found on home PC Desktops. And although it wouldn't be very wise to do so, I could also use Desktop grade systems as Server or Workstation deployments. Sticking to the more classic definitions however:
I think the MacPro was originally engineered to be a Workstation class system but just barely. Apple themselves don't even use enterprise class storage for example and this business of EFI means it often can not be legitimately/officially configured properly for professional Workstation-like deployment. Still, the ECC memory, the Xeon processor system, and the system cooling design do place it as a Workstation. Now whether or not the end user uses it as a Server, Workstation, or Desktop is entirely the purview of the user in question.
Sorry for the long-winded pedantic post but I thought it would be somewhat useful to define some of the terms we're throwing around here.
Well, these definitions are a bizitch - loosely defined and morphing over time.
We have to deal with three general class terms which overlap to a confusing degree: Server, Workstation, and Desktop. Search those terms and you will find as many different definitions as there are authors of them - and they have changed radically over time with the advent of differing form-factors, technologies, and the general march of higher and higher spec. components.
Today I think we have to lose some or most of the performance distinctions which in the very recent past were a big part of the class distinctions. Right, today we have Desktop classed systems with 6-cores running at 4 and 4.5GHz while just a couple of years ago were limited to Workstation Grade systems. Heck, even Hand-Held devices (another class distinction) these days have 4 processors and are running at very high clock speeds.
Currently we seem to be generally limited to just system purpose/deployment and system integrity as the two central distinguishing variants.
System purpose or deployment, is of course subject to implementation. I can use a Server blade as a Desktop for example, just by placing it in the home and installing a wide range of applications commonly found on home PC Desktops. And although it wouldn't be very wise to do so, I could also use Desktop grade systems as Server or Workstation deployments.
Sticking to the more classic definitions however:- Servers are purposed to serving - serving up files, virtual environments to terminal-like Workstations, pooled segmented processing horsepower, and etc. usually located in a non-user-accessible area.
- Workstation purpose is more narrow than Desktop. Workstations are classically designed and configured to just one or two specific purposes like audio engineering, or FX compositing, as a terminal interface, or for video editing, and etc. In fact the term Workstation came as smart independent terminals were developed. Typically, a Workstation is meant to be configured as such, placed in the working environment, and left alone. As such it needs to use higher integrity components much like the Server.
- Desktops are meant to be very user accessible, in the home "desk top" setting, more independent or self-contained if you will, and extremely varied in purpose. Because users can fiddle around with them and reconfigure them so easily/frequently, component integrity is not that much of an issue - they are not designed to sit there and accomplish one specific purpose for weeks and months straight without flaw (even tho of course they can be used that way).
I think the MacPro was originally engineered to be a Workstation class system but just barely. Apple themselves don't even use enterprise class storage for example and this business of EFI means it often can not be legitimately/officially configured properly for professional Workstation-like deployment. Still, the ECC memory, the Xeon processor system, and the system cooling design do place it as a Workstation. Now whether or not the end user uses it as a Server, Workstation, or Desktop is entirely the purview of the user in question.
Sorry for the long-winded pedantic post but I thought it would be somewhat useful to define some of the terms we're throwing around here.
Last edited:
I think 'expandability' just turned into one reasone more to go with a hackitosh?
I mean TB is all nice, but for me it just can't replace SATA, eSATA, PCIe, DVI, DisplayPort and VGA all together.
Not only isn't the performance of e.g. external GPUs not as good, it's also more expensive to buy enclosures for everything. And more of a mess on your desk because you can't throw everything just into one case and put it under your desk...
I am even more glad I build a hack, now that I know the next-gen Mac Pro is nothing for me.
I mean TB is all nice, but for me it just can't replace SATA, eSATA, PCIe, DVI, DisplayPort and VGA all together.
Not only isn't the performance of e.g. external GPUs not as good, it's also more expensive to buy enclosures for everything. And more of a mess on your desk because you can't throw everything just into one case and put it under your desk...
I am even more glad I build a hack, now that I know the next-gen Mac Pro is nothing for me.
In my recent experience with hackintoshes using the best compatible hardware, yes, it runs pretty damn good. But its not without small occasional glitches, and the fear an Apple update will break it.
Part of the Mac experience is that "it just works". There are people that love to tinker with their computers, I'm one of them.. but I also appreciate the experience that comes from running a real Mac as well.
That being said, I was disappointed that my i7-3770k, 16GB ram, GTX 670 and 240GB SSD still geekbenched 46% slower than a 2010 Mac Pro.
Part of the Mac experience is that "it just works". There are people that love to tinker with their computers, I'm one of them.. but I also appreciate the experience that comes from running a real Mac as well.
That being said, I was disappointed that my i7-3770k, 16GB ram, GTX 670 and 240GB SSD still geekbenched 46% slower than a 2010 Mac Pro.
Which MP?
Do you have hyper-threading disabled?
You should be geekbenching around 13000 stock which is a couple thousand more than quad MP's
I was getting around 12000 and Hyperthreading is disabled. My 2008 Mac Pro on my desk here at work also benches at just over 10000. Not a very big difference. 2010 Mac Pro's (three years old now) are benching at over 20000.
re-enable hyper-threading.
The 20K MP's are two 6 core processors 12 cores 24 threads. You have 4 cores and 8 threads.
The first is my quad MP with an upgraded Processor, the second is my Hackintosh with a Sandy quad. My hack smoked my MP
Attachments
I would try again, but at this point I'm simply going to wait for the new rMBP to come out and use that instead and leave my box for Windows gaming. I've always said if you want to play games, use Windows.. if you need to get work done, use a Mac.
Thats cool too, but overclock it. There is no point in buying a k SKU and leaving it stock. Game on it Run OSX on it doesn't matter you built it enjoy it!
Is it that easy to overclock? I read some guides using a Gigabyte and 3770K CPU and it often looked a bit too complicated even for me, but then again, I did not get the GraphicsEnabler part correct for the first two days by not finding how to enter it.
I didn't overclock mine, but depending on the mobo it can be pretty easy.
I ended up buying the K because I knew it would have been tested and passed more stringent quality controls than the non-K variety.
My i7-3770k system uses a GigaByte Z77X-UD5H board, it's used as my windows gaming rig. I had it hackintoshed at one point, but had some oddball issues.
My other machine is an i5-3570k using a GigaByte Z77X-UD3H now running as a 10TB linux file server for home. As a hackintosh, the VIA audio chipset isn't supported.
I ended up buying the K because I knew it would have been tested and passed more stringent quality controls than the non-K variety.
My i7-3770k system uses a GigaByte Z77X-UD5H board, it's used as my windows gaming rig. I had it hackintoshed at one point, but had some oddball issues.
My other machine is an i5-3570k using a GigaByte Z77X-UD3H now running as a 10TB linux file server for home. As a hackintosh, the VIA audio chipset isn't supported.
It takes patience, adjust you turbo freq's adjust your voltage and droop and test if it fails a bit more voltage and droop.
The issue more is if you're running OSX you need to compile and ssdt for your clocks. Thank MacMan for getting you to 4.2 but beyond that you're on your own.
3770's because of the small lithography really should be de-lidded and under water if you want to play over 4.2/4.3 Ghz. The overclockers have not been keen on Ivy and Haswell, de-lidding is PITA and the gains aren't there.
OC'ing takes patience just read a bit and play you really can't break anything if you want some help shoot me a PM OC'ing is strangely fun for me..
I agree, to a point. Even with my MacPro I also had issues when it came to updates and things not working correctly, so that goes with any system. BUT, if you have the right programmer, than anything is workable. After you have whatever programmer is willing to help you to get your system up and running properly, then all you need to do is install all your apps and once every two to three weeks backup your SSD or HD that you use as your primary drive, so you always have a backup.
Just to let you know, I actually found that programmer. Had I found this guy from the beginning I wouldn't have had to go through many, many hours, days and weeks of frustration with asking, hounding and near bothering others for help. I felt embarrassed and also felt like (at times) that I was a burden on those that tried to help me and that did not sit well with me.
Now using this guy, you're up and running within 1 to 2 hours and you're done. With Apple you can only max out your warranty for 3 years then I was on my own. That WAS THE WHOLE REASON WHY I WENT WITH A HACKINTOSH SETUP. Not with him and other guys like him. Here's his contact info if you ever need him:
http://rampagedev.wordpress.com/premium-technical-support/
In fact he and I worked to get my SR-2 system setup perfected to the point now where he actually has separate page dedicated to helping others create the same system that I have set up.
http://rampagedev.wordpress.com/premium-technical-support/sr-2-production-system/
Lastly, you mentioned that your i7-3770K system setup is 46% slower. Well my system (and others like mine) found here:
http://www.insanelymac.com/forum/topic/287857-new-macmod-2013-build-lemon-lime-twist/
is actually faster than the "new" TrashCanMacPro that hasn't even come out yet...
My system's GB score: 24,800+
TrashCanMacPro: 23,000+
You either need to have Rampage help streamline your system to get it to work more efficiently and/or just update your CPU to a 3930K. It's worth spending the extra $500 for it and you might even find a deal on eBay for less and sell your 3770K CPU to offset the price as to not take a major hit on your wallet. In the end you'll only spend about $200.00 to $250.00 more for that CPU and have a much faster system. Lastly, you need to really need to get the 3GB 580 GTX for your setup if you really want the best all around GPU out there for the money. The 580 is far better than the 670. So you may want to consider buying that on eBay and selling your 670 as well. Just a thought and I hope this info helps. Later...
That would be the real world where benchmark testing determines which cpu performs better no matter what the machine is it is included in.
I might go the hackintosh route again if the "nMP" proves to be a disappointment for AE/Cuda workflows - i'd rather pump the $ into the videocard processing than a fancy, limiting case.
I have both and love them both.
I was faced with a dilemma at the start of 2013, my GF needed a Mac to do Photography and Design on and I had a 2008 3,1 Mac Pro Octo 2.8.
I looked into another Mac Pro, no dice waaaay too much even 2nd hand, the MBPr was out of the price range and even a decent iMac with a 27" screen was looking above 2 grand with limited upgrade potential (slim to none and slim just left town).
So I looked at making myself a Hack and letting the GF use the Pro full time.
I found a suitable case:
http://aarghaknot.files.wordpress.com/2010/01/lian_li_pc_v1000z.jpg?w=500&h=647
Put an Asus P67 Sabretooth board in it with an i7 2700k 3.5 GHz and a GTX660ti. Total cost including fan controllers, sickleflow fans, 240GB SSD and 16GB RAM: £1200.
My aim was to make an upgradeable Hack that would be comparatively swift as the real Mac Pro I had.
The Hack runs about 25% quicker but in everyday use both machines are plenty fast enough for all our needs, very happy with the pair of them and looking forward to lots more use from both. I upgraded the Hack with a 4GB Palit Jetstream GTX 770 this month. Noise wise the MP runs at 55 dBA under load and the Hack 58-60 dBA depending on what I'm running, I was aiming for the same noise level but only just missed.
http://browser.primatelabs.com/geekbench2/compare/2301747/1305357
If Apple ever make an affordable, upgradeable machine again I will be tempted to buy it, till then Hack to the future.
I was faced with a dilemma at the start of 2013, my GF needed a Mac to do Photography and Design on and I had a 2008 3,1 Mac Pro Octo 2.8.
I looked into another Mac Pro, no dice waaaay too much even 2nd hand, the MBPr was out of the price range and even a decent iMac with a 27" screen was looking above 2 grand with limited upgrade potential (slim to none and slim just left town).
So I looked at making myself a Hack and letting the GF use the Pro full time.
I found a suitable case:
http://aarghaknot.files.wordpress.com/2010/01/lian_li_pc_v1000z.jpg?w=500&h=647
Put an Asus P67 Sabretooth board in it with an i7 2700k 3.5 GHz and a GTX660ti. Total cost including fan controllers, sickleflow fans, 240GB SSD and 16GB RAM: £1200.
My aim was to make an upgradeable Hack that would be comparatively swift as the real Mac Pro I had.
The Hack runs about 25% quicker but in everyday use both machines are plenty fast enough for all our needs, very happy with the pair of them and looking forward to lots more use from both. I upgraded the Hack with a 4GB Palit Jetstream GTX 770 this month. Noise wise the MP runs at 55 dBA under load and the Hack 58-60 dBA depending on what I'm running, I was aiming for the same noise level but only just missed.
http://browser.primatelabs.com/geekbench2/compare/2301747/1305357
If Apple ever make an affordable, upgradeable machine again I will be tempted to buy it, till then Hack to the future.
since when the benchmarks are called the real world?
go play with your toys and leave the "pro" stuff to the pros