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 2011 "Sandy Bridge E" everything we know
- Thread starter Amethyst
- Start date
- Sort by reaction score
A lot of heavy users of the cluster on campus have done that. It also allows you to customize the cluster to better suite your needs. Though they also still use the university cluster for jobs that will run many hundreds of cores, since its generally cost prohibitive, even for relatively rich labs, to have more than ~100 cores locally (ie 100 cores is probably in the ~$100K range once everything is said and done).
No it isn't. The car analogy is so pointless and overused it's not even worth using.
Not everyone plays games or does 3D graphics work based on a GPU and in that situation the CPU and the bandwidth of I/O are all that matter.
For AUDIO and I mean hosting software synths, streaming multi-track audio to and from the hard drive and handling audio interfaces it's all in the raw CPU speed and throughput, not the GPU or number of PCI-e expansion slots.
A core i5 or i7 based Mac handles:
A Geekbench score of 5839 - 11581 vs 8839 - 22510 at nearly 4 x the cost at the low end!
8 - 16Gb RAM, available for around £200 maximum from third parties, try that with a Mac Pro!
SATA 6 vs SATA 3 - This comes into it's own once SSDs are being used. I have a very fast 7200RPM drive I use for audio recording and general storage, it's one of the fastest 64Mb Cache Caviar Black drives available and won't get any quicker on a faster SATA bus but the SSD I boot from on a SATA 1.5Gb/s interface has already being superceded by drives 3 x as fast that run at half their potential speed in the current Mac Pro.
You can't imagine the difference an SSD makes to booting, virtual memory, loading apps etc... There's a lot of very high end mulit-sampled audio libraries with GBs of orchestral patches that literally REQUIRE an SSD to be used. Being able to run the system off a disk that handles 500Mb/s+ and still has enough bandwidth to run software like that while using a second drive for audio recording is a very real option with any Core i5/i7 based Mac Mini/iMac.
It's not all about frame rates etc... For half the cost of a Mac Pro I could buy a 3.5" Firewire 800 case to put my existing audio drive in, the crossgrade to Protools 10 from LE 8, a 2.3Ghz Mac Mini, a lower flex cable for the SSD and a 16Gb RAM kit. I'd lose just over a 3rd of the CPU power an entry level Mac Pro offers but for half the price, I don't think it really makes DOUBLE the price for such a small increase in raw CPU power worth the outlay!
Still meaningless. You buy a Mac Pro for expansion and longevity. If you buy a MacMini as replacement you have some interesting cable management to enjoy. A MacMini is a macbook without a screen. They have limited memory allotments and relying on Thunderbolt for expansion gets expensive fast.
It totally will work as a lower end DAW. It is a very fast computer for what it is. But what it is is not a Mac Pro or even desktop replacement. No DSP via PCI kills it for anything higher level.
SSD's on SATAII vs. SATAIII really rely on your entire system being compliant to the speeds of the SSD. The fastest SSD's available have 4K r/w that is slower than a 1.5Gb link. How often do you use this type of r/w cycle? 85% of the time. If a SSD claims 550MB/s you only get that to another capable device. Streaming is different but a Mac Po can RAID0 4 WD Blacks and net 400MB/s+ internally on 8TB of data. Whisper quiet and no external case sleeping when you need it most. Like when your DAW crashes from waiting for a sample to spin up in your arrangement. If you feel your cash savings are worth the little details, that is cool and a personal preference. But it is not for everybody. I would gladly take a slower processor in a Mac Pro over any iMac or Macmini with consumer components. My preference.
Why can't you off load to SSD since that's essential RAM and with SATA 6G the bandwidth and latency are pretty good, obviously not as fast as regular RAM but not too far from it.
I can't see Virtual Cores doing much in your situation.
Obviously your tasks/requirements are far different from mine.
Rob
----------
As far as DAWs, one of my MacPro 2006 (one in my sig) has been an awesome unit for use with Logic Pro 9. It has not missed a "beat" even when do heavy surround sound work with all kinds of FX and virtual synths ... if fact my 2006 MacPro has been relegated to exclusive Logic Pro 9 use and it has turn out to be a VERY good and still useful DAW to this day ... and BTW, I'm selling it if anyone is interested.
I have built a few workstations for bioinformatics and definitely sympathize with the RAM problem. Why not use one of the dual socket 2011 motherboards that have between 8 and 16 RAM Channels?
I have an ASUS dual-socket 2011 in front of my right now with only 64GBs, but I could just as easily replace those with 16GB sticks and hit 128GBs. If I switched to the big brother of this motherboard I could hit 256GBs with RAM being the largest component cost by far in the entire build.
I even run OS X on this, but triple boot with Ubuntu and Windows as well.
I just built this for my girlfriend's lab
A few RAIDed SSDs could be a good option to help in some cases, but I also have the issue of needing large storage/scratch space, and I just don't have room with 4 bays to have an SSD and HDD RAID. Now I guess you could cram up 10 drives with those PCI expansion kits and the optical bay, but for now, moving stuff to the cluster that has the 1 TB is probably the best way to go.
----------
I have built a few workstations for bioinformatics and definitely sympathize with the RAM problem. Why not use one of the dual socket 2011 motherboards that have between 8 and 16 RAM Channels?
I have an ASUS dual-socket 2011 in front of my right now with only 64GBs, but I could just as easily replace those with 16GB sticks and hit 128GBs. If I switched to the big brother of this motherboard I could hit 256GBs with RAM being the largest component cost by far in the entire build.
I even run OS X on this, but triple boot with Ubuntu and Windows as well.
I just built this for my girlfriend's lab
If my PI gave me the money, I'd do that too. Though I don't know if I'd care to run Windows. Anyway, my PI likes fully supported systems when spending $7000 on a computer. I can't say I blame him. And while I'd sure have fun building a $7000 computer, I'm not sure it would really help me get my PhD any faster.
I'll be building one on either the SR-X or the Asus board. What's the 64-bit Geekbench 2 and Cinebench 11.5 Score on OS X on the Asus board and what Chameleon installer did you use and which CPUs are you using?
Hi Tutor. I am maleorderbride from insanelymac. I assume you are the same name from there.
I did not do a Geekbench score yet, but Cinebench is just over 26 for dual e5-2687W's. You can only overclock by 5%.
I just used the latest RC2.1 and honestly it is a very easy install. Working audio via AppleHDA took the most time (about 30 hours) for me to modify the driver myself via codec dumps. Granted I could never have done so without standing on other's shoulders (King, MasterChief, & especially Toleda).
For everyone else that will just take minutes as they use what I have already done. As far as I know this is now the fastest computer in the world running OS X
Last edited:
Hi Tutor. I am maleorderbride from insanelymac. I assume you are the same name from there.
I did not do a Geekbench score yet, but Cinebench is just over 26 for dual e5-2687W's. You can only overclock by 5%.
I just used the latest RC2.1 and honestly it is a very easy install. Working audio via AppleHDA took the most time (about 30 hours) for me to modify the driver myself via codec dumps.
For everyone else that will just take minutes as they use what I have already done. As far as I know this is now the fastest computer in the world running OS X
30 Hours?!!! seriously? is your time worth nothing? that 30 hours would easily make the difference between a fake and a real mac for me. I'd be really close to buying a real mac with no time invested, actually the real mac would be cheaper.
Hi Tutor. I am maleorderbride from insanelymac. I assume you are the same name from there.
I did not do a Geekbench score yet, but Cinebench is just over 26 for dual e5-2687W's. You can only overclock by 5%.
I just used the latest RC2.1 and honestly it is a very easy install. Working audio via AppleHDA took the most time (about 30 hours) for me to modify the driver myself via codec dumps.
For everyone else that will just take minutes as they use what I have already done. As far as I know this is now the fastest computer in the world running OS X
Hello again. You assume correctly. Thanks very much for the update and info. Congratulations! That Cinebench 11.5 score of yours tops my 24.7 on my underclocked 5680's. Thus, I assume that it tops my Geekbench 2 score of 40,100. The last time that I was in your neck of the woods was in 2008 for my UCD 30th reunion. Time flies. Next time I'm there I'll look you up. Hasta la vista, baby. Oh, or should I have said, "The second time's a charm."
Last edited:
Some people like to tweak and others don't. If you don't like to tweak, then STOP here! However, if you do tweak or feel you'd like to hear more about the things that tend to interest tweakers, then read on.
The pros of the "real mac pro":
1) Likely flawless if Sandy Bridge E5 real mac pros were to exist.
2) If they come into existence in the future, almost all of them are likely to cost less initially than tl.Mob's (aka maleorderbride's) system referenced in posts #s 681 and 684.
The cons:
1) But they don't exist. So, nobody can buy one because they're not for sale. Maybe in the future they might exist if the Mac Pro line is continued rather than designated End of the Line ("EOL").
2) Probably all (but one) 2010 Mac Pros are less expensive initially and if there are Sandy Bridge E5 real mac pros for sale in the future, the base model(s) will probably be cheaper, but not cheaper for the top of the line similarly configured model, excluding having the same CPUs. Moreover, if there is ever a real mac pro Sandy Bridge E5 with dual 2687W for sale in the Apple store, it'll be the End of Time, which I hope occurs long after the EOL.
3) Thus, undoubtedly Sandy Bridge E5 real macs will not be as fast as the system referenced in posts #s 681 and 684. You will not be able to tweak the clock of the real mac pro one iota, nor otherwise tweak any other performance feature with ease and to a significant degree. But you will, as in the past, be able to buy things to replace things you've most likely already paid an inflated price for in that sticker price for the real mac pro. If you'd like to learn more about the $599 motherboard about which tl.Mob references, see [ http://www.asus.com/Server_Workstation/Workstation_Motherboards/Z9PED8_WS/ ] and by all means download the manual and take a look especially at the section entitled "Ai Tweaker menu" and/or if you'd like to learn more about the CPUs that he references, see [ http://www.cpu-world.com/CPUs/Xeon/Intel-Xeon E5-2687W.html ]. If you'd like to learn about an alternative $599 - $640 motherboard (already announced, but not projected by EVGA to be released until the end of this month) for housing Sandy Bridge E5's see [ http://www.evga.com/articles/00668/ ].
4) The EVGA and Asus motherboards referenced in point no. 4, can also house Ivy Bridge CPUs, with freely released, manufacturer supported bios updates to upgrade to more advanced pin compatible CPUs, unlike systems from another vendor.
In sum, I doubt that you'll ever see a real mac with Sandy Bridge E5s equally as fast as tl.Mob's. The entry level price of a real mac pro system and the cost to swap out the CPUs to install dual 2687Ws ($3770) or 2690s ($4114) would make the upgrade cost excessive. Nor can they be tweaked. You're unlikely to ever see an equally fast, comparably equipped real mac pro for anywhere near the price of tl.Mob's system, including the monetary value of tl.Mob's or your own labor/time. This projection excludes the value to tl.Mob of the experience of modifying a driver - tweakers can get a charge from such accomplishments [kudos to tl.mob], which you won't have to do because he's already done the heavy lifting for you. Thus, there's no need for you to repeat his efforts. All that you have to do is to ask him where do you download it. Here's how I'd do it: tl.Mob, where do I download that audio driver (and Chameleon RC 2.1)? It's that easy now, thanks to tl.mob - scout/tweaker extraordinaire.
The pros of the "real mac pro":
1) Likely flawless if Sandy Bridge E5 real mac pros were to exist.
2) If they come into existence in the future, almost all of them are likely to cost less initially than tl.Mob's (aka maleorderbride's) system referenced in posts #s 681 and 684.
The cons:
1) But they don't exist. So, nobody can buy one because they're not for sale. Maybe in the future they might exist if the Mac Pro line is continued rather than designated End of the Line ("EOL").
2) Probably all (but one) 2010 Mac Pros are less expensive initially and if there are Sandy Bridge E5 real mac pros for sale in the future, the base model(s) will probably be cheaper, but not cheaper for the top of the line similarly configured model, excluding having the same CPUs. Moreover, if there is ever a real mac pro Sandy Bridge E5 with dual 2687W for sale in the Apple store, it'll be the End of Time, which I hope occurs long after the EOL.
3) Thus, undoubtedly Sandy Bridge E5 real macs will not be as fast as the system referenced in posts #s 681 and 684. You will not be able to tweak the clock of the real mac pro one iota, nor otherwise tweak any other performance feature with ease and to a significant degree. But you will, as in the past, be able to buy things to replace things you've most likely already paid an inflated price for in that sticker price for the real mac pro. If you'd like to learn more about the $599 motherboard about which tl.Mob references, see [ http://www.asus.com/Server_Workstation/Workstation_Motherboards/Z9PED8_WS/ ] and by all means download the manual and take a look especially at the section entitled "Ai Tweaker menu" and/or if you'd like to learn more about the CPUs that he references, see [ http://www.cpu-world.com/CPUs/Xeon/Intel-Xeon E5-2687W.html ]. If you'd like to learn about an alternative $599 - $640 motherboard (already announced, but not projected by EVGA to be released until the end of this month) for housing Sandy Bridge E5's see [ http://www.evga.com/articles/00668/ ].
4) The EVGA and Asus motherboards referenced in point no. 4, can also house Ivy Bridge CPUs, with freely released, manufacturer supported bios updates to upgrade to more advanced pin compatible CPUs, unlike systems from another vendor.
In sum, I doubt that you'll ever see a real mac with Sandy Bridge E5s equally as fast as tl.Mob's. The entry level price of a real mac pro system and the cost to swap out the CPUs to install dual 2687Ws ($3770) or 2690s ($4114) would make the upgrade cost excessive. Nor can they be tweaked. You're unlikely to ever see an equally fast, comparably equipped real mac pro for anywhere near the price of tl.Mob's system, including the monetary value of tl.Mob's or your own labor/time. This projection excludes the value to tl.Mob of the experience of modifying a driver - tweakers can get a charge from such accomplishments [kudos to tl.mob], which you won't have to do because he's already done the heavy lifting for you. Thus, there's no need for you to repeat his efforts. All that you have to do is to ask him where do you download it. Here's how I'd do it: tl.Mob, where do I download that audio driver (and Chameleon RC 2.1)? It's that easy now, thanks to tl.mob - scout/tweaker extraordinaire.
Last edited:
For me the cost of switching would be an estimated 200 hours of rewriting my code to access all the different system services through their different APIs*, with the UI being the gnarliest part (plus having to learn a new development environment!) . . . and if you don't know how software development estimates go, well, if you come in double you should feel lucky....
[ * GUI, millisecond timing, screen services, filesys, OpenCL, OpenGL . . . . ]
If MP goes EOL (not likely I think), I'd do my best to stockpile a few 12-cores.
Last edited:
I never have understood why people worry so much over what other people do with their time and money..
If a hackpro, custom bicycle, restored car, or anything else that takes love and time isn't for you great! But please keep contain the need to condescend, no one who likes this stuff will really care anyway.
If a hackpro, custom bicycle, restored car, or anything else that takes love and time isn't for you great! But please keep contain the need to condescend, no one who likes this stuff will really care anyway.
I agree about the bandwidth limitations of Thunderbolt, both as a PCI-e replacement and as SAS alternative. This article proves it with almost a 3 x difference between SAS and Thunderbolt!
http://www.barefeats.com/tbolt01.html
But for audio work and as a boot drive, you've really misunderstood what makes an SSD absolutely smoke a hard drive, RAID or otherwise.
It's the massive amount of IOPs, 10,000s of them compared with all a drive can do which is move the heads around the platters at several MS making the random access time in the several hundred IOPs at most.
That's pathetic compared with an SSD and only useful for streaming low quantities of high bandwidth files.
That's exactly why hard drives score high in MB/s tests on large block sizes in the 1024Kbyte or more range but quickly drop off into single figures once they approach 4Kbyte sizes while SSDs just keep offering almost the entire bandwidth of your SATA interface at all times!
This is why they're such great boot drives and a dual drive system with an SSD to boot from and fast 7200rpm hard drive is the way to go.
I agree 100% about the longevity issue of a tower system but the cost issue is still the losing factor. My G4 has lasted me a long time simply because I've being able to upgrade the CPU and I/O etc... It's too long in the tooth now to be of any use as more than a server and a noisy one at that.
A Mac Mini has far more uses as a second system in the home too because of it's size and low power consumption.
For example I "Could":-
Get a used 2007 Mac Mini for about £250 on eBay and use it as a tie over system while I save up for a i5 based Mac Mini.
Make do with a Firewire 400 external case. One with the appropriate Oxford Semiconductor chipset can be had for under £40.
Buy a Core i5 based Mac Mini. Use Migration Assistant from the Core 2 Duo model to move my Pro Tools LE installation to the i5 system.
(The ONLY way of running it under Lion and it's not even supported, it just works for now and AVID want BIG money for a crossgrade if you want to run any new version of Pro Tools on a Mac running anything higher than Snow Leopard, which is their entire range).
Buy the lower-flex cable, a Vertex 3 and a USB 2.0 universal drive adapter.
Clone the i5 Mac Mini's drive to the Vertex 3 replacing the i5s internal drive with it and fitting my existing SSD in the lower bay with the flex cable to use as my recording drive, completely eliminating the external recording drive issue entirely and then use the Firewire drive for my iTunes library and general storage!
Then I'd just need to...
Use the older Mac Mini model through an LCD TV as a media centre with the addition of a Crystal HD Mini-PCIe card for use with XMBC and buy the Pro Tools crossgrade last, getting myself a nice little media centre into the bargain and a brand new Mac with a good few years use.
Or....
I could look for a used Mac Pro in the price range of the entire setup, notice how much value they lose over time and pick up a 2.8Ghz system with roughly the same CPU power as the 2.3Ghz Core i5 Mac Mini for around £1,100 used.
That would take me close to a year to save up for, I'd forfeit the whole media centre idea altogether because using my old G4 is out of the question because it's just too slow and far too noisy to be on 24/7 and I'd still have to shell out for the Pro Tools cross-grade eventually on top.
Last edited:
^^^ Yes. Sorry to not have listed EVERY benefit of SSD (as there are plenty). Response time, IOPs, and yes, 4K is super important for the majority of OS level stuff. This is not misunderstanding, this is actual understanding. IOP's are nice but faster 4K will feel faster than a couple thousand IOPs faster. Guarantee. I stopped caring about bandwidth years ago as a speed indicator.
I couldn't care less what anyone does with their time and certainly don't worry about it. Just making the point that spending 30 hours to fix something negated most if not all of the savings of using a hackintosh ( in fact, I'd argue that the hack is now far more expensive ). If someone wants to do that for enjoyment, hobby, learning - great. The poster didn't sound like that was the case.
How valuable would you think that effort was if Apple EOL's the MP and users still need to run OS X on cutting edge hardware? I bet there would be a huge interest in DP hackintoshes, and as these things go, the more interest the easier it would become to set one up.
Hi Tutor. I am maleorderbride from insanelymac. I assume you are the same name from there.
I did not do a Geekbench score yet, but Cinebench is just over 26 for dual e5-2687W's. You can only overclock by 5%.
I just used the latest RC2.1 and honestly it is a very easy install. Working audio via AppleHDA took the most time (about 30 hours) for me to modify the driver myself via codec dumps. Granted I could never have done so without standing on other's shoulders (King, MasterChief, & especially Toleda).
For everyone else that will just take minutes as they use what I have already done. As far as I know this is now the fastest computer in the world running OS X
tl.Mob, where can we download that audio driver and Chameleon RC 2.1? Is the onboard ethernet working natively or do we have download a driver for it (or for any other feature) and where can we download them, or did you have to install another compatible ethernet card like many have done for the SR-2?
I think you are right. I don't think the interest will be "huge", but there would definitely be interest. Most who do professional things on a mac pro would probably go elsewhere rather than do a hack. People that get paid for their time just want a machine that works. Having to spend time tweaking kexts, bootloaders, prefs and hardware doesn't earn a paycheck. Creating products and services does.
I agree 100%. If someone wants to have 6 displays...hack their own Mac...buy a MacPro instead of an iMac...why is that anyone else's business?
For the record...I am the third guy in the previous paragraph.
Only small brains question or criticize someone else over such things.
I agree 100%. If someone wants to have 6 displays...hack their own Mac...buy a MacPro instead of an iMac...why is that anyone else's business?
For the record...I am the third guy in the previous paragraph.
Only small brains question or criticize someone else over such things.
Riiiggghht. Isn't that exactly what you have done? Strong people enjoy an opposing point of view.
Except a true point of view would have been..
"I really don't have the spare 30 hours to put into building a mac pro"
Some people de have the spare 30 hrs and pass along the benefit of their work. The Hackintosh community is much like the OS community, rarely is something done solely for the benefit of the individual doing it..
I just finish my HackPro I needed one extra cd and 500mb of stuff down loaded for my vid card and MOBO..
For people that build especially at the level we're talking about, fixing will not take long at all, how could it the person built it in every sense of the word?
No. The difference isn't between SAS and Thunderbolt. The difference is between a PCI-e x4 RAID card and a larger caches, bigger processor, PCI-e x8 RAID card. Not particularly surprisingly if you stick enough high bandwidth drives on a x8 card you get more bandwidth as raise past what x4 can handle with overhead headroom left over. It is exactly what you'd expect if it was two classic format PCI-e cards in a bake-off with those differences in characteristics. Most RAID card vendors have x4 cards which are lower priced than x8 cards. The value difference is throughput.
Thunderbolt just changes the packaging. The PCI-e card tends to be embedded in the device ( Promise Pegasus TB devices ) and makes it external. (yes, there was productivity before wireless keyboards and mice).
Contexts focused upon higher IOPs the difference between 4x and 8x PCI-e is likely not to be that great. At least in the 10's of K range.
The critical issue though is how high the IOPs have to go to be effective for most people. Pointing at the less than 1% that need 1M IOPs for some project really doesn't impact the applicability of Thunderbolt to the rest of the 99%.
1. Pragmatics. When the memory to densities higher than 96GB is prohibitively expensive for most , then most do not install it.
2. The Mach Kernel that OS X ( and iOS ) use use a one page for every physical page. So if physical pages are 4K then another 32GB of memory is an additional substantial increase in the kernel's page table tracking data structures. (~ 8,000 more pages to track. ) These are a bit more problematical since this a structure you probably want to pin in memory (does not page out to disk ever.). So can't just allocate willy nilly and just grow it.
Throw on top the NUMA set up with dual packages with controllers in them and you have something that needs to be tuned to be highly efficient. ( Again it goes back to how many need that kind of work versus kernel updates which impact everyone.
Given Apple's move to 64-bit kernels (which are slightly larger anyway) it wouldn't be surprising to see a boost come with Mountain Lion. It definitely would be a bit tardy in arrival if have to wait till 10.9 to get it. Since there are no more OS X 32 bit kernels, the pressure from iOS to shrink the kernels 'pinned' memory footprint should be reduced. It can serve as a natural 'fork' between those two objectives.