Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Where are you guys getting the idea that there will only be single processor models? They just teased us with a video and threw some specs out there but I don't recall any mention of it only coming in a single processor version. There's still an awful lot we don't know about this box.

Apple is saying "up to 12 cores." That either means dual 6-core cpus or one 12 core.

Given that...

1. Current Sandy Bridge Xeons offer 8 core cpus so it would seem strange for Apple to stick with 6

2. We know there is a 12 core variant coming up

3. We only see one processor in the photos they've released

...I think it's reasonable to assume only single cpu models. That, and the size issue Jason brought up.
 
And this is the center of the debate. Who should Apple build a Mac Pro for?

I'd argue that HPC is nowhere on their list. Many of the things that have been pushed here by one user or another are not on their list. Now that does not imply that those things are not important to that user, but that they are not important to Apple. If for example you want to do HPC, I'd be looking at Linux or Windows HPC. Apple is just not in to that.

HPC clearly isn't on their list, but for a long time, including now, Apple has made excellent client-side machines. While I have no idea what size the market is (though I'd suggest its substantial), the prevalence of Macs in scientific work has been going up in my observation. It's a system that has clean, polished commercial products, but also plays nicely with the command line.

I'd argue that its far easier to create a flexible product when you can, than to try to design with specific users in mind. The current machine, which is arguably designed for the same audience the new one is designed for, is vastly more flexible. I also think its a dangerous approach for Apple, because The User They Design For Now and That Same User Three Years From Now don't necessarily have the same requirements.
 
And this is the center of the debate. Who should Apple build a Mac Pro for?

The debate is over. Apple built the new Mac Pro for Final Cut X users. Of which there are 3. They are driving the wedge between Premier as best they can. They omit Nvidia to kill off CUDA and then build a Mac no one really wants with much less features. Exactly like Final Cut X, no? It's like an Xbox for Final Cut. Of course this is only how it appears. Maybe they will tweak some things and make it really special. They have their entire line to make things pretty. Mac Pro should be purposeful and deep.
 
hey are driving the wedge between Premier as best they can.

Premiere Pro CS6 has supported OpenCL on the Mac since it was introduced last year. On a limited number of AMD cards, yes, but it was easy to work around (just remove the file).

The next version of Premiere Pro, to be released Monday, will have further support for any GPU. It'll use OpenCL or CUDA acceleration for nVidia GPUs and OpenCL for AMDs. Further, it'll (finally?) support multiple GPUs.

So, while I'm not keen on the new design, it'll actually benefit Premiere Pro CC users, if there are any. :)
 
You're not the only one. I'll never get a LaCie again either. Isn't "LaCie" French for SuckyTrash? Pretty sure it is. :D
La Cie drives are just very stylish. They have ZERO thermal design. The "rugged" drives that are so ubiquitous in video production (remember when it was called "Post"?) are worse than the other La Cie. Plastic holds heat in more than steel or aluminum!

----------

It's like an Xbox for Final Cut....
Ouch, that is a stinging comparison!
Too apt!
 
La Cie drives are just very stylish. They have ZERO thermal design. The "rugged" drives that are so ubiquitous in video production (remember when it was called "Post"?) are worse than the other La Cie. Plastic holds heat in more than steel or aluminum!


What are you talking about? The rubber sleeve on those drives covers probably an eighth of the surface area. So heat dissipation is not a problem in that regard.

I must just be lucky with the rugged drives I've used (knock on wood). A friend of mine who does a ton of stuff for Nat Geo uses them a lot without issue as well.

I'm not saying there aren't issues. I just haven't experienced it first hand.
 
I don't seem to get your point. Are you saying I'm wrong and Firewire connected through Thunderbolt is inferior?

For professional audio recording, yes. A pure FW to FW connection is ideal, as TB to FW introduces additional latency. When it comes to recording, having the least amount of latency is paramount. The best connection latency-wise is through PCI-E, actually. Unfortunately, that is not possible with the cylinder.
 
Also, the Ivy Bridge E5 Xeons go up to 12 cores on a single chip, the same as the old Mac Pro had with two chips.

Something people don't realize is 12 cores on one chip != 12 cores on two chips.

Because it only has one CPU, the new Mac Pro only has 4 RAM slots versus 8 RAM slots in the existing Mac Pro. This makes it very hard to get over 16GB RAM because you'll have to buy really expensive 8GB sticks.

Similarly, memory bandwidth will be effectively cut in half with only 4 channels instead of 8
 
Because it only has one CPU, the new Mac Pro only has 4 RAM slots versus 8 RAM slots in the existing Mac Pro. This makes it very hard to get over 16GB RAM because you'll have to buy really expensive 8GB sticks.

Just because memory modules with a higher density cost more than memory modules with a lower density does not make it harder to get over 16GB of RAM. Simply buy and install the more dense modules.
 
Something people don't realize is 12 cores on one chip != 12 cores on two chips.

Because it only has one CPU, the new Mac Pro only has 4 RAM slots versus 8 RAM slots in the existing Mac Pro. This makes it very hard to get over 16GB RAM because you'll have to buy really expensive 8GB sticks.

Similarly, memory bandwidth will be effectively cut in half with only 4 channels instead of 8

I guess you haven't priced 8GB DIMMs recently? I've now heard more than a few people making this claim but it just simply is NOT the case.

$300 to $350 for 4x8GB (32GB) of RAM is not expensive. Anyone making this claim is kinda out to lunch IMHO! It's actually cheaper (or about the same is some cases) than an 8x4GB configuration at present. Four years ago you might have been right but not any longer - not for a very long while.

Don't believe me, search it yourself of just click here (and stay AWAY from rip-off venders who advertise "Apple specific" memory.):
http://www.ebay.com/sch/i.html?_odk...70.l1313.TR0.TRC0&_nkw=4x8GB+ecc&_sacat=58058

Also I think you're mistaken about channels. MP5,1 had triple-channel access not 8-channel. The new MP6,1 actually has quad-channel access and thus faster, not slower.
 
Last edited:
Unbiased CPU rendering

The design is excellent IMHO. This thread currently is most relevant to me as I render unbiased on the CPU.

The design can also scale in volume pretty easily. The internal triangle could become a square with 2 CPU's or even a hexagon with quad e7 chips. I don't know if the thermal calculations will scale as easily.

Right now I'm hoping my software updates to support open cl (maxwell render) this is my best case scenario I could access the 7 teraflops. Otherwise my upgrade will be on the simple calculation of benchmark/price. I am swayed by running an all osx system which loads the calculation.

Ultimately the BTO and upgrade specs will have a hand to play. Perhaps although its a long shot integrated graphics could be specced with the 12 core thus keeping it affordable as a node.

Future proofing has produced a gap right now while we wait for software & external hardware to adapt. I assume there is no tb2 external storage available right now ? I also have confidence that apple know what they are doing in the long run. Perhaps multiple processors is inefficient in economies of scale. I know when I spec a 4x10 core e7 tower it comes to 18k UKP.

These are interesting times and I believe that the Mac Pro will be a success for apple although not currently guaranteed to be a success to me.

Open CL and grand central dispatch have been available for some time now so it should be maturing. This machine with hindsight was actually predictable in its configuration.

I've always used FW800 for external drives on my current MP.These have been fine but at some point i feel it would be good to consolidate them into an array i'll wait to see what 3rd parties bring to the table. I always understood the FW was superior because of it's constant speed rather than the bursts of USB, i also assume that thunderbolt will not have bursts of speeds but be constant. I also understood that USB used cpu cycles and hogged the system while FW had it's own chip controller.This i also assume will apply to TB2 ?

I also steer clear of lacie drives now and use OWC enclosures. Some years ago my relatively new 1tb lacie big disk died. This drive contained my entire itunes library which i had created by recording meticulously my vinyl collection scans info the lot. I paid someone to recover them all, it went well until i realised i now had 100,000+ untitled tracks. Needless to say i remained calm and re recorded nearly all of it until now.

I'd like to hear what other OSX native 3d artists have to say on the subject.

On a lighter note I hope someone makes a thunderbolt powered harness to mount double MP's jet pack style.

http://www.youtube.com/watch?v=4K8zz9eI4-8
 

Attachments

  • THUNDERBALL.jpg
    THUNDERBALL.jpg
    304.8 KB · Views: 64
Last edited:
I am okay with the single processor. For my needs it will work great. I have had a Mac Pro in the past and now a Mac Mini. I will enjoy this Mac Mini Pro for my photo work and for serving up media files.

It is rather amusing that given that it has really no upgradable guts, it is with all its connectivity nothing more than a Mac Pro turned inside out. It should be fun little machine.
 
Right now I'm hoping my software updates to support open cl (maxwell render)

Here's hoping for Maxwell / open cl, too. They did announce it somewhat if I remember correctly, opposed to CUDA support. I think for v2, which I luckily own. Intersting as a node, theoretically yes. And I also don't understand why everybody assumes that we won't see any 2-proc solutions in the future anymore. If anything, this redesign shows that Apple is still interested in it's Mac section.
 
Mr

I remember a couple of updates back, the mac pro featured maxwell benchmarks on apples site showing the improvements made. I believe next limit have answered this request before but they found the GPU to be inaccurate in calculations. Perhaps Open CL will circumvent this somehow.The only other option is the xeon phi route at a long shot.

My best case scenario is a special event launch with autodesk showcasing 3ds max native on the mac with next limit also making a demo of the quantum performance gains !

i wonder how 7 teraflops compares to my 2.93x12 setup ?
 
swapped one CPU for a GPGPU

Apple have made the design decision to swap a CPU for a GPGPU.

For software that is suitable for GPGPU processing (one data set with lots of parallel processes) this will give a big speed up.

BUT for many users (like myself) it has significant drawbacks.

1.) Many tasks are multi-data such as running several copies of a program on different data sets and these aren't suited to GPGPU speedups.

2.) For development software, academic experimental software etc the cost in run time is a sum total of development/experimentation and actual processor time and it is far too much of an overhead to write the code for efficient GPGPU speedup. So the GPGPU won't be used (the exception might be if good library code is available.)

3.) Even for commercial code suitable for GPGPU speedup much of this has been written for CUDA or openCL on nVidia which doesn't run properly on ATI hardware so users of such software will need to rely on it being rewritten. This will not happen for more than the most popular programs. Maybe not even for some of those - if most customers run it on Windows workstations with nVidia cards it won't make commercial sense to write a particular Apple version.

4.) If you're not making good use of the GPGPU it is annoying to have to pay for it!

Apple is trying to force developers/users into a software model that is optimised for Apple hardware. If they succeed then the new Mac Pro will be a powerful workstation for such use. If they don't succeed then much of its power will be restricted to a few programs which may become too small in terms of market share to be properly supported with further development long term.
 
The design is excellent IMHO. This thread currently is most relevant to me as I render unbiased on the CPU.

I'd like to hear what other OSX native 3d artists have to say on the subject.

I guess 90% of the CG industry renders 95% of their frames using only CPU. GPU rendering is primarily for GUI and previews. The minor percentages mentioned are primarily single layer affects comped in over the larger fuller plates. This seems like it has the immediate potential of changing with time however as more and more people are enamored by and therefor in want of GPU based rendering. Will it happen and when remains a question to me. I'm not sure. And if/when it does happen what hardware will we need? Will 2 GPUs be enough? Will we want 4, 8, 32 GPUs? With the current MacPro the GPUs being used can do a whole lot in real-time. Will that be enough? Will we want faster than real-time or massively heavier FX? All good questions and all have coefficient dependancies mostly to do with business models and production scales.

Typically render farms and smaller render gardens are employed to handle the compute volume which no single average-sized box could. So why would GPU rendering be any different? If 128 GPUs are needed in the future why would we think they would all be in one system. As the CG industry has matured we have employed the division of labor rule to good measure with artist workstations being very well defined within its division. The industry including most or all application developers share general system expectations and exploit system resources realistically within those bounds for the various task divisions. The MacPro6,1 seems to sit perfectly in the current definition of "artist workstation" even though the new MacPro attempts to redefine that definition somewhat - the definition of "Workstation" that is.

A lot of course will depend on the system price point - which of course no one knows yet. My hunch based on the system fabrication and assembly methods seemingly employed as well as the parts and materials selected slash excluded, is that Apple will attempt to compete favorably in the area of user expense - initial cost - sticker price. If I'm not mistaken this will result in the MacPro becoming a CG industry sweetheart.

There are a few divisions within the CGI electronic labor force I don't see much of a place for the new MacPro in. For example the default inclusion of dual high-end GPUs which Apple said all MacPro6,1 machines will be equipped with, may exclude the machine for consideration as render-nodes in beowulf style render-farms for reasons of unit cost and power consumption. And so on and so forth.

IMO, of course depending on the price-tag, I think the new system is almost perfectly suited for CGI boutiques and Artist Workstation deployments. As presented at WWDC it may be a little overkill for CG students but this could be addressed by offering a slightly lower spec configuration. Before WWDC I was considering to build a Hackintosh and had concluded a six 3.5GHz core, modern (670 or 570) GPU, and 32GB of RAM would be perfect for my purposes and needs - which are very similar to those of a present-day CGI student. :)
 
Last edited:
Out of curiosity, what do people do with 24 processor cores? I remember a documentary about South Park where they said the entire show is animated and rendered on regular iMacs. 4K video editing can put a lot of stress on a computer but it looks like the new Pro was designed specifically for that.

They put dual GPUs. Nowadays, GPUS commuting I sued for this.
 
Something people don't realize is 12 cores on one chip != 12 cores on two chips.

Because it only has one CPU, the new Mac Pro only has 4 RAM slots versus 8 RAM slots in the existing Mac Pro. This makes it very hard to get over 16GB RAM because you'll have to buy really expensive 8GB sticks.

Similarly, memory bandwidth will be effectively cut in half with only 4 channels instead of 8

Superbiiz are one of the cheapest ECC DIMM retailers on the internet, here are their cheapest prices for ECC 1600MHz DIMMs:

2GB ECC - $11.50 per GB
4GB ECC - $10.00 per GB
8GB ECC - $8.63 per GB
16GB ECC - $8.19 per GB

So in fact the densest configuration of 4x16GB is the cheapest per GB.

Now you can get 32GB DIMMs for a 128GB config, but they are $17 per GB for slower 1333MHz or three times the price for 1600MHz, so that is a disappointment for some, but most don't use more than 64GB.
 
Maybe Apple will have a secondary processor or rendering module? Rendering nodes?
 
Hmm

I think it's a long shot Jester that they make something specific. Would thunderbolt be usefull to connect another machine ? Or is ethernet still the best way ? I noticed they retained the dual ethernet ports i presume machines could be daisy chained this way ?
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.