New 12-Core Mac Pro Once Again Shows Up in Benchmarks

[LightningIceFox]

The whole explanation doesn't hinge on there being 2 GPUs.

Having memory slots for each CPU is quite a reasonable assumption. I'm not sure if technically speaking they could share the same memory (my gut says that's not possible, but an EE would have to verify). But beyond that, that's part of the advantage of having a dual socket station, is more memory slots.

Finally, as mentioned, You probably wouldn't want to run a QPI link (much different than running PCIe lanes) across boards. They would likely have to be on the same (larger) board.

I'm sure if it were technically as easy as most folks around here seem to think it is, Apple would have offered that option. They wouldn't have anything to lose, after all.

More likely is the fact that within the constraints they've chosen, it isn't technically possible (or perhaps feasible.)

Yes it is technically possible to share memory between processors.

It does take some overhead as far as additional hardware control on the memory (added cost to motherboard), but is feasible.

See this link below for more info:
http://www.netlib.org/utk/papers/advanced-computers/sm-mimd.html

Wikipedia kinda explains it more clearly...
http://en.wikipedia.org/wiki/Shared_memory#In_hardware

 

[mabhatter]

So the bottom line is, even with all this spiffy new hardware, the new 12-core model is not even 18% faster (29721 vs. 25208) than what is an essentially an upgraded Mid-2010 Mac Pro (a.k.a. "Mid-2012") model?

That's why Apple hasn't really rushed to get one out.

This is a SERVER chip, not a desktop chip. Intel really hasn't gotten much faster in YEARS at the top end. They just shrink the die to save power and make more money from fabs but they're not FASTER like they used to be. Combine that with multi-core processing still being an extreme niche... The majority OS has been just plain poor at it so most of the work is in highly custom apps like encoding video, mathematics, etc. there's just not a lot of software that can even warm this system up properly. The Quad-core iMac is about the best your going to get for "commercial" software... so if you aren't writing your own multi-threaded,multi-core supported code this box isn't really for you.

 

[simonmet]

For the price of the 12 core, I would expect better results. But more important than the results is the lack of a swappable GPU. No point in having Firepro GPU for sound engineers or those who don't need it. A standard GPU like an AMD 280 would suffice and bring the cost down. I'd be all over the new Mac Pro if it weren't for Apple's continued stupidity in the graphics card department.

We won't know for sure until release, but the D700s may be BTO for the high end models. In other words you'll hopefully be able to buy an 8 or 12-core model with D500s to keep costs (somewhat) down if you're a sound guy and don't need/want the high-end graphics.

Having said that it's always worth waiting if you can on new generations of Apple hardware. The Mac Pros have always been pretty solid but you never know what issues could arise from such a radically new design.

 

[mrhick01]

Wating 3 years for a mere 16% increase in power is just not acceptable.
I was expecting at least 50%. It doesn't make sense that the base model is slower than 3 year old technology.
The box is cool and all and lots of ports are always welcome but you don't "see" the power, you must feel it.
It's like the old Pontiac Fiero and the Ferrari kit. It looked like one but it under the hood it was still a POS…

Blame Intel for that. Apple would not have anything to do with that.

It has yet to be seen, however, how OpenCL and leveraging the two AMD GPUs in tandem with the CPU can improve performance. I would imagine once the software is in place, tremendous jumps will be seen.

 

[simonmet]

That's why Apple hasn't really rushed to get one out.

This is a SERVER chip, not a desktop chip. Intel really hasn't gotten much faster in YEARS at the top end. They just shrink the die to save power and make more money from fabs but they're not FASTER like they used to be. Combine that with multi-core processing still being an extreme niche... The majority OS has been just plain poor at it so most of the work is in highly custom apps like encoding video, mathematics, etc. there's just not a lot of software that can even warm this system up properly. The Quad-core iMac is about the best your going to get for "commercial" software... so if you aren't writing your own multi-threaded,multi-core supported code this box isn't really for you.

There are plenty of pros out there that have plenty of software (or simply make their own) that can take advantage of as much power as they can get. There is always demand for more power. As a scientist it's not hard to script or batch process your intensive research on large datasets.

The point is that Apple chose to go from dual-socket to single CPU and prioritise graphics instead. If they had gone for a larger design or dual-socket/single GPU option with up to 24 cores then these new Darth Pros would have smashed the previous generation in CPU intensive tasks, but they chose not to.

 

[lilo777]

That's why Apple hasn't really rushed to get one out.

This is a SERVER chip, not a desktop chip. Intel really hasn't gotten much faster in YEARS at the top end. They just shrink the die to save power and make more money from fabs but they're not FASTER like they used to be. Combine that with multi-core processing still being an extreme niche... The majority OS has been just plain poor at it so most of the work is in highly custom apps like encoding video, mathematics, etc. there's just not a lot of software that can even warm this system up properly. The Quad-core iMac is about the best your going to get for "commercial" software... so if you aren't writing your own multi-threaded,multi-core supported code this box isn't really for you.

Stop blaming Intel for Apple's lousy job. Others used same Intel chip and built computers with much higher geekbench score. Here is an example - computer with multi-core score of 54017. Sure, it uses two Intel CPUs (but I bet it does not have a sexy can shape).

 

[handsome pete]

Ah, but is it more cannibalizing just to outsource people with needs to Linux or Windows completely?

This isn't a pro computer. It is a high level consumer computer with some pro Applications. A real pro computer would be expandable. This is not, and it isn't impressive. You can get a more powerful machine for half the price that runs windows from what I can see. This is a major problem for genuine power users.

I think most people trying to defend what Apple is doing here aren't professionals. I don't see any real defense for forcing pros to use consumers. This is different than locking down iOS and making an "iToy"-this is a dramatic "We don't care about your needs" to any content creation. You know, Apple as nothing more than an entertainment provider-the ultimate media company.

Frankly, it'll probably do alright. And I can see Jobs signing off on this-remember where a lot of his money came from (Disney)-but, I think it is a very bad idea because this is the final nail in the coffin after recent problems on things like Final Cut for Apple products to be the products of producers and artisans. It is an easy, convenient to use platform and little else-there's no actual reason I can see to invest in this-none-that you couldn't do on a combination of Linux/Windows, aside from the OS X platform itself.

And that's asking for a $1000 premium for the right to run OS X and get yourself a nifty little container for the package. Again, highly innovative-in consumer space. It isn't cannibalization to make a bigger one of these with expandability, and charging more. I am not interested in an iMac, for instance, but would be interested in this competitively priced. And I'm not gonna shell out for one of these unless they're actually worth it beyond being a nice piece of art.

Please enlighten us with a PC equivalent that is cheaper and more powerful. Building your own doesn't count as that's not who this is targeted for.

And what are you talking about regarding expandability? In all of the shops and studios I've worked in I've rarely seen workstations upgraded over the course of their lives other than maybe some RAM.

I'm not even sold on the system myself, but there's a lot of nonsense being posted around here. I'm waiting for some real world benchmarks and I'll weigh my options then.

 

[Quu]

We won't know for sure until release, but the D700s may be BTO for the high end models. In other words you'll hopefully be able to buy an 8 or 12-core model with D500s to keep costs (somewhat) down if you're a sound guy and don't need/want the high-end graphics.

Having said that it's always worth waiting if you can on new generations of Apple hardware. The Mac Pros have always been pretty solid but you never know what issues could arise from such a radically new design.

Based on the options posted on Apples site you'll be able to upgrade the base configuration from a 4 core to a 12 core CPU and keep the D300's.

The Mac Pro comes in two base configs, one with 4 Core CPU + D300 and one with a Six Core + D500's but you can configure both up to the 12 core CPU.

 

[simonmet]

Based on the options posted on Apples site you'll be able to upgrade the base configuration from a 4 core to a 12 core CPU and keep the D300's.

The Mac Pro comes in two base configs, one with 4 Core CPU + D300 and one with a Six Core + D500's but you can configure both up to the 12 core CPU.

You're right, thanks!

So basically the only thing you can't configure is 2 CPUs and 1 GPU!

 

[helicoil]

As expected, haters gotta hate. How do you know there won't be a BTO option to eliminate the high-end graphics cards? You don't. You're just spouting speculation. Maybe wait until there is an actual product out to complain.

Hardly hate, justified (speculative) criticism i say -but then fanboys have got to erm... fanboy!

 

[throttlemeister]

For the price of the 12 core, I would expect better results. But more important than the results is the lack of a swappable GPU. No point in having Firepro GPU for sound engineers or those who don't need it. A standard GPU like an AMD 280 would suffice and bring the cost down. I'd be all over the new Mac Pro if it weren't for Apple's continued stupidity in the graphics card department.

I do too. It's obvious that it gets it power by the sheer number of cores in that machine, but I find it rather disappointing that the single core performance is actually lower than the single core performance of my Macbook Pro. I mean, the Mac Pro has a Xeon workstation processor; the Macbook Pro a lower power laptop processor. And in my case, 'only' an Ivy Bridge one at that. One would expect the former to be significantly faster.

 

[bigpics]

But there isn't much that Apple can do in regard to single-threaded performance, computing power per (Intel) core simply hasn't increased very much in the last couple of years. And what do you do when you hit a wall, you try to go in a different direction.

A12X, Lolz...

Yeah, not directly responsive, I know. But seriously, how far will Apple take its chip chops in another 5-10 years? You guys have read that Google (Intel's fifth largest customer) is making noises about manufacturing its own server chips, I'm sure. facebook too.

And guess who else already designs chips and runs a server farm or two or twelve with many more to come?

---

PS: A few tendentious convos aside, the most interesting MR thread I've read in many moons. Lots of interesting, insightful and informative contributions.

 

[MacDav]

It's not that hard at all because Intel has a very well defined roadmap.

The situation is as follows: The Mac Pro (the new one) uses the LGA 2011 socket. This socket became available in November 2011. This is an enthusiast, workstation and server socket. It has so far been used for two architecture of chip Sandy Bridge-E and Ivy Bridge-E which come under both Core i7 and XEON branding.

I built my LGA 2011 system in March 2012. A bit after the November 2011 date and I stuck a six core CPU in to it. I could if I wanted to stick the same 12 Core CPU the Mac Pro is going to ship with in my system right now.

Now you might be thinking okay that was 2011 technology surely the 12 core XEON processor doesn't work in all those motherboards released in November 2011? Well actually .. they do. Just stick the chip in to the motherboard and you're done.

This why this is a big deal because with Haswell-E which is releasing in 6 months Intel is making a new socket to replace LGA 2011. This means Haswell-E will not physically fit in to the LGA 2011 socket that the new Mac Pro uses.

This is what I'm talking about with regards to Apple getting in on the end of an architecture. With the old Mac Pro (LGA 1366) Apple got in at the ground floor and some people who bought those systems upgraded the CPU when new architectures came out because LGA 1366 spanned three years just like LGA 2011 has done. But with the new Mac Pro there is no CPU upgrades because its at the end of its life architecturally speaking.

So with the new Haswell-E architecture coming in 6 months time it changes the socket, the CPU architecture and the memory system. It no longer utilises DDR3 memory it now uses DDR4 memory and where in Sandy Bridge-E and Ivy Bridge-E the chips start at 4 and 6 cores for the low and midrange it goes to 6 core and 8 core for low end and midrange.

My whole point I'm trying to make here is the Mac Pro in my opinion missed its boat about a year and a half. It's just too late to ship a brand new system with LGA 2011 now at the prices Apple is talking when it's going to be completely outdated with no upgrade path.

The 2014 Mac Pro probably released in Oct-Dec 2014 will be the one to buy in my opinion, at that point it will still have 2.5 Years worth of chip upgrades coming.

Yes, you can time your purchase if you're building it yourself. I'm not talking about that. I'm talking about trying to time your purchase of CPU's to fit Apple's roadmap. As I said earlier and as you also mentioned it's probably best to wait till the next MacPro release and stay away from the first one. Although are you sure Apple doesn't have the latest Motherboard revisions already? I'll wait to see what happens with this first release and when all the facts are in then i'll decide which way to go.

 

[spaz8]

It's not that hard at all because Intel has a very well defined roadmap.

The situation is as follows: The Mac Pro (the new one) uses the LGA 2011 socket. This socket became available in November 2011. This is an enthusiast, workstation and server socket. It has so far been used for two architecture of chip Sandy Bridge-E and Ivy Bridge-E which come under both Core i7 and XEON branding.

I built my LGA 2011 system in March 2012. A bit after the November 2011 date and I stuck a six core CPU in to it. I could if I wanted to stick the same 12 Core CPU the Mac Pro is going to ship with in my system right now.

Now you might be thinking okay that was 2011 technology surely the 12 core XEON processor doesn't work in all those motherboards released in November 2011? Well actually .. they do. Just stick the chip in to the motherboard and you're done.

This why this is a big deal because with Haswell-E which is releasing in 6 months Intel is making a new socket to replace LGA 2011. This means Haswell-E will not physically fit in to the LGA 2011 socket that the new Mac Pro uses.

This is what I'm talking about with regards to Apple getting in on the end of an architecture. With the old Mac Pro (LGA 1366) Apple got in at the ground floor and some people who bought those systems upgraded the CPU when new architectures came out because LGA 1366 spanned three years just like LGA 2011 has done. But with the new Mac Pro there is no CPU upgrades because its at the end of its life architecturally speaking.

So with the new Haswell-E architecture coming in 6 months time it changes the socket, the CPU architecture and the memory system. It no longer utilises DDR3 memory it now uses DDR4 memory and where in Sandy Bridge-E and Ivy Bridge-E the chips start at 4 and 6 cores for the low and midrange it goes to 6 core and 8 core for low end and midrange.

My whole point I'm trying to make here is the Mac Pro in my opinion missed its boat about a year and a half. It's just too late to ship a brand new system with LGA 2011 now at the prices Apple is talking when it's going to be completely outdated with no upgrade path.

The 2014 Mac Pro probably released in Oct-Dec 2014 will be the one to buy in my opinion, at that point it will still have 2.5 Years worth of chip upgrades coming.

I remember bringing this up a few months back too, the thing is a very small percentage of users upgrade their CPU's. I'd say the Rev 2 of the new mac pro is summer 2015. That's far too long to wait for me, productivity is being lost now. And nobody buying the rev 1 isn't gonna help there be a rev 2 come to fruition. I was not under the impression the Haswell xeon's were nearly that close to release.

 

[Mr. Wonderful]

I remember bringing this up a few months back too, the thing is a very small percentage of users upgrade their CPU's. I'd say the Rev 2 of the new mac pro is summer 2015. That's far too long to wait for me, productivity is being lost now. And nobody buying the rev 1 isn't gonna help there be a rev 2 come to fruition. I was not under the impression the Haswell xeon's were nearly that close to release.

Right. I agree with Quu, but then I also have a feeling it's not going to be until ~mid-2015 before we get the MP refreshed again. No matter what, this is a stupidly powerful machine. DDR4 and HUMA/HSA in the next model will make it even more stupidly powerful, however.

 

[Jibbajabba]

Where do you people get this crap?

Seriously, if I could get paid for just posting nonsense - then I am all up for it. Sounds more fun than my stupid dull and boring IT job It's the internet - who cares what people say

 

[everything-i]

For the price of the 12 core, I would expect better results. But more important than the results is the lack of a swappable GPU. No point in having Firepro GPU for sound engineers or those who don't need it. A standard GPU like an AMD 280 would suffice and bring the cost down. I'd be all over the new Mac Pro if it weren't for Apple's continued stupidity in the graphics card department.

As these are only CPU based bench marks it is not surprising that the score is only marginally better than the previous models. The GPUs are what give these machines their power. Those are very high end GPUs and anything leveraging OpenCL will be extremely fast on these machines. An AMD 280 would give no where near the performance of the D700. Sound processing written in OpenCL would see a huge performance improvement. These machines are leading the field in this respect and it will doubtless be a while before software comes along that can truly leverage the GPU processing power in these.

 

[prowlmedia]

I do wonder why Apple doesn't offer an option to swap out one of the GPU for a second CPU card. I get the OpenCL argument but the reality of today is that some types of code just run better on a CPU then a GPU, compiling code for example. I am a programmer and would much rather prefer 2 CPUs and one GPU for faster builds. I'm not a compiler designer but I am pretty sure dividing up the compiler's tasks across a ton of relatively useless GPUs isn't practical.

Until it appears on the site we don't know 100% what they are selling... A dual core may well come in a few months. The original dual quad cores came later during that generation cycle.

 

[Tech198]

What the.....

What those four way things ??

Looks like like alien invasion.

These Mac's keep getting faster & faster.....

I guess all that hot air must go somewhere huh... so straight up the wind pipe.

 

[cl-user]

Parallel processing

I've like to see an exhaustive listing of applications and software components that take good advantage of multiple cores. Not being a server system, aren't we more geared for interaction than batch grinding.

Thinking where Apple could put on convincing demos, where fast local memory/store really pays dividends. For that matter, given that any professionals that would have need of such a system won't need any arm twisting or shill persuasion, how can Apple use this launch to shower brand allure down the product range (sell more pads and phones in China).

I can't see a real pro being impressed by any black T-shirt introductory master-class, they'll want to see some equivalent to themselves demonstrating how this system allows them to do what they need to do better (so presumably a lot faster). At the same time, not be compromise by unreliability or the fear of support falling between camps wrt external storage, specialist cards, communications.

 

[gibkibonzo]

At Apple's discount on parts, $2000 would be more than they'd pay, and would be more than I'd pay putting together the equivalently spec'ed machine.

So is OS X worth over $1000?

So what you suggest is that Apple should charge exactly what they're paying for the parts? Designing this whole thing (not only in terms of its look but mostly in terms of what its internal structure is) and assembling it (also note it's not some squaky cheap plastic with thousands of screws) is a bonus?

Go back to kindergarden if that's your logic, and stop whining. OS X isn't a must? Get a Windows machine with i7, gamers card and a hybrid drive - what's the big deal?

 

[charlituna]

Indeed. Should be a great machine for its intended audience... but this is MacRumors! Haters, start your hate!

I'm sure the first one will be the lack of an Apple released 4k Cinema Display. Something tells me it won't happen this year. I just hope we get a 1080p with HDMI I can use as a display or a TV for an Apple TV or even blu ray player. Maybe a larger model also. Keep the 27 inch and add a 36/40

Second thing will be when ifixit does their typical 'it can't be repaired' which really means they can't make oodles of profit selling stuff to 'cheap' users.

Third will be some small possible issue with a handful of units made out to be a major flaw. Hyped even more when there are alleged shipping problems

Fourth will be the typical hackintoshes, apple tax etc stuff.

Did I miss anything

 

[charlituna]

The problem is that GeekBench tests only CPU and RAM, not disk speed or GPU.

That's a very important detail that a lot of folks glance over. It's like the iFixit scores. The site makes out that devices etc can't be fixed but when you read closely you see the real issue is that it can't be fixed by common users buying stuff iFixit could sell.

Geekbench is rating this as if the CPU and RAM are the end all and be all of the machine and what will be used. But it's not. This machine is designed for a very small audience which needs that generally hit the GPU as much if not more than the CPU. So you can't exclude those bits of the system. The Mac Pro isn't for the common user but the prosumer using Apple's Pro Apps at the minimum and going up to the Pro using those apps plus After Effects, Maya, Pro Tools and similar.

And you can't really fairly assess the machine without looking at the intended uses and the whole software/hardware package.

----------

Good,hopefully that's it for another 12 months

Get real. That rumors about the next new Mac Pro will start. Possibly even before this one hits the shelf.

Heck we had iPhone 6 rumors and iOS 8 rumors before this years offerings were announced.

 

[milo]

So the bottom line is, even with all this spiffy new hardware, the new 12-core model is not even 18% faster (29721 vs. 25208) than what is an essentially an upgraded Mid-2010 Mac Pro (a.k.a. "Mid-2012") model?

I'm not sure what you're confused about. The fastest 2013 chip from intel is 18% faster than the fastest 2012 chip from intel. It's a drag that there haven't been bigger improvements but there's nothing apple (or any of the PC makers using these same chips) can do about it.

are current DAWs taking advantage of OpenCL? I don't think so, but soon they should.

I keep hearing that audio isn't well suited to OpenCL processing. If that's not correct, Apple can certainly prove that wrong with a Logic update that heavily leverages OpenCL. While I think these are good machines, a lot of the cost goes into CPU that users like audio folks likely won't be able to take advantage of.

My theory all along has been that the dual GPU is less about giving users that power and pushing OpenCL and more because it was the only way they could offer two channels of thunderbolt 2. If they could do dual TB2 with a single card and save some cash, I wish they would sell that as a BTO option.

And I still don't get why this particular case design doesn't have a second socket for pcie SSD on the other GPU card (or even more than two). Seems like a no brainer and it would go a long way toward the complaints about internal expandability.

 

[charlituna]

You were comparing some battery life benchmarks. Apples to oranges man. The fact that your battery benchmarks don't give useful information to the end user doesn't mean that these computer benchmarks also wouldn't.

Unless you are a computer engineer or super geek, specs and benchmarks often don't matter as much to users. Even pro users. My boss himself, who is a rather major figure in VFX, doesn't care if something has six cores at 3.5 or twenty at 1.2. He cares about how quickly and cleanly we can get our work done. If iMacs running connected to a Linux render farm gets him the best results he's fine with that. He'll test this new machine of course but in real world use. Because things like Geekbench just can't tell him what he needs to know.

Not unlike dealing with general users. You can say that a MacBook has a faster processor, more ram etc than an Air or even an iPad. But it might not be the right machine for a particular user. And that user might not know this until they test it in their own 'lab'. Which is why return policies exist.

This is the point he was making and it is valid. Those to whom benchmarks mean something are a small fraction of the world compared to those that just want to know if the job will get done.