New Mac Pro

[Wayne Morellini]

The Re-design and MOQ for parts is not necessarily based on the overall complexity. It's based on the amount of sub components the actual hardware component partners have to purchase. I reference mass produced as the amount of hardware that has to be purchased for a certain price range. Tooling the line is common but not cheap especially when they are systemboard type components that have a completely different structural design. Yes this does translate into overall cost the Manufacturer has to charge Apple to make per unit. They also have to adjust for sub component availability and price fluctuation. A perfect example of this is laptop design and MOQ for a single model. Go to a ODM such as Clevo and request what a MOQ is of a standard laptop production model. I think you would be surprised at the number. Then change the model specifics and engineering completely outside of the normal production structure and that number goes way up. The Mac line for Apple is a very small portion of their overall sales and production. The expected sales on this 1 model is not nearly the numbers of a single model from HP. Therefore the overall cost per unit is much higher. I deal with direct manufacturer engineers and these numbers. They are allot higher than people expect even for a 2 dollar sub component change.

OK Eric,give us a list of these custom components and sub components that are forcing up the price? Apart from In tell chip set, AMD and flash SSD etc, there should not be much more.

If I went to a OEM the price would be high for a laptop, if HP or Apple sent to the right OEM they have or will have regular dealings with the price will be better. But you still don't realise that most if the cost of design are in the individual components, that cist being spread across all buyers in the industry, likely with top buyers paying less of it per unit as they get heavy discounting.

As for numbers, Apple laptops uses to be highest number of individual brands, I don't know what they are now. But take individual numbers, does HP have more models of Laptop then Apple or less, probably not, and likely a number of their models outsell many HP ones. Yet we even see low volume brands not to far off the HP commodity laptop price (but that us another thing to do with firm rebadging models, but still a lot to do with seperate models sharing commodity parts).

Now, as for assembling the small PCB components,big manufacturers have for years automated assembly lines. I don't know if solder wave flow machines are even still used.

Now for the Mac Pro itself, let's be realistic, they are advertising it in cinema sexy sexy style, as a Russian friend of mine would say, who is probably wanting to get one though she doesn't need it, because they know they can sell a million, or millions, if you look at individual laptops that round put their numbers in the top bracket if laptops and top end desktop models. Now take a step back from all this and realise a local computer shop can put together something as powerful using boards with less manufacture volume and likely a nice case with a lot less volume (heaps of case models out there). If Apple can't. or don't want to do better on price that is up to them. But it requires specific very high end and volume experience across the specific sector to resolve this. Maybe it is just the American assembly line pushing things up. I mean, you can buy foreign manufactured parts and put them together locally, not too much extra cost, not dealing with low volumes like on Red products, but you have to pay for the entirely new plant before your new popularity falls off the biol, which means three years.

Eric, I have just woken up sick (now finishing this I am picking up) hardly been able to focus, out of it and not thinking to well, to check for anything significant, to find this. I suggest we give it a break, and.I've gotten diverted checking something in the text and my heads off the boil and I've forgotten what I was going to finish with

 

[M Most]

geeez, wayne, if you don't like Apple's products or their pricing then just don't buy them. But please stop trying to guise their pricing policies as some sort of evil conspiracy, and stop trying to tell everyone else that they have to agree with you. Because I and apparently a lot of other people here don't.

 

[Wayne Morellini]

No, I don't agree with you, for good reason. Now that is moral. Not only that I've even often produced physical examples that disproves the fudge being peddled.

As you go up the scale different dynamics and things are possible. The simple fact I've never met anybody up the volume scale here, tells.

I suggest a chill pill, because I don't want to spend more time proving what could already be read. You are also as well, all welcome to be wrong without trying to convince me of it (I would have thought people stop debating because they can't prove something right). You see interfaces to things but not what is behind things, it is like talking to an applications programmer about how you can design an OS or hardware, they don't see what's behind the api to the bottom of how it all works and generally can't figure out how to do it better. Cinematography forums have been full of people convinced the marketing hype is fact, much as they were convinced that higher camera prices then Red's were truly related to cost. A never ending cycle that will eventually end.

The fact I'm not interested in debating tells of how wrong the debates against reason is. You have to all examine what motivates you, and slow down. I'm motivated by holes in the truth, what is wrong eventually will be wrong, even if I said it. Which is different from being motivated to 'be' right when one feels right, is not t not?

Yes, I'm not buying already, yes 90% of other people are not buying, the only people not getting the message are supporters of Apple, who "demand" the profit. As far as 'evil conspiracy', I didn't say that did I? It is more a matter of business profit, which is controlled by supply demand economics. If people wanted they can choose to only buy at a acceptably similar price and therefore force the price down. Simple business principles done from the consumer side. Now apart from looking like a fancy capsule, it is actually worth a lot less to people on the top end than a regular replaceable card/processor/MB system and should be charged for as such. Wouldn't you agree M?

Thanks M.
Not well today, sorry.

Now I have to get to a job that has come up. If they all want to argue the obvious too, they can sink or swim on their own, I'll wait on the opposite shore.

--------

One of the few things in life that is right is most people are consistently wrong, and it is a waste of time courteously trying to explain things to them, but practically needed if you don't want to ho backwards

 

[Stephen Gentle]

OK Eric,give us a list of these custom components and sub components that are forcing up the price? Apart from In tell chip set, AMD and flash SSD etc, there should not be much more.

What does the price of the components have to do with the price of the product?

You give a product a price not just to cover costs but based on how much value your product gives the user. If people feel that there are benefits that Apple hardware (and more importantly, the software ecosystem that runs on them) give over the competition, and they feel that those benefits are worth a slightly higher price, then they will buy them. Simple as that.

I'm confused as to what your big issue is - if you don't believe that you would get that value out of it, that's fine, but that doesn't mean that nobody else does.

 

[Alex Kiritz]

We still don't have all of the details. Maybe Apple will justify the price by including all of their pro software for free with it.

 

[Andrei Mihailescu]

I understand where you are coming from Jeff, but there are a few points I don't agree with.

I agree with pretty much everything you have said, however I still maintain that there isn't a GPU on the market today that can truly saturate an X8 PCIe v3 slot. The Quadro K6000 is a beautiful example. By all the marketing specs, it should be able to bury an X8 slot in data, even bury an X16 slot and then some. It has an onboard memory bandwidth of 288Gbps or over 33.5GB/s! But it's really not that rosy when you dig deeper. With the 2-bit positioning overhead and ECC the memory, in reality, chokes up far short of that. Onboard memory bandwidth is really more in the range of 24GB/s.

Actually, the theoretical bandwidth of the local memory subsystem on the K6000 is 288 GBytes/second - it's a 6 GHz memory bus of 384-bit width. Internal memory transfers are able to achieve about 95 GB/s sustained in the real world, including the ECC hit, well above what any kind of PCIe bus in existence can handle. This is why workloads that can hold their working set within the GPU local memory can be really, really fast, but as soon as the working set grows beyond the local memory, you can end up waiting for I/O.

Still nothing to sneeze at and yet a DMA-access memory dump of its 12GB onboard memory onto the PCIe bus, as fast as it can unload data, tops out right below 6GB/s as you say. And yet, we have SSD storage arrays connected through X8 on the same systems, tested on the same slots, that can do better, get right up there near the 8GB/s (more like 7.8GB/s max potential bandwidth). Installing the K6000 in an X16 slot yields slightly faster performance, but still below the X8 threshold. For all that onboard bandwidth, the ability to load data on and off the card is still the bottleneck and a tighter one than the PCIe slot. Before the Kepler architecture, Fermi had a huge issue with pushing data back onto the PCIe bus. It just plain sucked at it.

A K6000 in a x16 PCIe 3.0 slot achieves about 12 GBps sustained transfers to and from the main memory. Below the theoretical maximum, but well above what a x8 slot can achieve (around 6 GBps, obviously since the bandwidth is cut in half). Of course, it really depends on how efficient the memory allocation algorithms one uses are (or can be), so there is no guarantee that any one software / algorithm actually can use the practical maximum bandwidth. Whether this has an influence on any one benchmark is highly dependend of course on whether this particular benchmark / application is GPU memory bandwidth constrained. As you noticed, not all are, but there are plenty thereof.

All the bandwidth considerations taken into account here, the data I/O is never my bottleneck. I only shoot video in 4K+ (with the exception of a few GoPro shots) and 90% of what I work with is 5K Epic footage, I shoot at 96fps as often as I can. Most of my corporate work demands that I do this anyway, it's perfect for pulling stills and altering playback speeds. I do tons of post-processing, 3D rendering, etc.. My bottlenecks are always computational. Working in Resolve in 4K is a computational bottleneck, never an issue of data throughput -- we can't process anything fast enough to load up these interfaces. Same with compositing in Nuke. Or pretty much any other apps I work with. Yes, I'm looking at it from my perspective, but as powerful as these GPUs are, they can barely spit back 5K 30fps in real-time just decoding an R3D stream and applying a conventional LUT.

That is kind of the point: you are taking one particular workload that is not GPU-friendly, and certainly so far not very well optimized (wavelet decoding & debayering), and with it you are indeed overloading what current GPUs can handle. If you take out the wavelet decoding out of the equation, you will find that the same GPU can apply LUTs to as many 5k frames as you can realistically feed to it, without breaking a sweat.

In an realtime / faster-than-realtime playback optimized system, you might have a Red Rocket-X doing the decoding / debayering, easily decode 5K in realtime at well above 30 fps, push it to the GPU and get it back with a color transform applied as fast as the Red Rocket-X can serve it. It's really not a coincidence that Red designed the Rocket-X in a x16 mechanical & electrical format, since with Dragon frame buffers it can become bandwidth constrained by the practical bandwidth one can achieve on a x8 bus (and of course far more so on a TB2 port).

A 6K Dragon frame decoded as 48bit RGB (16bits/channel) is 114MB. At 60fps, that's 6.7GB/s, still well within the range of an X8 PCIe v3 slot. The dual PCIe X8 slots in the new Mac Pro have enough combined bandwidth to handle full 6K uncompressed at 16bpc/48bit RGB at over 120fps or 32bit float at over 80fps. Will those GPUs be able to handle that??? Nope. GPUs coming next year? Perhaps. The year after that? Probably.

You will not get a x8 PCIe 3 GPU to do sustained transfers to and from system memory with 6.7 GBps, you will top out at those 6 GBps, and quite likely less depending on exactly what kind of memory allocation routines you use, the word width, the number of threads and so on and so forth. Especially on OpenCL devices. But depending on what exactly you want those two GPUs to do, they can and will easily be able to handle all that and could handle far more if you had the means to feed them the data. Of course, if you expect them to do wavelet decoding on R3D files at those frame rates, then no, they won't manage that, but if you instead want them to do LUTs and/or other color correction transforms, scaling, blending, they will be able to handle it. It's all a question of what kind of workloads you offload to what kind of devices.

And to be fair, the general choice isn't between one CPU and two GPUs on a x8 bus versus two CPUs and one GPU on a x16 bus. You can easily have a (non-Apple) workstation with two CPUs and two GPUs on x16 buses and even two Rocket-X on x16 buses, all at the same time. Such a system is far less likely to be bandwidth constrained than the new Mac Pro, and it would have been technically perfectly possible for Apple to go this way as well, obviously if they had given up on the compact form factor.

The new Mac Pro is certainly a marvel from certain points of views and I don't doubt it will be sufficient for many workflows, but if you are looking for future-proofness, it's certainly not at the forefront. It's kind of loopsided that you can even today get cheap(ish) dual-port 40 GbE / Infiniband to the desktop/workstation, and be able to do network transfers faster than you can upload to and download from any of the GPUs in the new Mac Pro.

 

[Wayne Morellini]

We still don't have all of the details. Maybe Apple will justify the price by including all of their pro software for free with it.

I'm still not here to get baited, but that would be nice.

 

[Wayne Morellini]

What does the price of the components have to do with the price of the product?

Lol Stephen that was a good one.

>You give a product a price not just to cover costs but based on how much value your product gives the user. If people feel that there are benefits that Apple hardware (and more importantly, the software ecosystem that runs on them) give over the competition, and they feel that those benefits are worth a slightly higher price, then they will buy them. Simple as that.

Perceived value, not real value. I'm not saying a little markup, and I am saying for their own survival, they need to not charge one more cent than HP would. I would have a HP today if they had a Z2 4k 32 inch lay down touch with thunderbolt 2/latest display link ports, with 3d compatibility and descent gpu (the processor would be descent). You get quality rather than fancy, and quality features. The only reason to look at a Macpro at a cheaper price like you have in America, is because the lay down 4k+ touch iMac has been delayed, and HP hasn't done a 4k version in their consumer line either. HP would perceive the lack of value of a non carded or upgradable system that is no better than what a upgradable system starts at, and price it down.

The HP z1 has real pro value compared to an IMac for instance, flip that thing and start upgrading. HP, could do a lot more with it unfortunately. If they made it look less industrial and offered software packs to mach the Apple consumer stuff, they could sell a nice looking consumer gaming version.

All the fud people spoke about how great the last IMac was last year has come back and bitten, they are having sales issues according to Mac Rumors. That is what happens when you give a modest update to something that needed even more 4 years ago. Poor Apple. If Steve hadn't got sick, he would be crushing the competition by now, but once he got sick Apple started falling rather than keeping solidly ahead in direction. Now even Samsung wants to crush.

How come no haswell-e chips?

 

[M Most]

Perceived value, not real value.

No, Wayne, for some of us it is very real value. Because as much as you want to talk about hardware specs, the post you're replying to said something you've chosen to completely ignore, and that's the notion of a software ecosystem, along with a controlled and predictable hardware configuration. The fact is that OS X provides an awful lot of things that those of us involved in a heavily IT driven industry happen to value very highly, including (but not limited to) a rich scripting environment with multiple scripting languages directly supported, a rich development environment, a complete set of Unix utilities (including such command line essentials as grep, awk, sed, rsync, and many others), an integrated and free StorNext compatible SAN client, direct compatibility with many scripts developed in Linux, and access to a number of commercial programs not available under Windows (including but not limited to Autodesk Smoke). And all of this supported by a company that is likely to be around and healthy for a number of years to come.

It's not always about numbers and hardware specs. Some of us actually have to get things done with this stuff. And Apple hardware AND software often allows us to do that better, faster, and easier, regardless of whether you happen to believe it.

 

[L. Langer]

Now it's floating around from a third-party developer of display calibration software that the new OS update does in fact bring 10-bit color via Thunderbolt with compatible GPUs. No confirmation from Apple on that but the end-users are not corroborating the existence of 10-bit color and there is no information from the software company as to what the compatible cards are.

 

[Eric Z]

I think the new Mac Pro does support 10bit color. All the pre-requisites are there:

- Pro grade GPU (AMD FirePro) with accompanying driver, which supports the 10bit option.
- ThunderBolt2, which supports DisplayPort signal.


Now you just need a a couple more things:

- Real 10bit monitor (1.07 billion colors) with ThunderBolt or DisplayPort v1.2 port.
- Enable the 10bit color option in the AMD Graphics panel.

 

[Wayne Morellini]

No, Wayne, for some of us it is very real value. Because as much as you want to talk about hardware specs, the post you're replying to said something you've chosen to completely ignore, and that's the notion of a software ecosystem, along with a controlled and predictable hardware configuration. The fact is that OS X provides an awful lot of things that those of us involved in a heavily IT driven industry happen to value very highly, including (but not limited to) a rich scripting environment with multiple scripting languages directly supported, a rich development environment, a complete set of Unix utilities (including such command line essentials as grep, awk, sed, rsync, and many others), an integrated and free StorNext compatible SAN client, direct compatibility with many scripts developed in Linux, and access to a number of commercial programs not available under Windows (including but not limited to Autodesk Smoke). And all of this supported by a company that is likely to be around and healthy for a number of years to come.

It's not always about numbers and hardware specs. Some of us actually have to get things done with this stuff. And Apple hardware AND software often allows us to do that better, faster, and easier, regardless of whether you happen to believe it.

What I'm talking About is what practically could or should be done, these are the areas practical advancement lays in. So yes, they should be able to lower the price and still offer a good ecosystem and product experience, and like the HP z1 etc they could advance their design even further instead.

 

[Brian Merlen]

For me Wintel is a bag of hurt. I spend significantly more time rebuilding Wintel machines at work or for my folks then Mac machines. My Mac needed one rebuild in the past 4 years. Thats pretty darn good as far as maintenance goes I think. Granted I do not like the lack of PCI slots in the new mac pro and probably will not be buying one. Instead I am going to grab a macbook pro retina at some point either this generation or next. I am not in a rush, and my old machine works fine for my personal non commercial needs.

Anyway my point is as someone who uses both I def see advantages on the Mac platform stability wise. My family can't even keep their wintel machines running half the time, they are just useless door stops essentially that I constantly rebuild only for them to destroy immediately lol

 

[L. Langer]

I think the new Mac Pro does support 10bit color. All the pre-requisites are there:

- Pro grade GPU (AMD FirePro) with accompanying driver, which supports the 10bit option.
- ThunderBolt2, which supports DisplayPort signal.


Now you just need a a couple more things:

- Real 10bit monitor (1.07 billion colors) with ThunderBolt or DisplayPort v1.2 port.
- Enable the 10bit color option in the AMD Graphics panel.

The same has been said for years but Apple just never ever implemented it. I heard it was on the table a few years ago but never materialized. Maybe it was because Apple had no 10-bit display available due to lack of 10-bit panels to use for their monitors. That will change with UHD/4K since the panels are 10-bit for the most part.

 

[Hans von Sonntag]

...and access to a number of commercial programs not available under Windows (including but not limited to Autodesk Smoke). And all of this supported by a company that is likely to be around and healthy for a number of years to come.

It's not always about numbers and hardware specs. Some of us actually have to get things done with this stuff. And Apple hardware AND software often allows us to do that better, faster, and easier, regardless of whether you happen to believe it.

We are running a little post-boutique that supports our projects. It must work with clients in the neck, dead-lines, back-ups and has to meet the standards of big international public companies regarding archiving etc...

Therefore we don't care much about specs but a lot about reliability and quick workflow turn-arounds. As most post houses we use MacPros and HP Zs. Since we gave up Linux, because Smoke on Mac solves 99% of all post tasks we encounter, OSX is our main OS and the HPs are now burn nodes. Other companies that deal with more demanding clients than we do still run their flames on the HP Linux boxes. Additional I find the OSX ecosystem very cost-effective because we can run Adobe CS, Resolve and Smoke on the same machine. I cannot do this elsewhere.

Personally, I think the new MacPro will be a good pice of gear as all MacPros have been but won't invest in new machines until I know that all possible hick-ups and bumps are ironed out, earliest after NAB 2014. Reliability is key for us. For private uses this is different. Here you can tinker around and surf on the bleeding edge of technology. Much too risky for my business.

Hans



.

 

[M Most]

What I'm talking About is what practically could or should be done, these are the areas practical advancement lays in. So yes, they should be able to lower the price and still offer a good ecosystem and product experience, and like the HP z1 etc they could advance their design even further instead.

Why should they? Because YOU want them to? Like all successful companies, Apple prices their products based on supply, demand, and a combination of costs including but not limited to design, parts, manufacturing, labor, packaging, advertising, licensing, shipping, inventory, research and development, corporate overhead, potential and anticipated volume of both manufacture and sales, and countless other things. Unlike you, real companies like Apple cannot simply look at a parts list and say, OK, let's take this number and add 10% because that's all we should be allowed to make since we're in the business of "empowerment." They need to be just a bit more sophisticated than that. And since they're doing pretty well, I would judge the pricing to be quite fair, regardless of how particular individuals might feel.

 

[Wayne Morellini]

Why shouldn't they because you don't want them too,? Their desktop sales have been in the toilet for years, this model will change things around out of fashion, and fanboyism, at least temporarily. But in the end they must pay the Piper, and offer better designed value or lower prices. However they decide to setup things, if it is not effective enough it is their responsibility.

So before you say "YOU" in capitals like it is a drawn out wavery accusation, that has been devalued by logic after the last time you used it, you got to think why you are saying it. You might be a loyalist, that is defending the fan owned status, but in a market driven economy rather than a Royalty driven economy, your arguments are extra devalued, ineffective and illogical, Your arguments are not sophisticated enough, throwing up red herrings, there is more to it. So after I had already considered the sophistication of their position, they needed to do more, at least what I said to rebuild desktop sales, but need to do even more as a loss leader to rebuild desktop sales. The days of earning easy foolishly spent money is drawing to the close for them, they are loosing ground in most markets. The big picture is the forest. and the most important parts of the ecosystem.

By the way, my 10% is net, after tax, not gross, which I would have stated, not the insane gross as you accuse. That is the sort of margin basis (10% maybe high, even 3% or lower) that real competitive volume companies, not Apple, try to follow on low priced But you don't seem to know that? Companies that want high margins try to set margins on how much they can get people to pay instead, divorced from costs.

Depending on how well things care done, they are likely doing 10% already, if they are including full version of a list of pro programs as suggested before (not me) then that would be nice (however how many they would do is debatable, as most people won't need every program and that is a lot of lost profits).

 

[jimhare]

Hi Wayne,

Completely understand your perspective, makes perfect sense but I don't think it applies to Apple.

Apple tried to act like other companies in the late 90s and were a few weeks away from bankruptcy when they took back their original approach when Jobs returned.

I don't think acting like other companies would be good for Apple and I will happily pay extra for their products, Mac Pro included.

With the desktop argument, I'm pretty sure the only people to buy Mac Pros will be pros, and for those that have the need for them, price isn't the determining factor.

The iMac is a wonderful example of an amazing device at an affordable price for the masses.

I say Apple should leave their margins high and continue to bring interesting things to the world.

I'm glad they didn't bring out a $99 iPhone, that's not what I want them around for.

 

[Wayne Morellini]

Brian, it is true. Something overlooked is previous articles us the grey parts market. Extra cheap machines might use parts from batches with high defect/low quality rates over reduced periods plus Windows did not make it easier). People had to spend the money to get a higher quality machine that hopefully did not use any of these parts. Still these are cheaper than a macpro. HP pro and high end consumer are ones worth comparing.

 

[Wayne Morellini]

Hi Wayne,

Completely understand your perspective, makes perfect sense but I don't think it applies to Apple.

Apple tried to act like other companies in the late 90s and were a few weeks away from bankruptcy when they took back their original approach when Jobs returned.

I don't think acting like other companies would be good for Apple and I will happily pay extra for their products, Mac Pro included.

With the desktop argument, I'm pretty sure the only people to buy Mac Pros will be pros, and for those that have the need for them, price isn't the determining factor.

The iMac is a wonderful example of an amazing device at an affordable price for the masses.

I say Apple should leave their margins high and continue to bring interesting things to the world.

I'm glad they didn't bring out a $99 iPhone, that's not what I want them around for.

I also think Apple should make top machines with leading features for the higher prices, but they just failed in the 90's and were lacklustre outside of Newton. In truth I'm more of the Steve Jobs in a community focused way, than M is. Just remember, Apple was advertising their macpros in theaters over here, maybe a first, they are obviously going to pickup a lot of non pro sales.