Next MacPro: Predictions...

[JanneJansson]

I like the idea of smaller mac pro, but I don't like the idea of limiting the performance to favor the design. Mac pro is about pure POWER. For those who want a cute computer, there is imac. I'm also not liking the idea of having more cables on my desk. Better make a box and put all add-ons inside. ..like the current design

 

[Sanjin Jukic]

Guys,

you don't need to discuss and waste your time with it just because Apple and also other PC vendors with many partners involved already tested hardware and they made their product line decision for 2011.

Also you as a consumer just have to wait, see and decide what is the best for you.

 

[Theo Valich]

I've been reading this thread and while it is quite visionairy, unfortunately there are significant practical difficulties which render most of the discussion here - a vision and nothing else. Don't get me wrong, I would really-really-really like for most of this to come true, but there are two problematic issues called: Standards and the Laws of Physics.

First and foremost, everyone who's been following the computer industry has seen two particular trends: high end gets beefier and beefier until it reaches physical limits, and then the high-end part gets miniaturized to the point of a low-power, mass production silicon. There is no way around this, unfortunately. You design a part with 512 cores sectored in 32 cores and in the end, you have a $500 part and an $30 part, all based on the same architecture - only with less transistors and a dramatically smaller power draw.

Can you put a high-end graphics card and a powerful CPU onto the same motherboard? The answer is quite simple (and sad): No. The reason for it is the amount of layers which would be required to put these two together, and just the idea of signaling becomes an engineering nightmare. GPUs are already 10-12 layer PCB (Printed Circuit Board), motherboards are 6-8. Move a motherboard of a decent size to 10-12 layer design, ad complexity here and there... we'd approach the raw cost of the motherboard PCB to a $80-100 range. Without any components on it.

Could you connect the CPU onto the Lightpeak (or just about any other) peripheral standard? For data, yes - but what about power and signal stability? The major issue with the silicon of today is that we take the numbers for granted - 45nm, 40nm, 32nm, 28nm - all is nice and engineers that create silicon are masters at their work, but the sad reality is that as the technology progresses, it becomes more frail and requires much more power to operate - something which engineering architects are already calling "dark silicon".

In order for a transistor inside a CPU/GPU to work, you need serious current, in the 85-105-110-120 Amper range. By following the laws of physics, at this current a human could potentially receive a deadly shock with miniscule Voltage. Thus, the idea of shifting silicon parts with 1-3 billion transistors around and attaching them using a standard that is supposed to replace the USB is just not a physical reality. PCIe as a standard has a lot of issues moving from a motherboard to a cable and reaching greater lenghts, yet alone something unproven as LightPeak.

Getting back onto the subject, why Apple did not launch a new Mac Pro? The answer is quite simple - nothing new under the yellow Sun. For starters, the new-Xeons with Intel's Sandy Bridge architecture will not be ready until March 2011, and in the future, the switch to AMD and Bulldozer / Fusion architecture - something that was recently leaked in AMD's documents and promtply got taken down by tech publications (such as bit-tech)... there are plenty of challenges just to design those parts, yet alone radically innovate.

For expansion slots, you need to conform to the existing PCI Express, ATX, WTX standards - and unfortunately, industry (Apple included) did not show the will to bring forward new and more advanced standards. As Mac (technically) became a PC with either a EATX motherboard (Mac Pro) or a PC notebook (MacBook, MacBook Pro, iMac), the room where Apple innovated shrunk down to the level of very intelligent packaging.

Next Apple's Mac Pro is not going to be as revolutionary as people believe, but it should bring significant performance upgrade, to the tune of how Intel's Sandy Bridge Xeon is behaving faster than the current Westmere architecture, to the tune of faster storage, new GPU architecture such as Fermi-based Quadro, the introduction of LightPeak for peripherals and faster SATA standards. That's it.

Unfortunately, Apple (and just about anybody else in the field of computer design) is limited by the chips offered by the chip vendors. Regardless of going AMD or Intel, the chipset will still feature the same functionality as the ones available on the open market. My Mac Pro has a motherboard manufactured by GigaByte, recent ones were made by ASUS and Foxconn, and the list goes on.

Apple is genius at packaging, there is no doubt in that. But there are physical challenges and international standards the company has to follow. And lastly, it cannot have features which are not present in silicon of today - since everybody would have them, then.

Unless Apple starts making silicon. I don't mean rebranding Samsung's silicon (A4), or Intel's one for that mater. I really mean - making silicon. Only then Apple would be able to innovate at their own pace. As we all know, the company opted to take another route, and not make chips themselves. Yet alone chipsets attaching to a 3rd party CPU/GPU.

 

[jake blackstone]

Because Lightpeak is supposed to be mass market AND therefore cheap (and also idiot proof), at least relatively future proof, and capable of meeting the connectivity performance needs of almost everybody.

The lament was, that he would like to have the capability right now, which is indeed available today, if needed. As far as future proof, not sure what you mean. Why do you think, that 10GB fiber has no future road map? Light peak doesn't have monopoly on that.
Now, if you mean lower price of interconnection, based on economies of scale, than that's different story altogether...

 

[Irmeli R]

quote: you don't need to discuss and waste your time with it...

Posts: 7,478

:Yawn:

 

[Hrvoje Simic]

Separate peripherals would lead to a bit of desktop clutter, true, but Apple has the most innovative and brilliant industrial designer in the world in Jonathan Ive. Something tells me he could probably come up with a simple "stacking" design (probably cable free between modules) that would quickly solve that problem.

Has potential. Let's say MBO could stay in main housing and they could use stacking module for different cards (let's say signaling is not an issue) and/or HDD bays...that would mean they would have to meet average customer demands for standard system. In that case I'd expect additional stacking ability for power users...customizability which is not typical for Cupertino logic. If the logic stays the same, you are stuck with current expansion limitations but with a different product design.

A modular (I did think of that when I started this thread...) approach would allow them to develop and sell many more pieces of gear than they do now, at least for higher end users of the core Mac. It they were interested in going the route I'm speculating, the line would have multiple sized screens, but the same inner electronics. That would allow for different market segments, just as the current iMac does.

Makes sense, but based on their track record, Apple's investment into each individual piece of gear, i.e. separate modules, seems like something Apple is not interested in spending resources - Pro segment and customer# in mind.

For example - multiple screen choices - from a manufacturing standpoint it is another expenditure for Apple's smallest market segment.

BTW, what I don't see is a touch screen, at least not on the vertically oriented iMac case. Apple has already said that they don't feel that people want to hang their arms out to control the screen, and I agree.

Hopefully Apple won't come up with yet another "creative" outburst such as one button mouse, buttonless touchpad or mouse with multitouch on it.

The main issue with touchscreen is reduced visibility of the screen being covered with a hand and that finger can never be precise as a scaled movement ratio high dpi cursor for detailed work. With a mouse or Wacom pen screen estate is enhanced, with touchscreen it's 1:1.

Plus, Apple has legal issues with multitouch...

Of course, I also think that Apple would go for the most forward thinking, least limited, and most unique approach possible, which to me would be optical interconnects, even if they have to wait for that.

I agree that Apple has a creative and bold approach. In some cases.
But those aren't the only and deciding factors with Apple. Some of them seen so far are not exactly lovable.

 

[M Most]

First and foremost, everyone who's been following the computer industry has seen two particular trends: high end gets beefier and beefier until it reaches physical limits, and then the high-end part gets miniaturized to the point of a low-power, mass production silicon. There is no way around this, unfortunately. You design a part with 512 cores sectored in 32 cores and in the end, you have a $500 part and an $30 part, all based on the same architecture - only with less transistors and a dramatically smaller power draw.

This sounds like a description of the primary reasons Red changed their program to a modular design - and why I'm thinking that Apple might essentially do the same thing.

Can you put a high-end graphics card and a powerful CPU onto the same motherboard? The answer is quite simple: No.

Which is exactly why I would think they wouldn't try. That would be a prime example of a technology needed by the high end users but not needed by everyone else. I.e., a perfect candidate for an expansion module. And this is not to mention the possible inclusion of on-board graphics co-processing in future CPUs.

Could you connect the CPU onto the Lightpeak (or just about any other) peripheral standard? For data, yes - but what about power and stability? The major issue with the silicon of today is that we take the numbers for granted - 45nm, 40nm, 32nm, 28nm - all is nice and engineers that create silicon are masters at their work, but the sad reality is that as the technology progresses, it becomes more frail and requires much more power to operate - something which engineering architects are already calling "dark silicon".

See above.

In order for a CPU/GPU to work, you need serious current, in the 85-105-110-120 Amper range. By following the laws of physics, at this current a human could potentially receive a deadly shock with miniscule values. Thus, the idea of shifting silicon parts with 1-3 billion transistors around and attaching them using a standard that is supposed to replace the USB is just not feasable.

Not so sure about that (the feasibility, that is). Cluster computing has been around for quite a while, and even in our little industry, both DaVinci (using Linux and Infiniband) and Filmlight (using Linux and 10GigE) have employed this principle for the last few years. All I'm suggesting is that Apple could conceivably standardize and commoditize the specific methodology.

Getting back onto the subject, why Apple did not launch a new Mac Pro? The answer is quite simple - for starters, the new Intel Sandy Bridge architecture, and in the future, the switch to AMD and Bulldozer / Fusion architecture - something that was recently leaked in AMD's documents and promtply got taken down by tech publications (such as bit-tech)

That's certainly one view, and at this point it's just as valid as mine.

For expansion slots, you need to conform to the existing PCI Express, ATX, WTX standards - and unfortunately, industry did not show the will to bring forward new and more advanced standards. As Mac (technically) became a PC with either a EATX motherboard (Mac Pro) or a PC notebook (MacBook, MacBook Pro, iMac), the room where Apple innovated shrunk to the level of very intelligent packaging.

I think that's easily handled by marketing either their own or others' PCIe expansion chassis. And for the future, Apple would "encourage" their primary 3rd party hardware suppliers to conform to their new "standards." I could easily see a Kona3 (well, it would probably have to be Kona 4....) that's simply a breakout box with all the electronics integrated, and a fiber connection to the bus. The heart of what I think Apple might do is centered around ultimately eliminating PCIe, but for the moment simply replacing it with a LightPeak based interconnect. The third parties will catch up in time, just like they did with things like Firewire and in software, OS X and iOS. And Apple has shown an extraordinary ability to put things out there and wait for everyone to catch up.

Next Apple's Mac Pro is not going to be as revolutionary as people believe, but it should bring significant performance upgrade, to the tune of how Intel's Sandy Bridge Xeon is behaving faster than the current Westmere architecture, to the tune of faster storage, introduction of LightPeak for peripherals and faster SATA standards.

Quite possible. As I said, all of this is anyone's guess. Mine is just one among many.

However, Apple is limited by the chips the company use. Regardless of going AMD or Intel, the chipset will still feature the same functionality as the ones available on the open market. My Mac Pro has a motherboard manufactured by GigaByte, recent ones were made by ASUS and Foxconn, and the list goes on.

Hence one of the reasons for my thinking. With the iMac, they have none of these issues.

Unless Apple starts making silicon. I don't mean rebranding Samsung's silicon (A4), or Intel's one for that mater. I really mean - making silicon. Only then Apple would be able to innovate at their own pace.

Time will tell.

 

[Brook Willard]

If it's faster and has more expansion and configuration options than the current Mac Pro, they can merge it with the iMac line. They could merge it with the iPhone line and I'd be fine with it. I just need more!

 

[J Davis]

I know this may have been lost in the conversation but earlier in the thread I posted about one of the macs now has HDMI out. I for one was really surprised by the addition of HDMI out and hope that if it is useful (re: gamma) then it would start appearing on macbooks for onset work as well as mac pros and iMacs. What do you guys think of this or have any of you had tested it?

 

[Sanjin Jukic]

quote: you don't need to discuss and waste your time with it...

Posts: 7,478

:Yawn:

Interesting is that some people here all the time tracking me,

what I'm saying and doing and even in the case that they were not posting so much

because maybe they can't say so much about anything anyway...

Again very interesting sort of a social network... :laugh: :001_tt1:

BTW, I just uploaded the latest of my Quick & Dirty shot on Vimeo that sometimes makes some of you so dizzy :laugh:

(now shot on GH2 with Kessler Cineslider™ and Gitzo Fluid Head).

My dear fellows and muse then just because of that you should see that new Q&D test

here >>>also switch to Flash Player @ full HD (720p) because then you'll get the accurate gamma >>> :hurray:

 

[Albert Astorga Zepeda]

?? Apple are one of the key partners in Lightpeak. They're certainly not excited about USB3 but Lightpeak is a big part of their roadmap.

they are part of Blu-Ray and have not adopted it yet. ohh but they did adopt HD-DVD. got pissed when it lost and have not touched DVDSP ever since. how about that?...

 

[Albert Astorga Zepeda]

Why? 10GB speed is available today with existing fiber technologies.

That is false.
you probably meant 10Gb.

 

[Albert Astorga Zepeda]

I like the idea of smaller mac pro, but I don't like the idea of limiting the performance to favor the design. Mac pro is about pure POWER. For those who want a cute computer, there is imac. I'm also not liking the idea of having more cables on my desk. Better make a box and put all add-ons inside. ..like the current design

I could care less about looks. all I care is about performance (it is not like I carry a Mac Pro w/30" ACD around showing it off to friends). as it is. the Mac pro with all its beauty and performance it lacks a few very important things...
I am sure they could not go smaller with a mac pro.

 

[Lucas Wilson]

?? Apple are one of the key partners in Lightpeak. They're certainly not excited about USB3 but Lightpeak is a big part of their roadmap.

This is not a comment about Apple/Lightpeak per se, but...

Manufacturers frequently contribute large sums of money to specific ventures that they don't necessarily initially plan on supporting. But they need to be part of the group that steers things so they can spy on competitors and jump in if things appear to be getting serious.

Can't even begin to name the number of specs that had significant major manufacturer "support" that died quiet little deaths.

Lucas

Lucas Wilson
------------
Business Development
3ality Digital

 

[Peter Mosiman]

Manufacturers frequently contribute large sums of money to specific ventures that they don't necessarily initially plan on supporting. But they need to be part of the group that steers things so they can spy on competitors and jump in if things appear to be getting serious.

Can't even begin to name the number of specs that had significant major manufacturer "support" that died quiet little deaths.

Maybe I missed something, but I don't understand what you mean by this?

 

[N_Villers]

This is not a comment about Apple/Lightpeak per se, but...

Manufacturers frequently contribute large sums of money to specific ventures that they don't necessarily initially plan on supporting. But they need to be part of the group that steers things so they can spy on competitors and jump in if things appear to be getting serious.

Can't even begin to name the number of specs that had significant major manufacturer "support" that died quiet little deaths.

I agree with you completely, however, it is my understanding, and correct me if I'm wrong, but it was Apple who approached Intel with the idea behind lightpeak. Innovation with I/O interconnects has long been a strong point with apple having first introduced to market both usb and firewire.

With the speed on cpu clocks approaching the theoretical limits of silicon and beginning to encounter problems associated with quantum physics, the industry as a whole has realized that the solution is to work laterally. With dynamic multiprocessing, a ubiquitous and fast interconnect would be key to allowing scalable infrastructures to become more mainstream. And mainstream is what apple is, and or the most part, has always been about.

It would be super cool to see something more like the mac mini as the core of this new theoretical development. Need more processing, I/O or monitors, just add a new node. Each could be fairly modest in performance as long as in aggregate we could see high performance and real distributed multiprocessing.

I hope Mike is right about this and would reaffirm my belief in Apple who I admittedly have begun to really wonder if I can continue to expect support for professional work.

-noel

 

[M Most]

Maybe I missed something, but I don't understand what you mean by this?

I think what he means is support != product. I think that was in reply to the notion that Apple will integrate LightPeak into next gen products because it was an early participant in the R&D effort. I only suggested LightPeak because it's a known project and might suit the need. For all I know, Apple could be doing its own high speed interconnect, something along the lines of Infiniband, and it would accomplish the same things, especially with a fiber connector. They could call it the Apple Desktop InterConnect. ADIC. Oh, wait, that's already taken. OK, it could be Apple Connective Transport, ACT. Oh, that's taken too. Well, I'm sure they could think of some clever acronym. They've always been pretty good at that.

My point is that none of this is dependent on adoption of a specific known interface. The concept of modularity is what I think might make sense to a company bent on innovation and with a known distaste for the traditional.

 

[Peter Mosiman]

My point is that none of this is dependent on adoption of a specific known interface. The concept of modularity is what I think might make sense to a company bent on innovation and with a known distaste for the traditional.

Thanks for clarifying.

My question to then would be - as a longtime Apple enthusiast and once upon a time employee, I find the idea of modularity to be a PC thing, and Apple doesn't want to ever go there - Q: Wouldn't modularity take away from Apple's business model of consumers can find something that works for them with the "small" amount of choices provided (ie a few versions of each lineup, but not endless DIY configs)?

"Giving more choices to the consumer only makes it more confusing, and it's unnecessary because we have come up with the best standard configurations possible." - This is how I feel Apple does things. I guess the PRO lineup is different from the consumer, but I find it hard to believe that Apple will have the PRO lineup become so radically different than its consumer lineup because then it doesn't all fit in a nice neat package.
Does this make sense?

 

[M Most]

"Giving more choices to the consumer only makes it more confusing, and it's unnecessary because we have come up with the best standard configurations possible." - This is how I feel Apple does things. I guess the PRO lineup is different from the consumer, but I find it hard to believe that Apple will have the PRO lineup become so radically different than its consumer lineup because then it doesn't all fit in a nice neat package.
Does this make sense?

Yes, it does. But in some ways, you're actually advocating what I outlined. Right now, the Mac Pro is the oddball in Apple's lineup. It is the only box that is sold to the user as something they can open up and put things in. If you want to add storage to the iMac, or any of the laptops - well, you can't, at least not internally. You attach an external box. This is no different than attaching an external box for, well, something else, like pro level video I/O. At the moment, they're manufacturing a machine in the Mac Pro that shares very little if anything with any of their other machines. My point is that they don't have to. All they really need to do is find a way of connecting something to an iMac in a way that eliminates the speed limitations inherent in current interfaces like USB and Firewire, and has hooks into the system that would allow it to essentially operate as a bus level device. This would, to me, be completely in line with everything you've just said regarding prebuilt configurations of somewhat closed boxes, but still accommodate the pro user within that architecture.

 

[Peter Mosiman]

Aha, I copy you now. And this is how the suggestion of using LP as the interconnect has come into the equation?