All future computing power in the cloud? {Groan} let's hope not

[deconstruct60]

RAM,

4 DIMM slots per CPU package is perfectly aligned with the future Xeon E5 move to DDR4. Single controller per DIMM slot. 4 controllers, 4 slots. This is pretty much pruned as far as it can go unless start pruning off x86 cores (i.e., cap the Mac Pro at 4 cores). Pruning x86 cores is an extremely dubious move at these price points.

Unless Apple is going to start moving whole Mac line up down to lower price points then abandoning these price points don't make much sense either. ( Besides the 27" iMac has four slots. The 21.5" iMac doesn't have four slots for any good technical reason. It is more vanity (can't let folks see RAM door) and/or bad packaging by Apple. It is also a bit of artificial product segmentation to push more folks into the 27" model.)

optical drives,

At least one likely tossed. The fact that many configurations posted here have 2.5" and 3.5" drives stuffed into the 5.25" bay only points that even the users are aligned with two ODDs bays is too many. However, that directly points to the space needing to be repurposed; not removed. This is actually something Apple is grossly behind the curve responding to.

Both 5.25 bays? I'm sure there would be much wailing but yes it is possible. Either a Xserve like sled ( so could service from the front without opening lid) or two transverse mounted 3.5" in the upper bay would likely get more traction over time. ( could move then out of the PCI-e card zone for more room. )

and 3.5" hard drives (in favor of 2.5") all seem like good candidates for cutting.

I can see 3.5" number going down but not down to zero. Most Mac Pro users want a SSD in the configuration. That is easily done with a 2.5" drive sled.

Possibly also dual processors?

Over the long term, the Mac Pro is doomed if a healthy variety of software that scales by core does not show up.

if the bulk of the software doesn't scale then 4 cores is going to be enough. 6 or 8 cores isn't going to matter much either.

Most of the embarrassingly parallel code will drift to GPGPU cards. But there lots more code that either has a mix of scalar and embarrassingly parallel code or users that run multiple program concurrently.

What has been lacking is applications with a the mix of OpenCL and GCD usage. Over the next 5-10 years apps built to leverage the combo of those should grow. If not then the overall Mac platform is in deep trouble.

I think the chances of Apple dropping dual processors are less though. More likely seems like introing a low end with an i7 maybe.

There is small possibility they could split the Mac Pro into two boxes.

At least for the short-medium term, an Xeon E3 would be a better fit with Mac Pro. More PCI-e lanes than a mainstream i7, ECC memory for double/triple digit RAM installs like the rest of the Mac Pro, and not particularly catering to overclocker crowd (take risk out of long term hardware support coverage). Thunderbolt is still trivial to implement as with an i7, so probably can do a typical Apple margin (30+% ) offering near the $2000 border with that kind of box.

The "extreme" i7 models don't buy anything significant. They cost exactly the same as the E5 Xeon models. They run just as hot TDP wise so the size of the box being driven by that constraint isn't any better. Using Xeon E5 1600 and 2600 CPUs allows for shared infrastructure and component costs. i7 throws that out the window for almost zero net benefit outside of tinkering flexibility.

All those things also nicely dovetail with getting a Mac Pro more rack friendly.

It doesn't have to be rack friendly, just not rack hostile. It is primarily just the handles that are the problem. They have absolutely nothing to do with the device functioning as a computer. Cutting into the storage and computational functionality because of the handles is horrifically bad design.

Modest adjustments for shape but keeping the same volume would make the it neither friendly or hostile. It would still be a little quirky, but much less a gratuitous pain.

A smaller box has far more synergy with being on top of a desk instead of nominally being targeted for the floor (or optionally a rack). With Thunderbolt and targeted at actual desktop would make it more aligned with the "tail wagging dog" position of pairing it with the Thunderbolt docking station (Display). The dangling MagSafe connector would still illustrate that it is a contrived configuration, but it far less contrived as the Mac Pro sitting on the floor and the TB docking station not being able to reach flexibly.

Apple delivered MBA along side MBP so maybe they'll do two versions for several years. It is a bigger risk than they have done previously.

(I also think compatibility with the new wave of PC EFI cards is on the list. Making it so Apple doesn't have to make special cards takes some of the development costs off the Mac Pro.)

That has also been an inhibitor to the Mac Pro market. If the value add of the box is being able to insert PCI-e cards to then minimizing the size of the PCI-e card market isn't going to help. With Windows 8 being UEFI by default it looks like the rest of the market has caught up to where Apple is.
If Apple can get the Mac Pro back on a positive growth path then 3rd party vendors will likely contribute to the ecosystem.

 

[deconstruct60]

Indeed. The idea of server and "dumb" terminals has been around commercially since around the 1960s. It is isn't some new and nefarious plan that was dreamed up in the 1980s. We are now starting to come full circle,

It not quite a circle but a pendulum. In the early PC era central, time/disk/resource sharing servers were evil so there was almost mindless stampede into PCs. It over corrected. One reason VMware is doing billion dollar business consolidating too many physical servers back into a more reasonable number. Every collection of 2+ people didn't need their own physical server.

I am always vaguely curious to listen to the doom and gloom about the Mac Pro on these forums. "It's all about the iToys" blah blah blah. People seem to forget that the reason why the iToys are successful is because of an ecosystem.

They are successful because they are not toys. The fundalmental flaw here is labeling them as toys at all. Not that there is an ecosystem. There is an ecosystem for every non dead platform.

An ecosystem of apps and content. If Apple releases Xcode that can run on Linux, or Windows, that will be a reason to start worrying.

Xcode runs on OS X ; not just on Mac Pros . Again it is fundamentally flawed to equate the Mac Pro to the overall Mac product line. They aren't the same. Without a Mac Pro there would still would be content and apps for the iOS platform.

When a quad core A7 ( ARM A15-A7 arch era ) rolls out with 2GB of RAM the only reason XCode and development tools aren't running on iOS devices is because Apple is managing the security of iOS deployments more so than the capabilities of creating content. Apple never needs to jump to Linux or Windows if OS X dries up for app development for iOS. The whole notion that "has to" keep the Mac Pro just to do content creation is huge farce. No, they don't.

I don't believe that Apple is run by idiots and I am sure they have not forgotten that content needs to be created in order for it to be consumed on iDevices.

Media content is standards based is platform neutral. Yes, Apple isn't run by idiots and they know that they will still have content without a Mac Pro in the line up. This "threat" that the narcissistic level at them is completely empty. The vast majority of media content consumers do NOT have Apple products. Those folks drive the content market.

3.4GHz Quad Core, 32GB , Fusion Drive iMacs can likely compile just about everything in Apple's application/OS stack portfolio in blinding speed with the new LLVM tool stack. For corner cases where the builds are long all that is needed is to bring back parallel build tools and rack full of Mac minis. Time and file system synched, the app build process can easily be made embarrassingly parallel unless the code has been designed in a brain damaged way.

 

[deconstruct60]

If Apple chose to skip as large an upgrade as the one to E5 generation Xeons - and let's be blunt, that was a *choice*, not a necessity - it's a reasonable bet that the MP as we have known it is dead.

There is no indication at all that Apple is skipping the first generation of the Xeon E5 branded products on purpose purely based on those part's properties. That is largely been driven by lots of hand waving analysis that doesn't have much substance to it.

A new Mac Pro arriving in late Q3 - early Q4 is much more likely arriving at that time primarily because that is when they would have something ready. If going to arrive at that time then E5 v2 would be a natural choice since would basically be compatible with the motherboard for a v1 design. ( same support chipset and socket pin configuration. )

It was by choice. But the choice isn't skipping as much as doing the work at all (i.e., choosing to start working late so finishing later then most. ).

There may be something called a "Mac Pro" in the future, but it seems quite unlikely it will resemble the tower we have come to know and love.

It is very likely a tower. The "rectangle" resemblance will likely still be there. It will look far more like previous versions than any of the other Mac models.

Will probably get resurrected but slightly modified 9600 -> G3 complaints.... but what isn't released without complaints on these forums?

Will it be faster than previous models on most applications? Yep.
Will it require some folks to leave behind some legacy oriented baggage? Yep.
Will it not have enough options to satisfy any possible customer? Yep.

 

[xVeinx]

Apart from the 2012 iMac that is. Great for consumers who like the 'thin' look, a horror show for people who actually use computers as, y'know, computers.

Way to go apple, you reminded me why I like the pre iStuff machines on my desk, thanks. Sure, sell all of the iMacs you can to people that just want to read the news and play games (and drive up the cost of my apple stock ). How about a computer for the rest of us (now where have I heard that phrase) to use?

I think people take themselves too seriously when it comes to computing (I used to make fun of people who used Apple computers for instance, only to see their value now!). I'm well aware of what it takes to do high-performance computing-scientific computing in particular. The iMac has never been about being packed to the gills with high-power components and customizability with any fanciful gadget that comes to market. Ever. It was designed to serve as a reliable office/home machine, and it fulfills this role admirably even now. If anything, Apple has begun conceding to the "gamer" market by providing high-power graphics (albeit at a cost). Perhaps Apple will release a smaller, separate box that people can customize (without warrantee or support for upgraded parts), and that will be sufficient to appease the dissatisfaction. A PC with Mac OSX basically. Leave it to the manufacturers to support and supply drivers with their products. The problem with this model is it devalues the engineering that Apple does and substitutes it with a basic box that gets pummeled with random components but has some increased functionality.

The Mac Pro is the machine that bears the power and flexibility to add components useful for audio/visual/scientific computing. I understand that people don't want the machine to become less powerful, but I don't see any reason to assume that the machine is suddenly going to become a terminal connecting to a cloud environment. Like I mentioned before, they will very likely release a machine that is well designed and powerful. Will it be everything that anyone could ever dream? If we're lucky, it'll be better .

 

[Tesselator]

It not quite a circle but a pendulum. In the early PC era central, time/disk/resource sharing servers were evil so there was almost mindless stampede into PCs.

I blame Jeff Bridges!

Interesting how now that the "pendulum" as you refer to it, is scheduled to swing back the other way we have old Jeff escaping into a world of handhelds. And I guess there won't be sequel where the evil MCP is now Google and/or other Cloud providers. LOL

No, instead we have a bunch of lunatic (psudo)scientists telling us we ourselves are a product of "the cloud" and the entire universe is a computer simulation. Hehehhee....

 

[deconstruct60]

Interesting how now that the "pendulum" as you refer to it, is scheduled to swing back the other way we have old Jeff escaping into a world of handhelds.

The pendulum is in motion because of gravity ( assuming that friction and air resistance don't get in the way). It swings because responding to natural forces.

Changes in technology , adjustments to marketing, changing preferences of customers are the natural forces here. As long as they are active the pendulum will keep swinging and incrementally moving to equilibrium.

 

[damir00]

There is no indication at all that Apple is skipping the first generation of the Xeon E5 branded products on purpose purely based on those part's properties.

I don't know what this sentence is trying to convey.

Apple has either decided to skip E5, or is going to be so late in delivering that they'll (again) be a release behind. When all is said and done, it doesn't really matter much what the reason for that choice is.

 

[Photovore]

Some content creation is done with proprietary software. I can't imagine anyone who owns rights to any ground-breaking trade-secret code being willing to put their life's work up into the cloud, and therefore potentially out of their hands. I'm sure not putting mine up there....

 

[deconstruct60]

I don't know what this sentence is trying to convey.

Apple has either decided to skip E5, or is going to be so late in delivering that they'll (again) be a release behind.

Some folks have proposed that Apple is skipping Sandy Bridge E5 because somehow they aren't .... fill in the blank (fast enough , low power enough , etc. ).

There is no solid evidence at all of that.

Be release behind again? Why? The upcoming Ivy Bridge Xeon E5 fit in the exact same motherboard (minus some firmware tweaks) as a Sandy Bridge Xeon E5 would fit in. They could ship with either one as long as the board and the rest of the system was complete.

The closer they get to the ship date of the Ivy Bridge E5 though it doesn't make much sense to ship the Sandy Bridge variant 2-3 months and then change. They can skip it but that is because the rest of the system was late.

That the tick/tock cycles Intel has for this class of CPU package are longer than most of the frenzied mainstream expectations would warrant. There is no new motherboards coming every 11-12 months. The architecture codename isn't going to match the mainstream CPU versions. Yikes the iMac is on haswell and we're on Sandy bridge oh the horror. That is typically not as "behind" of late as most complaints make it out to be from a x86 microarchitecture and I/O perspective.


Consistent delivery of new models is the primary problem the Mac Pro has. Being tightly synched to latest out of Intel isn't.

When all is said and done, it doesn't really matter much what the reason for that choice is.

If it is because they had terminated it and then revived it then it is a on-going issue. If customers don't buy in numbers that demonstrate growth they'll stop selling it.

 

[deconstruct60]

Some content creation is done with proprietary software. I can't imagine anyone who owns rights to any ground-breaking trade-secret code being willing to put their life's work up into the cloud, and therefore potentially out of their hands.

This code has never been put on a laptop or box that ventured outside the company to do work local at another site? If it has, there is no material difference.

There is nothing inherent in the cloud that prohibits security. What policies are laid on top of the resources gives security.

There are apps that are disconnected from the entire internet. Or pretty much stationary on single systems ( or LAN cluster of systems). However those are relatively few.

 

[Tesselator]

There is nothing inherent in the cloud that prohibits security. What policies are laid on top of the resources gives security.

Hahaha... even you should laugh at that.

I doubt thee are many systems connected to the public grid that aren't hackable. And most are hackable by 16 to 20 year-olds. Seriously, I can't believe you just said that.

 

[seveej]

Here's the problem. I started thinking about what Apple's competitors are doing, and about how much money they're putting into cloud computing. Then I started thinking about how Timmy's next great idea could be to come out with 'Cloud Pro' instead of a Mac Pro.

On one level, I share your worry that this will be the future, and (should it be made to work) it in many ways holds huge promise.
Just think: Everyone would have "terminals", on their desktops, in their laps, under their arms, in their pockets. All these terminals would allow you to work your data, view your pics, listen to your rips etc. All these terminals could give you your interface of choice.

Problems: The network infra's not there yet, but might be soon (10 years or so), and the international legal infrastructure is not there yet (no matter what you think, many companies and individuals neither can nor will give their data to be stored without regard to borders). And finally: There are instances where any feasible amount of bandwidth will not do away with the need to have on-site computing power and storage.

Honestly, I do not see Apple pushing the cloud as a solution to serious "creators". with huge amounts of working files. The cloud is (as it is) currently targeted at those people who consume media and do no heavy lifting with regard to files.

RGDS,

 

[Photovore]

This code has never been put on a laptop or box that ventured outside the company to do work local at another site? If it has, there is no material difference.

The Mac Pro goes out, does its work, and comes back. If I have to leave the building for a couple hours, it's padlocked to something sturdy. I also have a copy on my laptop which is always with me, and one in a secure location at my day job.

There is nothing inherent in the cloud that prohibits security. What policies are laid on top of the resources gives security.

I could encrypt data to/from the cloud, but that doesn't matter in this situation. If the app itself is running up there, then the algorithms are potentially exposed. Also, the app is always under construction. Is XCode going to be in the cloud too? Source files?

There are apps that are disconnected from the entire internet. Or pretty much stationary on single systems ( or LAN cluster of systems). However those are relatively few.

That describes this app. On location for a few hours and then back home. The internet would be no help anyway unless 2GB/s bandwidth were guaranteed -- maybe some day.

 

[arjen92]

I'm going to guess that it will be a powerful and well-designed machine. There isn't anything to suggest that Apple would attempt such a thing, especially since they don't have the infrastructure anyway. So, sit back, relax, it was just a bad dream

If you're talking about cloud computing not going to happen. I wouldn't be surprised. Perhaps I'm mistaken on what these datacenters are for. But I wouldn't say they don't have the infrastructure.

 

[goMac]

Editorial today on Apple going back to minitowers:
http://www.macworld.com/article/2029740/the-time-is-finally-right-for-a-mac-minitower.html

I don't necessarily agree with everything in here. Thunderbolt wouldn't be great for things like video cards. But still interesting.

 

[deconstruct60]

Editorial today on Apple going back to minitowers:

I'm sure it was good for adviews. I reads like a trolling for ad views article.

I don't necessarily agree with everything in here.

Perhaps not..
" ... So when it came time to upgrade my desktop Mac a few years back, I opted for a 27-inch Core i7 iMac with 12GB of RAM. ... "

Because the preamble highlights that besides the fixed on the form factor he previously bough ( a "form over function" priority) that iMac is a very close fit to what he needed.

The xMac is almost solely because he wants an largely overlapping alternative to the iMac. None of this is particularly about the Mac Pro real target market.

Thunderbolt wouldn't be great for things like video cards. But still interesting.

Hardly. Read a several of the "xMac" threads on various Apple rumor/forum sites. Massage the argument and ignore the market ramifications for Apple and ta-da .... probably a 10% boost in adviews for his column.

Everything in his "making the case" section applies equally to the Mac mini and iMac as it applies to his xMac. So it isn't much of a case at all.

 

[Giuly]

Why not? You only pay for the computing power you actually use, if you need more RAM or CPUs they're added to the instance automatically (and you will get billed), no fans running in front of you, keeps running even during a power outage and all you need to connect to the cloud is something akin to the computing power of an Apple TV with an AxX chip. A 10MBit/s internet connection will be sufficient, and as the data centers are hooked up via multiple Gbit/s downloading will appear blazing fast (and you will get billed by the GB).

All problems with mainframes and terminals have been fixed over the last 30 years, so it's the way to go.

 

[goMac]

All problems with mainframes and terminals have been fixed over the last 30 years, so it's the way to go.

The bandwidth problem hasn't been fixed.

I have a business connection and it is still multiple factors slower than an SSD, and even slower than a traditional hard drive.

Why would I take performance improvements from SSDs and throw those out the window for cloud computation?

Cloud computing works well when the files are small but the processing is big (like say, computing the human genome or solving for pi.) It doesn't work so well when the files are big (like video editing, music editing, image editing.) Which market is Apple more interested in?

 

[Tesselator]

^^^ Exactly correct goMac! And it's NOT likely to change in the near future either.

Editorial today on Apple going back to minitowers:
http://www.macworld.com/article/2029740/the-time-is-finally-right-for-a-mac-minitower.html

I don't necessarily agree with everything in here. Thunderbolt wouldn't be great for things like video cards. But still interesting.

That would actually be cool! I can imagine a half-sized MacPro with:

• Dual Xeons (various configurations),
• 2 (dual width) PCIe v4 x32 slots,
• 4 or 6 RAM slots,
• Zero internal drive bays,
• Front loading slim slot for an ejectable SSD,
  Or maybe internal dual mSATA SSD ports,
• Thunderbolt x8,
• BlueTooth 4.0,
• Same grade aluminum case.

I'd buy one without hesitation.

 

[bearcatrp]

So the notion that somehow a workstation is diminished because several services are moved to being on a wire coming out of the back of the box really doesn't hold alot of water.

Your so wrong on this statement. The big push to cloud computing is rolling full steam ahead. Every month more software vendors are pushing to do everything on THEIR servers, for a fee of course. Looking at the new OS'S, you can tell that's the plan. Plus, looking at the prices of server chips, getting higher every year to keep allot of folks from buying. I build my own ever since selling my 2008 mac pro. Looking at the prices to build the latest systems is outrageously expensive. Soon it will be all subscription based. Just the icon on your desktop to click to log in to whoever.

----------

Editorial today on Apple going back to minitowers:
http://www.macworld.com/article/2029740/the-time-is-finally-right-for-a-mac-minitower.html

I don't necessarily agree with everything in here. Thunderbolt wouldn't be great for things like video cards. But still interesting.

I could see apple building a 8 core single chip small mac pro, but still think they won't.

 

[Giuly]

The bandwidth problem hasn't been fixed.

I have a business connection and it is still multiple factors slower than an SSD, and even slower than a traditional hard drive.

Why would I take performance improvements from SSDs and throw those out the window for cloud computation?

Cloud computing works well when the files are small but the processing is big (like say, computing the human genome or solving for pi.) It doesn't work so well when the files are big (like video editing, music editing, image editing.) Which market is Apple more interested in?

The cloud has storage that performs way better than single SSDs, but I aknowledged that the initial upload of your data into the cloud will be a tough job with current broadband technology (hasn't changed much over the last years, has it?), at least for workloads > 1GB.

You have to put that into perspective with something like Google Fiber - which is 1GBit/s down- and upload -, because it's unlikely that general purpose cloud computing will be available before Gigabit Internet.

The cloud basically moves the whole computer (RAM, CPU, storage, graphics) out of your sight, and just leaves you with a screen that pops directly into the network. If you upload something from a camera for example, the camera will be the bottleneck, not the Internet connection (for Google Fiber or most commonly available business connections) or the cloud.

 

[GermanyChris]

I don't "get" the cloud and I don't "get" why people would like to make a tower PC smaller?? It's fits under/on top of your desk why is smaller beneficial??

 

[deconstruct60]

I don't "get" the cloud

The cloud is relatively nebulous term like computer. Sales wonks like it because can pour it over a wide variety of products like buzzword ketchup. Transforms known products into new exotic ones.

and I don't "get" why people would like to make a tower PC smaller?? It's fits under/on top of your desk why is smaller beneficial??

Few to non of these folks would pay Mac Pro prices for this "smaller". Most often it is, similarly to the above, just repackaged, beaten-to-death xMac arguments presented as something 'new'.

Smaller is going to get them a lower cost.

There are a significant number of customers who bought Mac Pro that were really too big with respect to workload and usage. It was bought more on image (perception of power rather than a need for the functionality) or future proof that is never significantly leveraged. Making it smaller will limit the functional so not as obviously an overextended buy.

 

[Photovore]

As I mentioned above, the cloud is useless to me unless two Gigabytes per second of downstream speed is guaranteed. If I am not mistaken, this is far beyond the current capabilities in the U.S. If I am mistaken, please inform me!

 

[goMac]

The cloud has storage that performs way better than single SSDs, but I aknowledged that the initial upload of your data into the cloud will be a tough job with current broadband technology (hasn't changed much over the last years, has it?), at least for workloads > 1GB.

But it doesn't. If I've got 50 gigs, 100 gigs, or (god forbid) terrabytes of data, it's going to take me up to a day to load that into the cloud. And if you're a pro shooting video, you're dealing with those sorts of data from your camera.

The cloud may have faster disks (it probably doesn't), but all that gain is lost immediately in the data transfer rates.

You have to put that into perspective with something like Google Fiber - which is 1GBit/s down- and upload -, because it's unlikely that general purpose cloud computing will be available before Gigabit Internet.

And for Pros, even Google Fiber is several factors slower than the local NAS attached by fiber channel.

The cloud basically moves the whole computer (RAM, CPU, storage, graphics) out of your sight, and just leaves you with a screen that pops directly into the network. If you upload something from a camera for example, the camera will be the bottleneck, not the Internet connection (for Google Fiber or most commonly available business connections) or the cloud.

I know what the cloud is. It's just not usable for Pro work. If I've got 50 gigs of data, I don't want to have to wait half a day to upload it to the cloud, and I don't want to have to worry about hitting a data cap that turns my computer into a doorstop.

The cloud may work great for consumers, but there is no way it could fit into consumer workflows at this point.

You also end up with situations were either by legal reasons or preference, pros cannot/do not want their data off site or on someone else's disk. I have worked some places where this is actually forbidden by state law. And if you're making the new Star Trek movie, you definitely don't want your data on the cloud where there is a risk it could be stolen. Google already got caught with employees looking at user's GChats, imagine storing movies on the cloud?

You're far more likely to see pro houses build out their own internal clouds (which most of them have already done), but there is no way pros are moving to the cloud on the internet.

Pros buy high end hardware because they don't like compromises. But the cloud is full of compromises.