Mac Pro to be discontinued?

[balamw]

MOD NOTE: Knock off the bickering.

B

 

[-hh]

If Apple PR says there is a Mac Pro coming in 2013 is the "new" adjective necessary? If working on Mac Pro to be released next year it is inherently new since it is a future product.

Yes, except that Apple PR didn't say "Mac Pro" in 2013 ... they said "Something New", and that's the nuance I'm referring to.

(speculation on what a change in form may be)

It is more likely the changes will be to a marginally less modular Mac Pro rather than more. For example, an embedded GPU versus being able to remove all the GPUs from the machine.

Agreed, that too.

The CPUs and GPUs won't migrate to different boxes. Modularity is far more likely to be more akin to the "external Superdrive" (as path already on for laptops and Mac mini). Perhaps extra Firewire , USB , and perhaps Ethernet as with TB dongles and TB Display. There are definately some candidates of built in peripheral sockets and peripherals themselves that may move to external boxes.

Wasn't really what I was thinking of. What I was thinking of was a single-CPU based "Lowest Common Denominator" as a building block so that if a particular customer's workflow process required 47 cores, he could buy them in multiples of, say, 8...so he would stack six modular boxes ... 6 x 8 = 48.

The components that need very high bandwidth interlink won't. (CPU, RAM, greater than x8 PCI-e slots, and GPU )

Yes, this does raise up a technical question of how to get the interconnects to talk to each other fast enough, but the payoff for Apple is that it collapses the cost of the Mac Pro product line down to fewer hardware configurations to design, build & support.

(Firewire)

Why? Because no personal computer system vendor or mainstream peripheral maker is interested in using them. Especially not Intel who is pretty adamant against Firewire. (won't find more than FW400 support from them and that only because of the wide use on video cameras at one point. ). With less than 10% of the personal computer market Apple can't push FW (or Thunderbolt) up the hill by itself. Draft USB 3.0 killed off those two before they every got out of the gate.

Understood, but my point is that since it is already a separate controller chip, the incremental cost to Apple of upgunning the FW800 to (higher) shouldn't have been all that huge, and the specialty-for-Apple aftermarket would have followed, just as how there's still FW800 devices for sale still today. Yes, they're all external HDDs, but the alternative before TB was eSATA, and TB still isn't quite there yet for being cost-competitive.

Frankly, Apple "gave up" on Firewire long ago when they used holding back FW800 from the entry level Macs a market segmentation technique. That killed FW800 momentum ( which was already weak) which essentially nukes the follow ons.

Yes, but it is still water under the bridge that we can't bail back upstream.

Thunderbolt has a chance because technically it is Intel that is pushing it forward. If Intel puts the TB controller on several design reference boards, then it will catch on in the overall market. They have (at this point with Ivy Bridge designs) and it is growing.

Sure, but TB is still not yet on the Mac Pro. Consumers don't care about the technical challenges or reasons why that is so.

(HDD I/O)

Actually, it is getting pretty close when drop to 2.5". It only works on large sequential files but the new 1TB velociraptor peaks around 200MB/s ( 1,600 Gb/s ). (http://www.storagereview.com/western_digital_velociraptor_1tb_review) That means two in RAID 0 with everything aligned can ( 3,200 Gb/s > 3,000 Gb/s ) and that with four present many early SATA II controllers with bandwidth issues ( since the traffic consolidates into a single link eventually). That's one reason the new top end ones come with SATA III links.

Okay, but that's (a) using a RAID0, and (b) a Velociraptor isn't a 3.5" disk, but a precursor to the 2.5" SSD age.

3.5" aren't performance drives anymore. They are drifting to being almost elusively being the $/GB leaders with "reasonable" access times.

Right, which is why so many MacPro users are looking at hybridized systems, to try to have a good compromise.

No. One reason the relatively (compared to the rest of the Mac line up) CPU in the MBA 11" does reasonably well is because it is not hobbled by a small , slow HDD. The SSD storage basically equalizes the MBA 11" with the MBP 13" models on many metrics.

True, but not my point. My point is that despite products with SSDs like the MBA notwithstanding, the generic mainstream PC consumer (not just Macs) is still mostly running hardware with a single spindle of spinning media for their boot+data drive.

So yes, there's no doubt that they would get a nice performance boost from an SSD...but the user demographic is such that they're not going to pay for it (yet...within the mainstream demographic). Similarly, for their application of an 'external drive' for data backups, it is going to typically still be a single spindle HDD for the generic "Joe Homeowner" mainstream demographic. We have to keep in mind that power users aren't typical consumers.

In the iPod and iOS space there is only one, largely comatose, HDD model right now. Again the success of the iPad/iPhone is in part because don't have seek latency hiccups that HDDs incur.

YMMV. I see it more as a system architecture that isn't calling for having even 100GB of data storage, let alone a couple of TB. As such, the cost premium for upgrading to solid state is modest and it pays for itself in its system performance gains.

It would not be surprising if every Mac in the 2013 line up had a SSD on board, including the Mac Pro. ( if only as a cache accelerator to create a virtual Hybrid drive or as a OS/Apps drive. XServe had the latter long before it should up on other Macs. ). And yet another issue which should be driving a revision of the Mac Pro

Given Apple bought a SSD controller design company, it is likely that SSDs are more tightly aligned with long term strategic objectives than HDDs are. Sure they will be incomporated into ARM based SoC designs first, but it isn't like SSDs aren't displacing in Mac designs either.

Agree that that should be the direction...but time will tell with Apple. Frankly, I don't really expect the MacPro to get an SSD as standard unless there's other major changes in store, such as to form factor, etc.

The aftermarket bracket is kind of thing that Apple will attack (e.g., like no need for 3rd party sleds in current model).

Too late for the 2012 model ... these sorts of extra goodies fit the day it shipped.

Likewise, a need for 3, 4, 8, or 12 'bulk storage' 3.5" drives is exactly the solution that Thunderbolt enables. Older, slower connections ( SATA II or SATA III ) in external boxes. That is far more likely to be directed at "modular" solutions than central core bandwidth tasks like connecting CPU and GPU.

And an iMac with a Promise array prices out to roughly the same as a Mac Pro + Apple LCD + internal HDDs. Doesn't make it compelling at all.

A "snap on" , stackable enclosure with 4-5 bays for 3.5" drives makes much more sense. There are many existing server , storage systems , and even NAS boxes that have "expansion" boxes along these lines. It would likely be labeled "revolutionary" by Apple, but it already been done and has a proven to be effective.

Make it a HDD enclosure with a SuperDrive bay and could chuck both HDDs and ODDs out of the core Mac Pro box. Since the enclsure would likely include a SATA controllers effectively it would "look like" an internal ODD for those that "need" old Windows and App DVDs and CDs to see it as an "internal" drive.

Given how Apple gave this away with the one XServe RAID model, I don't see them keeping such a "stack" as an in-house product when it is just a bunch of hard drives. There needs to be a brain inside that box for Apple to want to keep it as an Apple product.


-hh

 

[Bubba Satori]

Yes.

 

[imageWIS]

Apple will have to come up with a new Mac Pro or the equivalent very, very soon. Some new machine with a new, smaller case design that still allows for user replaceability / scalability.

 

[ActionableMango]

Apple will have to come up with a new Mac Pro or the equivalent very, very soon. Some new machine with a new, smaller case design that still allows for user replaceability / scalability.

We already know it won't be very, very soon.

 

[imageWIS]

We already know it won't be very, very soon.

I said has to, not will

 

[deconstruct60]

Yes, except that Apple PR didn't say "Mac Pro" in 2013 ... they said "Something New", and that's the nuance I'm referring to.

No this is incorrect. Go back to post 51 in this thread for the link outlining the actual facts.

Spinning Tim Cook's bungled attempt to push the info out is just spreading FUD at this point.

Wasn't really what I was thinking of. What I was thinking of was a single-CPU based "Lowest Common Denominator" as a building block so that if a particular customer's workflow process required 47 cores, he could buy them in multiples of, say, 8...so he would stack six modular boxes ... 6 x 8 = 48.

Like the Mac mini ? Already on the market. It is multiples of 2 or 4 now but wouldn't be surprisin to be all 4's when the next revisions roll out on Ivy Bridge.

If it is cores without a large amount of memory per core Apple's new design efforts should be trying to incorporate cards like the following

http://www.cpu-world.com/news_2012/2012080201_Details_of_Intel_Xeon_Phi_coprocessors.html

Modular expansion could be done with a x8-x16 PCI-e lanes for this kind of scaling.

Yes, this does raise up a technical question of how to get the interconnects to talk to each other fast enough, but the payoff for Apple is that it collapses the cost of the Mac Pro product line down to fewer hardware configurations to design, build & support.

Not really. If have to invent a new interconnect that raises cost. If have to significantly raise costs to get to a simplification, it isn't necessarily true have lowered overall costs.

Effectively, Apple has already done a config simplification. That's why there is a CPU/core support chip/RAM daughter card in the current models. Those simply leverage PCI-e as the interconnect .... just like all of the other designs without daughter cards. No "new" interconnect technology needed.

Understood, but my point is that since it is already a separate controller chip, the incremental cost to Apple of upgunning the FW800 to (higher) shouldn't have been all that huge, and the specialty-for-Apple aftermarket would have followed,

No. It isn't that one sided of an equation. It is an additional controller chip and design costs for the peripheral vendors. It isn't a viable strategy if only 2-3 "only for Mac" aftermarket followed. That isn't sufficient.

In terms of units deployed, there are much more than just FW800 external HDDs out there.

Yes, they're all external HDDs, but the alternative before TB was eSATA, and TB still isn't quite there yet for being cost-competitive.

Whether Thunderbolt is successful or not doesn't depend at all on the external single (or even dual) HDD market that eSATA and FW800 enabled. That isn't its primary usage and that is the primary driver why it isn't "cost competitive" ( use for something not particularly designed for).

Sure, but TB is still not yet on the Mac Pro. Consumers don't care about the technical challenges or reasons why that is so.

The being deployed to the Mac Pro is also extremely irrelevant to Thunderbolts eventual success or not. TB solves a problem workstations like the Mac Pro don't have: PCI-e expansion and mutliple video output streams. Those are largely a non issue for that class of machines.

Thunderbolt is oriented to machines which have embedded graphics on the motherboard. That is not the standard workstation class architecture. It is also oriented toward machines for which there is little to limited (e.g., ExpressCard) expansion. Again this is not standard workstation class architecture.

There is some push for TB on a Mac Pro to make it "consistant" with the other Macs. However, the Mac Pro (and workstations like it) are not going to make TB viable. It is the other models it is actually targeted at that will assure that.

Okay, but that's (a) using a RAID0, and (b) a Velociraptor isn't a 3.5" disk, but a precursor to the 2.5" SSD age.

2.5" is only neccesary in that is where the higher densities are deployed earlier. With the 3.5" drives there is always the temptation to just use more platters to crank up the storage capacity. So they trail on densities.
As the densities go up the peak sequential transfer bandwidth goes up also.
"Slower than SATA II" is not a rule of thumb that is going to hold up as the densities increase. Increasing the densities is the only way HDDs survive against the SSD onslaught so it probably will.

Frankly, I don't really expect the MacPro to get an SSD as standard unless there's other major changes in store, such as to form factor, etc.

There is really no need for a major form factor change for that. XServe has substantially less internal volume and had SSDs as an option. Apple's mSATA card derivative could be incorporated relatively easily.

Too late for the 2012 model ... these sorts of extra goodies fit the day it shipped.

Immaterial. the 2012 model is obviously a placeholder. Whether Apple wants to tackle the 3.5-2.5 adapter market would be addressed by them. There is almost nothing new to be discerned about Apple's future direction from the 2012 model than from the 2009 models.

And an iMac with a Promise array prices out to roughly the same as a Mac Pro + Apple LCD + internal HDDs. Doesn't make it compelling at all.

A Promise array incorporates a RAID card. A JBOD box would be significantly cheaper because all that need to enclose is a straightforward, standard SATA III controller.

Given how Apple gave this away with the one XServe RAID model, I don't see them keeping such a "stack" as an in-house product when it is just a bunch of hard drives. There needs to be a brain inside that box for Apple to want to keep it as an Apple product.

There is no "brain" in the Thunderbolt display. don't think they are going to dump that as a product any time soon. No "brain" in the Ethernet or Firewire TB dongles either.

Pushing the drives sleds to another box with another set of fans & power supply is just a direct attach peripheral.

XRAID is a substantially different product. It is an independent system and not direct attached. It was not a simple peripheral, but a system unto itself. So yes it needed a "brain".