Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MP 7,1 Waiting for Mac Pro 7,1 A1991 (no more)

Boil said:
We still need three PCIe slots to meet the needs of the broadcast video world...
Click to expand...
Since it's being designed for high end AV, that makes sense. I don't know if broadcast has benn involved with the project like the film industry, however.
frankiee said:
Support for NVIDIA is mandatory!
Click to expand...
Says who? No it doesn't actually.

I don't know what the deal is between the two companies (nor do I care) but their respective headquarters are a bicycle ride away from each other if either side wants to do business with the other.
 
Boil said:
This is where I would see the "modular" aspect come in...
Click to expand...

If modular is primarily center around storage it probably not significantly like these "lego" brick concepts that have been floating around.

There could very well be four M.2 slots, but they would be for Apple-proprietary M.2 SSDs, all controlled by the T2 chip, and you have to install them in pairs...
Click to expand...

It is very unlikely Apple is going into some "sold in the store" blade SSD business. If there are secondary storage slots keeping those proprietary storage slots that doesn't almost nothing there for Apple.

First, there is almost no "volume". Entirely optional BTO components don't have a significant base volume. Apple putting a specific blade into every MBP 13" sold is an entirely different economies of scale equation. ( Case in point they have never sold these Apple specific SSD blades before in open market. Why would they start now? )

Second, all of Apple's internal SSD controller work is to incorporate the controller into a wider functionality system on a chip (SoC). Putting a T-series or yet another A-series derivative not only doesn't have scale it would probably cost even more from the solutions already out there. So no volume economies of scale or BOM win here. [ The option of just chucking the now old SSD controllers from the old SSD. .... How much is that going to help to put dated SSDs into the new Mac Pro ??? ]. All of Apple SSD controller work is away from SSD blades [ controller mounted to logic board].

Third, Apple providing "trimforce" means Apple already tacitly admits that folks will use non Apple SSDs. If buying a couple of several, providing a path for that is already there ( with Thunderbolt on the systems is always there anyway ). If Apple sells one SSD with every Mac they already have that volume , revenue, and profits. To throw it on the secondary one(s) borders on ultra ridiculous Scrooge McDuck myopia. They are going to loose more in aborted sales than they'll gain. It would be far more cost effective and cheaper to just spend the money to qualify one/two SSD M.2 supplies and perhaps slaps an Apple marked firmware on them ( like what they did early on. )


Some mix of M.2 (and maybe 2.5" to allow in lower $/GB SSD) is the only sensible path to "more internal capacity". It would be more cost effective for Apple too. ( since not creating some very low volume device in the "copious spare time" they don't seem to have anyway. )


That's how they get you...! ;^p
Click to expand...

They only "got you" after you bought the system. There are a significant number of folks who have the budget and want one big pile volume for all of their data. Those folks will buy the very high capacity default SSDs and Apple will live just fine off of that. However, there are also other folks with tighter budgets than Apple's inflated SSD $/GB pricing enables. They'd already be seeing a drop off is prune off the 3.5" and 2.5" crowd if they go that route. Making the customer base gratuitously even smaller won't help the system's ecosystem over the long term.

Four standard 2280 M.2 slots on secondary storage daughtercard with integrated RAID controller
Click to expand...

That's probably a punt to Thunderbolt (if there were no slots). Don't really need optional SSD slots where there is some requirement where have to fill more than one at a time. The Mac Pro 2008-2012 had four independent sleds. That would be entirely sufficient for a wide spectrum of use cases . Apple isn't likely going to sell a BTO RAID card with just one SSD on it. Nor a RAID card with the set to JBOD. Far cheaper would be just to put 1-4 M.2 slots on a PCI-e switch ( so they all appear independent. And primarily just get an "added aggregated" capacity uptick from it. Which is the main objective here "more capacity" not "more storage bandwidth speed.". ). [ Even more cheaper if just use the PCH chipset to be that chip.... but I'd suspect they'd want to manage the DMI bandwidth pressure; so hang that switch off of the CPU> ]

If someone wanted to stack that back up into a software RAID ( future version of SoftRAID) they could and Apple doesn't have to do much directly. ( primarily just like the Mac Pro 2009-2012. And they probably won't mind the hackery going away of folks doing redirects off of the "drive sleds" to 3rd party raid. )


For third-party add-ins (Blackmagic Design will partner to provide a 12G SDI 8K daughtercard at launch & BTO)
Click to expand...

That makes about zero sense. If Apple is trying for 3rd party, bus connected, add in cards intenrally ...... they don't need a proprietary slot at all. Perhaps it is only a x8 ( or x4 ) PCI-e legacy standard slot ( and nothing but bus power allotted ). And if have one of those and someone has a huge, burning need for a RAID card with M.2 slots on it... ta-da ... they'd have one.

That's RAID card above from "Apple". It would fit just fine an a x4/x8 electrical slot. So Apple doesn't have to compose it for sale. ( or even stock it if they don't want to. )
 
deconstruct60 said:
If modular is primarily center around storage it probably not significantly like these "lego" brick concepts that have been floating around.
Click to expand...

Again, I am not talking about stacking modules; I am talking backplane & proprietary daughtercards...

deconstruct60 said:
It is very unlikely Apple is going into some "sold in the store" blade SSD business.
Click to expand...

Yeah, which is why I shifted away from that to a T2 / dual Apple M.2 SSDs (for that faster parallel thing they do) for the secure boot / primary system drive...

deconstruct60 said:
The Mac Pro 2008-2012 had four independent sleds. That would be entirely sufficient for a wide spectrum of use cases .
Click to expand...

Yeah, had one of those with all sleds filled & the Apple (proprietary) hardware RAID card, set to RAID 5 with the fourth drive as a standby, was the main server for an eight person architectural office...

deconstruct60 said:
Apple isn't likely going to sell a BTO RAID card with just one SSD on it. Nor a RAID card with the set to JBOD.
Click to expand...

No, I would expect them to sell the M.2 RAID card unloaded & allow the end user to fill with normal priced 970 EVOs...

deconstruct60 said:
That makes about zero sense. If Apple is trying for 3rd party, bus connected, add in cards intenrally ...... they don't need a proprietary slot at all.
Click to expand...

I am pitching proprietary because it is Apple, and that would fit into a convoluted definition of modular, but still lock down a lot of the expansion to "modules" that are a higher profit for Apple, even if they are not exactly what the end user wanted (industry standard PCIe card usage)...

What I would LOVE to actually see from Apple would be a mini tower with a single large CPU socket, eight RAM slots, the T2 / dual M.2 SSDs primary drive we see in the iMac Pro, three PCIe slots (x16/x16/x8), four TB3 / USB-C ports, four USB ports, two 10Gb Ethernet ports, & one 3.5mm headphone jack... Maybe provisions for four 2.5" SATA drives...
 
Last edited:
Joe The Dragon said:
well with intel 1 cpu you are looking about 48 pci-e lanes
so X16 slot 1 / video.
X16 slot 2
gives you 32
and then 10G 4 for 2 ports
12 for 3 TB buses.

With all other IO + storage + TX on the DMI bus.

Maybe move dual 10G to pch and use it's X4 for one more TB bus or 4 more lanes into to the PCH.
Click to expand...

The three TB busses/controllers doesn't make much sense if going to allow some empty (in standard config) PCI-e standard slots. Two controllers ( 4 ports is enough). Getting into a port count pissing contest with the iMac Pro and MBP is really kind of silly and counterproductive. It won't drive substantially higher Mac Pro ales at all. The 6 count on the Mac Pro 2013 was about a goofy and the "sit and spin" light on the ports. Going up to 8 TB ports ( 4 controllers) would be just plain silly ( not here but on some of these recent concept.).


Here is a more sensible set up with same 48 lane budget.


option 1 ( one empty slot )

x16 primary GPU (apple slot )

x16 empty slot2

x4 ssd storage. Either one slot or PCI-e switch coupled to 2-4 M.2 sockets.

x4 PCI-e switch two 10GbE controllers

x4 TB controller "bus" 1

x4 TB controller "bus" 2

[ Essentially, that 'extra' x4 that a 3rd TB controller would soak up could be put to much better usages if going to have internal internal expandability at low-moderate additional power levels.

The two additional DisplayPort ports that the 3rd TB controller would have soaked up are sent to two external connectors as an alternative to Type-C output. For example, two MiniDisplay ports or one mDP and one HDMI. You'd still have the 6 monitor metric that Mac Pro 2013 was suppose to be saddled with. Just not OCD limited by the same exact port shape for all three. ]


option 2 9 ( two empty slots )

x16 primary dual GPU (apple slot)

x16 empty slot2 ( double wide )

x8 switch to (x4 phys) x2 10GbE and x8 electrical PCI-e standard slot ( just bus power. )

x4 TB controller "bus" 1

x4 TB controller "bus" 2

[ still could have two ethernet port if hand a 1GbE off the PCH chipset with little bandwidth pressure added to the DMI link in most contexts. It isn't exactly symmetrical in terms of bandwidth but would "look" the same if the industry design OCD would kick in on physical symmetry.

The 10GbE would add a small "bleed" on the x8 slot If the inserted card was hyper sensitive just don't use the 10GbE socket at all.

That x8 slot could be throw at SSD storage with a x4-x8 card so still coverage that path too for folks where that is a burning issue , but additional cost ]

In both cases, TB is primarily used to cover empty x4 standard slots. Dedicated M.2 slots would eseentially be used as replacements for the old HDD drive sleds.
 
If Apple decided to take the transition the long way, we could - so to speak - see ARM chips becoming more prominent and moving from just dealing with a few processes to being a fundamental coprocessor to an Intel Xeon (or AMD's Threadripper, as some of us wish) chip. They could run iOS apps faster and better, help developers test their products on one machine and perhaps even be upgradeable as new chips come out.

It is a dream since it would involve also Apple switching to AMD CPUs to have a better roadmap ahead and keep market share, which as well might lead to an even stronger bond between the two companies, keeping Nvidia farther away. :D

It would then give sense to having two daughter-cards one with the x86/x64 and the other with the ARM. And also buying a bloody expensive machine as very likely the mMP will be and not being able to sell it to anyone in 5 years (perhaps even earlier) because Apple has transitioned to proprietary chips. :D

Delirious scenarios aside, the latter can be a concern to a Pro user. It affects depreciation, therefore the value of the investment. The thought of waiting two (actually five) years to buy an expensive computer that becomes hard to sell very quickly, makes me shiver.
 
askunk said:
And also buying a bloody expensive machine as very likely the mMP will be and not being able to sell it to anyone in 5 years (perhaps even earlier) because Apple has transitioned to proprietary chips. :D

Delirious scenarios aside, the latter can be a concern to a Pro user. It affects depreciation, therefore the value of the investment.
Click to expand...
That's not the way depreciation is calculated in the US, at least in the business world.

A capital expenditure has a lifetime based on its function. For me, if I pay $130K for a server, or $10K for workstation, or $4K for a laptop - that cost is depreciated over 3 years (I'm billed for 1/12 of the amount each quarter). Other systems, like storage arrays and network gear, may have 4 or 5 year depreciation schedules.

The concept of residual value (resale value) simply does not exist. If I keep a server for 5 years, then "eWaste" it - I paid for it the first 3 years, then had 2 "free" years, and any money that the company can get back through resale or recycling goes into a different pocket.

Also our "resale value" is hurt by the fact that we physically shred all SSDs and HDDs. (And for the recent Apple laptops with soldered SSDs, that means that we shred the laptop.)

A prosumer who's making some money on the side and claiming depreciation on her taxes might be concerned about the resale value, but businesses assume zero residual value for the most part. (And, I suspect that claiming depreciation - and then reselling - is tax fraud unless you adjust the depreciation claims for the resale value.)
 
  • Like
Reactions: Nugget and barmann
anybody else waiting for the big reveal to be a slightly larger mac mini with user upgradable ram and hard drives? I have this feeling that they are not going to make us an actual computer that is designed to not thermal throttle and has actual expansion. If they dont make that computer we will know they dont care to give workstation users what workstation users need.
 
deconstruct60 said:
It has only primarily changed in hot rod, single threaded drag racing. The doubling of transistors hasn't really been that far off. For parallel computations the performance hasn't been off much. Software is more of the bottleneck than the hardware is. And that is the part that is far more subject to "rigidity" than the hardware is.

doubling the transistors doesn't mean performance is going to increase if the computation is stuck in some chokepoint. Brute force hammering things with clock speed doesn't scale over the long term.
Click to expand...

Fair point .
At the same time, hardware developers need to cater to what is available at any given time, like all manufacturers do .

This also ignores that making things smaller has also gotten more expensive. Not "mass producing" the same stuff with largely the same tools. The tools are getting more inexpensive. To get the same "economies of scale" means there are higher and higher thresholds to hit. It also means there are fewer competitors in the game so the price decreases due to that also play less of a factor.
Click to expand...

Why make things smaller, or use tooling and assembly station for proprietary tech and form factors, if it doesn't yield more affordable products and / or more performance for the consumer ?
That sounds to me like incompetence in product management , to use technology that isn't ready for an overall progressive approach , which would cover the arguably more important aspects mentioned above .

And while the underlying technology is a different matter, the form factor and size of standard components don't seem to have changed that much .
[doublepost=1558035776][/doublepost]
Manzanito said:
Maybe people don’t buy new computers at the same rate anymore? Also, a lot of people used to buy a computer just to browse the web and can get by with a smartphone.

Less sales will turn into less components being made, and thus more expensive to produce per unit.

Just a thought, I don’t have data to back any of this up, but could be a factor.
Click to expand...

Good point .
And while you can't really browse the web on a smartphone, you can on a tablet - which isn't a booming market anymore .
Laptops and other Lite-computers like iMacs have become plenty capable for many users though, maybe even most .

However, there are still boatloads of towers being sold for heavy lifting, flexibility, server use and renderfarms etc ., which makes the workstation a viable design still .
 
  • Like
Reactions: ssgbryan
To add to my theory at post #12165 (damn! 12k posts? LOL)...

And, to better align said theory to the rumor-mill of cube-this, modular that...

...I am thinking of a nomenclature change, also, in the horizon!

The Cube.

The Cube Pro.

What distinguishes the two apart are small, which saves on production and design cost. Big part are the architecture. One is based on Comet Lake S Mainstream Intel CPU's. And, the other is based on Cascade Lake Xeons....

Where the two merge is... perhaps changing on my original theory of PCIe slots... is the way GPU's or PCIe devices and expansion are handled... this is where a lot of the anticipation and wishful thinking seems to be mostly directed in regards to this thread. Ppl want the slots (for PCIe, RAM, SATA, NVME, etc)!

I have no clue if slots will return.

But, I think they will. Might not be PCIe standard slots. But, slots nonetheless?

The Cube will come with Comet Lake S CPU's with iGPU, which Intel has said will be better than prior Gen. So, I can see the Cube using just an iGPU. And where Apple can make money is the modular part to expand the Cube with a dGPU, etc. Might cost another grand to get a basic GPU and the expanded module!

The Cube Pro will have the expanded module already there with a dGPU since it's geared for "Pro's." But, of course, the base price will also reflect that. I am thinking $3000 for base, in the least.

The Cube with Comet Lake S i3 CPU (not soldered! Hooray!) and iGPU and 8GB DDR4 RAM (basically a big mac mini but comes with Apple KB and mouse) will be based at $1500....
 
I can see it now:

"After much effort, we’ve concluded the redesigned Mac Pro will not achieve our high standards and we have cancelled the project. We apologize to those customers who were looking forward to this launch. We continue to believe that the future is streaming video and are committed to push the Peanuts NASA Special forward."
 
  • Like
Reactions: sigmadog, Aldaris and AidenShaw
AidenShaw said:
If the "Nouveau Cube" doesn't have standard PCIe slots that support 300watt Nvidia GPUs, Apple should abort it now.
Click to expand...
I need PCIe slots. Lots of PCIe slots.

slots.jpg
 
  • Like
Reactions: wallysb01, Synchro3 and aaronhead14
CWallace said:
Not to mention the Cube, like the 6,1, was a market failure that put form over function so I would expect Apple would not want to refer to the 7,1 using the name.
Click to expand...

The Cube mostly failed because it didn't have a niche at all. A different situation than the 6,1.

filterdecay said:
anybody else waiting for the big reveal to be a slightly larger mac mini with user upgradable ram and hard drives? I have this feeling that they are not going to make us an actual computer that is designed to not thermal throttle and has actual expansion. If they dont make that computer we will know they dont care to give workstation users what workstation users need.
Click to expand...

I kind of feel like this is fundamentally a more sensical bet than all the more fanciful reads into "modular". But considering the iMacs and iMac Pros don't throttle (with the caveat that the iMac Pros are using slightly down clocked chips to make that happen) I think that's the least relevant criticism to throw at stuff.

If you want to criticize Apple with its performance, the issue has rarely been throttling, and much more potential longevity issues from running their machines hot and prioritizing quietness.

On a more concrete note—if the Xeon-W moves from quad to six-channel memory, does that mean we get a minimum of six DIMMs on the Mac Pro? What about the iMac?
 
filterdecay said:
...a slightly larger mac mini with user upgradable ram and hard drives? I have this feeling that they are not going to make us an actual computer that is designed to not thermal throttle and has actual expansion. If they dont make that computer we will know they dont care to give workstation users what workstation users need.
Click to expand...

While I am an advocate for the Nouveau Cube, I realize that industry standard parts would serve the end users better, so I am hoping for a proper mini-tower personal desktop workstation...

AidenShaw said:
I need PCIe slots. Lots of PCIe slots.
​
Click to expand...
​

I would bet on three PCIe slots, x16 / x16 / x8...

fuchsdh said:
On a more concrete note—if the Xeon-W moves from quad to six-channel memory, does that mean we get a minimum of six DIMMs on the Mac Pro? What about the iMac?
Click to expand...

AidenShaw said:
Apple may go with three. (The chipset in the 6,1 supported 12 DIMMs with a single socket - Apple went with four.)
Click to expand...

If Apple is using a CPU with hexa-channel RAM, then I would really hope they go with a minimum of six RAM slots...!
 
  • Like
Reactions: filterdecay and fuchsdh
Boil said:
While I am an advocate for the Nouveau Cube, I realize that industry standard parts would serve the end users better, so I am hoping for a proper mini-tower personal desktop workstation...

​

I would bet on three PCIe slots, x16 / x16 / x8...





If Apple is using a CPU with hexa-channel RAM, then I would really hope they go with a minimum of six RAM slots...!
Click to expand...

You think it will have PCIe slots? And user-servicable RAM? Wow... glad there are still som optimists around here.
 
  • Like
Reactions: Ifti
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back