Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Trends toward modular Mac Pro
- Thread starter MiJuConcept
- Start date
- Sort by reaction score
I am having this sudden flashback to Silicon Graphics super-computing hardware in the 1990's when I was seriously looking at making a purchase of system running the same software that Google started on. They had this cool piece of hardware called a Hippi switch. The more this hardware changes, the more it stays the same.
Lets agree that hardware is totally controlled by chip makers. Apple is not in a position to make its own CPU at this time due to the lack of scale.
Lets agree that hardware is totally controlled by chip makers. Apple is not in a position to make its own CPU at this time due to the lack of scale.
Even 10GBe is looking old now. The cost of SFP+ capable switching hubs has really dropped. I would hope that Apple brings onboard a quad port. Then you can cluster with or without a switching hub.
I hope Apple isn't spending thousands of manhours and millions of dollars dreaming up some "clever" new workstation design. There's no need to re-invent the wheel when it comes to workstations and their components inside.
Does anyone ever wonder why Dell, HP, Lenovo and others just use a standard tower form-factor?
It works.
Does anyone ever wonder why Dell, HP, Lenovo and others just use a standard tower form-factor?
It works.
I hope Apple isn't spending thousands of manhours and millions of dollars dreaming up some "clever" new workstation design. There's no need to re-invent the wheel when it comes to workstations and their components inside.
Does anyone ever wonder why Dell, HP, Lenovo and others just use a standard tower form-factor?
It works.
Have you ever tried to source a computer part off the shelf that was even a little bit novel or interesting? Shenzen, China is full of small companies wanting to get noticed by corporations for their innovation and design. However, they have no capacity to deliver anything above 10K units per month. Even 90% of the companies are just shopfronts for one of the 5000 factories scattered across China.
Lets take another perspective on how people control design. Mercedes Benz owns the seat factory (Recaro) and the brakes factory (Brembo) and it has a stake in the engine factory (Mistubishi and AMG).
Apple has saved up its money to plunge deeply into other things. The workstation business is not their main focus any longer. For that reason the 2013 Mac Pro will be one of the last self contained designs. Even Apple has conceded it no longer controls the design of computers and their best hope is to work towards a central processing core with a range of configuration options and allow the customer to decide.
The only sad part here is how much has been forgotten in hardware design compared to the absolute peak of quality demonstrated by American mainframe companies.
Wow, and most of the HIPPI links were almost as fast as USB 2.0 !
SFP+ is definitely cost effective - I have a couple of hundred 10 GbE ports, and it works out to under $1000 per port.
I like having all of my data channels at 10GBe or higher. The amount of time it saves is mind-blowing.
But that's exactly what I'm talking about.
These are supposed to be workstations.... not works of art.
The trashcan Mac was beautiful to look at... but it had MANY shortcomings when it came to being an actual workstation.
Let's save "novel" and "interesting" for consumer products.
(ironically... Apple's consumer iMac line has basically looked the same for over a decade... except for the time it got thinner....)
Yes... workstations are a tiny part of Apple... (and Apple itself a tiny part of the broader PC market)
But Apple still has to deliver a capable product to their users.
Look... I think the concept of the "Stackintosh" is cool. It's gotta be better than the trashcan. But I still think it's a little too outside-the-box... (literally)
If you polled Mac Pro users... I bet most of them wish Apple had kept the Cheesegrater and just updated the internals... rather than start over with some crazy new design.
Some things don't need radical changes.
But hey... any updated Mac Pro is better than NO updated Mac Pro, right?
I would bet that many who would say "keep the cheesegrater" would also be happy with an updated design.
The cMP is *huge* for what it actually provides. Keep the size (but trim those uncomfortable handles so that it fits in a 19" rack without using a hacksaw) and increase the number of PCIe slots and internal drives. (And put PCIe auxiliary power for GPUs on jumper cords directly from the power supply.)
And of course do something about the woeful lack of PCIe slots.
And (or) make a smaller (say Z-6 sized) model.
The cheese grater doesn't look good when compared to other current offerings - it only looks good when compared to the trashcan or the MacPro on a stick.
Last edited:
Crazy People (1990) with honesty in advertising.
Through all of the difficult times I consider Apple to be the most consistent and reliable company over the past 30 years.
Here is my contribution
Buy a Mac Pro, not the new one, the old one, because with this computer you have choices and being a consumer is all about choices, even though the expansion slots will almost always remain empty since there are hardly any useful cards to buy.
The Old Mac Pro ... being powerful is a choice.[doublepost=1520298876][/doublepost]
My personal opinion is, yes I agree. However, the "Stackintosh" and the "Jet Engine" Mac Pro both came from the same person.
Keep in mind that external PCIe cages are nothing new. They are a staple product in the Bitcoin community. By allowing Apple workstations to embrace expansion cages they also allow for clusters of GPU processing to facilitate the demands of numerical processing.
Stunning ad for a Volvo
Through all of the difficult times I consider Apple to be the most consistent and reliable company over the past 30 years.
Here is my contribution
Buy a Mac Pro, not the new one, the old one, because with this computer you have choices and being a consumer is all about choices, even though the expansion slots will almost always remain empty since there are hardly any useful cards to buy.
The Old Mac Pro ... being powerful is a choice.
But hey... any updated Mac Pro is better than NO updated Mac Pro, right?
My personal opinion is, yes I agree. However, the "Stackintosh" and the "Jet Engine" Mac Pro both came from the same person.
Last edited:
Agreed! What if it was possible to make something smaller, more powerful and yet also fully modular using off the shelf parts? Hmm.
A SFF system would be interesting, but I don't see any chance of the three Amigos being that practical. See http://www8.hp.com/us/en/campaigns/workstations-desktop-mlp/index.html to see a range of compatible systems from mini to dual socket.Agreed! What if it was possible to make something smaller, more powerful and yet also fully modular using off the shelf parts? Hmm.
I'd like to see:
- A cheesegrater/Z8 sized system with dual sockets, seven to nine PCIe slots, 2 TiB RAM, up to 1.8kW power supply
- A midi-tower/Z6 sized system with single socket, five PCIe slots, 512 GiB RAM, up to 900 watts (base is 650 watts)
Last edited:
That's a lot of slots. What would you need so many for? Couple of GPUs, a 10gbe card, and a couple of exotic expansion cards and you're only barely reaching 5.
Who knows what the future holds?
And isn't that one of the main benefits of PCIe slots - you can plug in cards for stuff that wasn't available when you bought the system? And forget the 10GbE card - that's on the mobo these days (but you might want a 25GbE or 50GbE or 100GbE card).
How about storage? My home Dell T-3610 (same CPU/chipset as the MP6,1 hex-core) has six 1TB 850 EVO internal drives, a DVD-ROM and a BD-RE. (Two of the SSDs are on the PCH 6Gbps links, and a PCIe card with quad 6Gbps links for the other four.)
It also has an LSI 8 port SAS/SATA RAID controller with 1 GiB writeback cache and 8 disks, and an LSI 16 port SAS/SATA JBOD controller. That's an external 24 drives, 144 Gbps, 96 TB of data. (Actual user space somewhat less, because of hardware RAID-6 and RAID-5 volumes.)
The point, though, is that although I didn't realize that I'd need 100TB of disk on my home PC - since it had free PCIe slots it was a simple matter to attach dozens of additional disks.
Last edited:
Apple still needs to push out into the darkness and bring back something beyond expectation. Putting an Apple logo on a PC isnt what this is about.
Apple shot for the moon with the 2013 Mac Pro. Nobody was expecting the trashcan Mac. But it landed with a thud.
And it sounds like you expect Apple to go even more crazy with the shape, design and function of the next Mac Pro.
Can Apple deliver another "alternative" to the traditional workstation? Is that what people want? I thought people's complaints about the 2013 Mac Pro was it was too different than any workstation they've ever used before.
I've heard pros on podcasts talking about wishing Apple would dip a toe into the Hackintosh community. Perhaps by letting MacOS officially work on 3rd-party hardware. That's essentially putting an Apple sticker on a PC.
I know Apple will never do that... but Apple hasn't really been a good steward to the Mac workstation either.
There are a couple trains of thought here. As you said earlier... the Mac Pro is a tiny part of Apple's business.
On the other hand... it's still a very important part of their entire operation.
People can spend upwards of $10,000 on a Mac Pro... so they expect it to do whatever they want.
There's not a lot of wiggle-room for getting crazy.
I feel bad for Mac workstation users. You get the MacOS that you love... but you get saddled with awkward hardware choices (or sometimes no choice at all... or no updates for YEARS...)
Or you can use Windows... and hate the OS... but you can get any damn hardware you want!
Apple needs to be somewhere in the middle.
Apple shot for the moon with the 2013 Mac Pro. Nobody was expecting the trashcan Mac. But it landed with a thud.
And it sounds like you expect Apple to go even more crazy with the shape, design and function of the next Mac Pro.
Can Apple deliver another "alternative" to the traditional workstation? Is that what people want? I thought people's complaints about the 2013 Mac Pro was it was too different than any workstation they've ever used before.
I've heard pros on podcasts talking about wishing Apple would dip a toe into the Hackintosh community. Perhaps by letting MacOS officially work on 3rd-party hardware. That's essentially putting an Apple sticker on a PC.
I know Apple will never do that... but Apple hasn't really been a good steward to the Mac workstation either.
There are a couple trains of thought here. As you said earlier... the Mac Pro is a tiny part of Apple's business.
On the other hand... it's still a very important part of their entire operation.
People can spend upwards of $10,000 on a Mac Pro... and they expect it to do whatever they want.
There's not a lot of wiggle-room for getting crazy.
I feel bad for Mac workstation users. You get the MacOS that you love... but you get saddled with awkward hardware choices (or sometimes no choice at all... or no updates for YEARS...)
Or you can use Windows... and hate the OS... but you can get any damn hardware you want!
Apple needs to be somewhere in the middle.
There are several Compaq servers which are alleged to have Apple compatible Micro-code. Meaning they would accept and operate with MacOS. I believe this was more to do with VMWARE authenticity of virtually operated systems. When your motto is "Think Different" you cant exactly run away from that history and take a seat in the audience.
From all of the arguments I have heard so far I still fail to recognise the point of difference. the HPE DL980 G7 is about 8 years old. Yet it still appears on the first page of the Corona benchmark 1.3 test sores and if you remove the duplication, it is one of the top 6 computer platforms. Keep in mind this score is with really ancient CPU technology and is not assisted by any GPU cards (about 8 slots free). The Generation 10 hardware goes to another solar system.
By the way, I was reminded by your comments regarding moon. HP Moonshot server.
When people say they want something new, it does not mean it has to be Xeon Gold or Xeon Platinum driven CPU architecture. It can be a really good implementation of a very solid and scalable architecture.However, nobody makes those chipsets now and buying into older hardware is a production nightmare. So who is at fault here ? For me, Apple is not deciding any more how computers are constructed (chipset, CPU, bus controller). Those choices are thrust upon them.
I just doubt that Apple, despite having the capacity, would be interested in breaking envelops while *still* offer a Mac Pro with substantial performance. We need to take into context of the situation, they have stagnated with the trashcan form factor, almost looked like they didn't care until last April, and suddenly they primitively announced that they have to play a catch up game, including VR/AR/AI/ML which Macs are laughably behind. Given how much effort they have paid on Macs post-Jobs, especially desktop Macs, I would consider them delivering "just a box" is already the max that they should bother doing for now.
The other perspective is also pointed out by others in this thread, that Apple doesn't control much if any of the technologies that goes into a workstation-type computer. You can see them desperately trying to own more and more component manufacturing and patents in mobile segment, but not on traditional computers. Therefore, a truly modular and scalable design is really difficult to conceive since the number of configs and anticipation to future industry trends is limitless. In essence this is what took the tcMP down, and arguably it "only" lacks PCI slots. If the new MP has even more proprietary inter-connectors that's stuck at a given bandwidth, then we are looking at even more chances for the overall system to be obsolete.
It seemed like they were pretty invested in shifting Macs into an iDevice, in both design philosophy and logistics. If the tcMP didn't flop, then the iMac Pro would have been more than enough of the top end for now, then a few years down the line perhaps we will see ARM-only Macs, running x86 over emulation. But as fast as Apple has progressed in the chip segment, it is not yet enough to be put in even a consumer Mac, let alone a professional / workstation Mac.
The other perspective is also pointed out by others in this thread, that Apple doesn't control much if any of the technologies that goes into a workstation-type computer. You can see them desperately trying to own more and more component manufacturing and patents in mobile segment, but not on traditional computers. Therefore, a truly modular and scalable design is really difficult to conceive since the number of configs and anticipation to future industry trends is limitless. In essence this is what took the tcMP down, and arguably it "only" lacks PCI slots. If the new MP has even more proprietary inter-connectors that's stuck at a given bandwidth, then we are looking at even more chances for the overall system to be obsolete.
It seemed like they were pretty invested in shifting Macs into an iDevice, in both design philosophy and logistics. If the tcMP didn't flop, then the iMac Pro would have been more than enough of the top end for now, then a few years down the line perhaps we will see ARM-only Macs, running x86 over emulation. But as fast as Apple has progressed in the chip segment, it is not yet enough to be put in even a consumer Mac, let alone a professional / workstation Mac.
Last edited:
The "Think Different" campaign ran from 1997-2002. It hasn't been used in 16 years.
A lot has changed since then (and especially since 1984...)
I'm coming at this discussion from the types of people or companies who spend thousands of dollars on hardware for work.
There's a reason the other guys still make tower workstations:
They work!
I don't think anyone complained when this year's HP Z-series workstation looked almost exactly like last year's model. That's not something those kinds of customers worry about. They just want it to work.
I think if Apple wanted to experiment with wacky hardware designs... they might wanna try it with their consumer computers.
You don't want to buy a workstation (or a whole lab full of them) and it be a dead-end.
Maybe Apple will come up with a new version of the stackable concept... CPU-block, GPU-block, storage-block, etc.
But Apple needs to support it with updated blocks for a specified amount of time. (either themselves... or allow 3rd-parties to officially play along)
Otherwise it'll be the same situation the 2013 Mac Pro ran into: no updates for years.
Apple can't take another 4-year vacation from the workstation market!
Silicon Graphics Onyx
Things only go around in circles. I cant explain it any other way.
I don't know what that Silicon Graphics illustration is. I'm obviously not familiar with all these mainframe comparisons. Forgive me if I misunderstood what your initial post was all about.
I thought you were referring to the idea of a future Mac Pro with a main body for the CPU and RAM... and then separate (but stackable) GPU and storage modules that can be replaced as technology advances. That actually sounds good, in theory.
My point was... for the last 4 years Apple's engineers have supposedly been in the lab cooking up something new and exciting. And we're still waiting to see it.
But for the last 4 years... the other guys have been continually releasing new and upgraded workstations for their customers. No crazyness... just functional. Dell and HP might not make beautiful computers... but at least they're releasing them. And they have the ports and slots people have come to expect.
And that's what I said in my first post in this thread: I hope Apple isn't spending a lot of time creating some crazy concept that their customers won't accept.
Apple took a big risk with the cylinder Mac Pro with its triangle thermal core. And it pretty much backfired. They cannot do that again. Schiller already put his foot in his mouth with his "Can't innovate anymore, my ass!" comment.
So maybe Apple needs to calm down and stick to a more traditional workstation.
That's not a mainframe. The Silicon Graphics Onyx is a 20+ year old workstation designed when workstations were innovative designs & more than just big PCs.
Apple expressing themselves creatively is what gave us the 2013, a machine which restricted their customer's abilities to express themselves creatively.
That is basically a miniaturized server rack cabinet approach, to put in today's terms. The difference is that each of the modules still use their own respective I/O, instead of having a universal encapsulating standard, such as how Thunderbolt 3 or USB-C is shaping up to be. This method has some merits over truly externally stacked modules, since the overall chassis has a common compartment to house cables and shared PSU individually.
However, when slots are that integrated into the chassis, this begs the question if there is even a need to standardize each modules physically anyway, there are just a lot of unused space for way too many configs. The ATX tower or even the Cheese Grater approach has minimized the possible configs to a reasonably manageable number, and then split the case in sectors/zones with differing space (wasted). The best thing that comes with this approach is zonal cooling.
Apple thought even that approach was way too ungraceful, then came the tcMP which eliminated the 2 places that waste the most space, 3.5" bays (rightfully so), and PCIe slot (pretty dumbly in retrospect). Furthermore, it is difficult to even begin to guess the form factor now, without knowing first and foremost what direction Apple wants to go with regards to their pro desktop lineup.
The bottomline, it is seriously difficult, if not fruitless to attempt a form factor that isn't a vanilla tower for now. Maybe have a few daughter boards instead of a single large mobo, sure, but trying to modularizing functions semi-externally will serve quite little benefit in exchange of a lot of engineering headache.
However, when slots are that integrated into the chassis, this begs the question if there is even a need to standardize each modules physically anyway, there are just a lot of unused space for way too many configs. The ATX tower or even the Cheese Grater approach has minimized the possible configs to a reasonably manageable number, and then split the case in sectors/zones with differing space (wasted). The best thing that comes with this approach is zonal cooling.
Apple thought even that approach was way too ungraceful, then came the tcMP which eliminated the 2 places that waste the most space, 3.5" bays (rightfully so), and PCIe slot (pretty dumbly in retrospect). Furthermore, it is difficult to even begin to guess the form factor now, without knowing first and foremost what direction Apple wants to go with regards to their pro desktop lineup.
The bottomline, it is seriously difficult, if not fruitless to attempt a form factor that isn't a vanilla tower for now. Maybe have a few daughter boards instead of a single large mobo, sure, but trying to modularizing functions semi-externally will serve quite little benefit in exchange of a lot of engineering headache.
Last edited:
You have seen it - the iMac Pro is the thing they spent the past 4 years working on. That's the thing that was going to be the only mac pro going forward. The new "modular" thing was in all likelihood not even an initial spec wishlist when the amigos chat happened.