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
Failure is a relative concept, consider beyond our likeness Apple design philosophy follows consistent trends.
I had hope the mMP at leas to offer DIY ram Upgrades and DIY GPU upgrades (with Apple-sourced gpu), but seems Cupertino prefer to lose some customers than Mac control.
Independent to what the leak said (consider it just speculation):
Here are few cues on the iMac PRO:
- RAM, while using std DIMMs, instead put a plastic window where user can access and replace it, Apple decided to drive you to an authorized partner to upgrade this, reason? APPLE TAX on CTO, avoid user ordering minimal configurations and at the same time memory from AMZ,Newegg, Macworld etc.
- SSD, Apple never adopted STD m.2 NVMe, even their NVMe are not easy to clone since still require the A10 chip to handle its encryption instead the NVMe controller (you need a custom NVMe controller to clone it).
- GPU, at least by now there is no cues on macOS to support native boot with new nVidia GPUs, so even a mMP with std PCIe GPU wont boot with nVidia.
Last edited:
Yes, the Pro and Mini.
Apple clearly states that they have a modular system already.
Apple defines their Mac linup like this: Portable, all-in-one and modular. Modular meaning, machine, keyboard, monitor and so on. They mention that they already have modular machine today so that would be the Mac Pro and Mac mini.
People have gone overboard with the whole modular thing cause it´s been a wet dream since the 80s with a computer that you can expand and expand as you need it. Whats not to like
And it would really be a great thing if it was realistic today. But with the volume of the Mac Pro today it would be insane to think Apple could maintain this in a sensible manner. The volume is too small. It´s a reason that those great sexy workstations from SGI and SUN and others went into history: they got to expensive. When a standard Mac or PC catched up to theese workstation and drove down the price they didnt stand a chance. As for Apple they have always been in the "Premium" segment but it´s a limit even for them (Us customers really) regarding cost/value.
This is the most promising clue about the new Mac Pro that we have today.
The reason all the RISC workstations died is that the P6 ate their lunch. They didn't "get too expensive" - they'd always been expensive.
Once the same level of performance was available from Intel at a much cheaper price, it was all over.
Yet Apple execs have openly admitted or allured both the Cube and the trashcans being failures. There is of course a broad range of qualities one would need to look at before claiming something a completely failure, in which case probably nothing never is. But in *relative* terms, since those form factors suffered criticism or didn't achieve what they were set out to do, if the next MP shares similar form factor again, then for an onlooker it is a constant reminder of the predecessors' short comings, which was, "form over function".
RAM; with the iMac Pro at least they have an excuse in tightening up the air pipes for better cooling, and the machine is already an AIO anyway. If like you said Apple prioritizes profits on really practical details like this then it doesn't take a genius to see the gesture as a middle finger.
SSD; this like the iMac Pro again, but actually may have a foot to stand on for the Mac Pro. If it takes a proprietary controller to provide better performance, considering it is where the boot volume resides, even considering the concerns on locked up alt boot ability, privacy, after warranty servicing etc, this performance gain is probably a welcome, and most users wouldn't mind or care if it means more speed.
GPU; Or perhaps it is because almost literally there are no nVidia Macs now so there is no need to release anything official (last one was MBP2014, and just an odd one really). Either way that was not what I was referring to, the detrimental factor isn't whether there are vendor choice, if you can choose from AMD or nVidia but they are both proprietary sanctioned connectors which only happens when licensed or cooperating with Apple, then it is an economic barrier for enough choices of cards to be readily usable for the platform. And this is not just GPU, but other PCI peripherals and interfaces which are more demanding than Thunderbolt, but too niche to bother doing more than one version other than a vanilla card.
raid 0 on 2 pci-e ssd's limited by an X4 link shared with an co cpu is slower then 2 pci-e x4 ssd's on cpu-pci-e
Consider in April'17 mea culpa they only blamed the thermal core thermal restriction not the tcMP lack of std components, so I mean they are not concerned about standard but Apple's flexibility to offer upgrades or new updates to the mMP, consider both the tcMP and the Mac mini by Apple catalogs are modular systems (I think they mean modular as NOT-AIO).
[doublepost=1520453920][/doublepost]
iMP' NVMe performance is not degraded by A10 processor, the co-procesor only holds the cryptograpical key for the Custom NVMe controller, STD NVMe controller requires the host to sent the encryption key thru PCIe, iMP approach beyond provide stronger safer encryption procedure (thanks to A10' s.enclave) it dont busy the PCIe channel neither system's RAM (maybe Next AMD based mac will feature this extent to the system RAM, since it a ZEN Feature).
The thermal core idea has a long way to go before it is fully depleted. One possible explanation is the trade-off between size and performance was calculated too much in favour of size. One inch more on each side of the triangle would potentially add 50% more airflow to the thermal core.
Then also why is the thermal core so empty. it would need equivalent peaks and valleys to expand the operational effectiveness. The most ineffective design for cooling is a flat surface. But there is adequate space inside the core for additional metal. The air at the centre of the core is touching nothing and doing nothing.
What they needed was turbulence and pressure controlled by a central element living just inside the thermal core. An Archimedes screw for the air to really make full contact with the heat transfer. This design borrows something from jet engine and leaving the core empty was asking for trouble.
I can safely say the cooling system is too small, too little metal and unfinished.
[doublepost=1520470419][/doublepost]
Apple has changed computing and everyone copies Apple.
Last edited:
In general, no.
E5-1650v2 (hex core) 130 watt
E5-1650v4 (hex core) 140 watt
You can find some specific cases where a v4 with similar performance uses less power than a v2 - but usually Intel will keep the power the same and increase the performance.
Almost all workstation/server vendors have a lot of experience with cooling 120 watt to 140 watt processors. Not much demand in the workstation space for "slower, but less power".
I like the heat sink on my Dell T3610 (E5-1650v2 - same CPU and chipset as the hex trash can).
![YH2R3-lg[1].jpg](https://forums.macrumors.com/data/attachments/668/668819-24f91beb3ae9777db225b0492f46ba65.jpg "YH2R3-lg[1].jpg")
Phase change heat pipes with two finned radiators - and the PWM CPU fan in the middle. Lots of airflow and turbulence.
Have you seen the ILM swap posted on the DELL forum ? It was created with the intention of breaking away from the DELL heatsink.
https://www.dell.com/community/Desk...T5600-T7600-heatsink-alternatives/m-p/5154768
This is the modular system that replaced about 20 Mac Pros at my workplace:
http://store.hp.com/us/en/mdp/business-solutions/z840-workstation
http://store.hp.com/us/en/mdp/business-solutions/z840-workstation
But not with that specific design from his "leak". All of the "close-ness" do not solve any problem for the user.
[doublepost=1520501163][/doublepost]
Nobody really cares about the moniker or what word to dub the machine, ultimately it is the function that matters. When looking at the Cube or the Trashcan, what people loathed wasn't how they *looked*, but that specific chassis was a result of its functions being limited by its form. This is what I mean. If the next MP's cross-section isn't in the shape of a square or a circle, but still maintain a sub-mini-ATX size chassis then it will just be the same. Because like the "leak" you offered, it doesn't have room to fit full sized cards, and that will be instantly recognizable right away.
Last edited:
Well, that is unequivocally untrue .
And just for the record, as you keep claiming otherwise, the tcMP is a türd .
That link isn't about replacing the Dell sink, it seems to be about adding aftermarket heat sinks when upgrading a single CPU T5600 to dual CPU. (The single CPU T5600 has a second, empty CPU socket - and the Dell mobo doesn't have the heatsink mounts compatible with "universal" aftermarket coolers.)
Last edited:
One thing its sure, given Both Xeon-W and Epyc are gigantic CPUs the next Mac Pro needless will be at least a 50% bigger even if Apple keeps the trashcan general design with an appropriated thermal-core (the solid-aluminum thermal core is out of discussion, at least its clear it wont back, at least not alone and as a "core" ).
Yes that is correct. It explains how to change the ILM so you can use any heatsink. The photos lower down show the Intel product installed in the machine. But it does not limit this design to Intel and Supermicro.
[doublepost=1520550057][/doublepost]
A guy at MIT wanted to make a lava lamp with dozens of feet of aluminium foil inside. This lamp linked to the cooling system of the PC. Surface area is what makes a machine cooler not physical mass. However, thin sheets of metal become damaged easily and dust is an issue.
You will notice that a really good heatsink is very heavy in the contact area and very large surface area at the extreme edge. This is because you dont really need air to cool a machine. Infra-red heat is EMR - electromagnetic radiation and it travels very nicely through the extreme conditions of space to heat the Earth. This is why oil and water can substitute for air as a cooling conductor. The only thing that matters is temperature difference and a physical property called "specific heat" of a material or substance. This is the reason why you often see a sandwich of materials in a good heatsink. One material wicks the heat away to outer parts of cooling system and the second or third material dissipates the heat into air / oil / water.
The 2013 MP does not have a good heatsink design. It would have ideally been two or three metals to push the heat around to the entire heatsink and then a forest of needles in the middle or a turbulence element to really heat up the air.
Maybe we need one of the oil billionaires to make a central thermal core out of gold and see if that solves the problem.
I really like the E7 processor. It can run at a very modest temp and still push out massive amounts of math.
[doublepost=1520551376][/doublepost]
I like having conversations with people who hold views different to mine. I believe there are PC copies of Mac Mini, iMac iPad and "Turdintosh" as you like to call it. Apple changed the way people looked at "necessary" hardware for a specific task. Sure it was a tough road for movie makers and 3D studio but we both know that number crunching has been the Achilles heel of the Mac for more than 10 years. Apple did not want to take the high ground in quad and octo socket machines and left it to people like HP / Dell / IBM. This much is clear.
Last edited:
People have gone overboard with the whole modular thing cause it´s been a wet dream since the 80s with a computer that you can expand and expand as you need it. Whats not to like
And it would really be a great thing if it was realistic today. But with the volume of the Mac Pro today it would be insane to think Apple could maintain this in a sensible manner. The volume is too small.
This is the most promising clue about the new Mac Pro that we have today.
This guy gets it.
I do remenber two year ago I calculated a possible thermal core for an hypothetical 750W tcMP (Xeon E5v4 + dual Polaris GPU), just switching to copper and a faster fan where enough, of course a TC in copper will be really expensive, a more conservative solution would be thermal pipes + aluminum fins, so the arguments from Apple about the thermal corner in the tcMP where BS, I believe simple they decided to wait until AMD got ready Epyc CPUs or there where really big issues with desktop development oversight.
There are also alternatives to copper as Aluminum-carbon nanotube constructs, exotic but seems a sound alternative, notwithstanding I believe the mMP will use std 2-phase heat pipe cooling .
It's easier to consider all arguments from Apple Engineering (and Marketing) to be BS.
[doublepost=1520564111][/doublepost]
You mean, like nearly every other PC/workstation on the market?
Apple has designed & built two thermal chimney workstations - the cube, and the cylinder. Both of them suffered endemic catastrophic failures as a result of cooking their own components. The cube destroyed its optical drives and top surface buttons, the cylinder, its graphics cards.
You have to question why anyone, inside Apple, or outside, would hold onto the belief that they understand heat well enough to do it again, and why anyone would buy another chimney-design product from them again.
I also think that they could have pulled off a newly updated tcMP if they had put their resources into it. I don't think they decided to wait for a CPU to arrive or something like that. I think they simply thought that having an iMac Pro would be enough.
They started with the discontinuing of the displays and promoted the Macbook pro as a workstation with external devices. At that time the iMac Pro would be in development as the replacement for the tcMP. It has all the features of the tcMP, faster i/o, faster Xeon, better graphics and a display built in. So in that vision, there is really no need for a standalone display from Apple themselves.
Here is what I think: When they launched the new MacBook Pro there was a massive criticism for lacking more than 16GB Ram and so on...It wasn't a machine for Pro´s was the mantra. Then there was the departure of the display business combined with the troubles with the LG displays and the lacking update of the tcMP which hadn't seen an update since 2013 caused the massive outrage that Apple does not care for the pros anymore... And they heard it, they would not in a million year had that briefing of the new Mac Pro otherwise. So they had to do it. If they had just stuck with their original plan launching the iMac Pro as the MP replacement everybody wanting a new MP would go on a wild rampage don´t you think?
As for a new Mac Pro, the biggest hurdle would still be the GPU. And I am curious what they will come up with. The old MacPro´s weren't exactly quiet so I don´t think they are going to go back to a straight classic Mac Pro approach with PCIe GPU´s with their own blowers. Even though this is just a rough concept (The GPU´s would blow the air downwards and so on) I think this is interesting. A well-constructed core of some sort with a GPU´s that would "hook" on to the core for cooling maybe?
http://www.palit.com/palit/vgapro.php?id=2823&lang=en
No, even w/o the mMP pre-announcement the iMP by itself would have been enough at least to hold FCP.X & ProTools rampage, the problem is Apple two years ago didnt foresee to need vr/ar/ai APPs in the iOS ecosystem, and those Apps requires for devlopers powerful GPUs, add this AMD delays on ZEN/POLARIS/VEGA development put the nails on the tcMP coffin.
The problem with Apple is they only foresee the things they invent, they use to dismiss market development until is too late, VR/AR/ML was almost an ambush to Apple's plan.
OK, that's thinking positive about Apple's leadership, I think they simple planned a spin on the tcMP to AMD, and delays on Zen/Vega put the tcMP update in limbo, then ML/AR/VR requirements where too much for the tcMP restricted TDP, an updated TCMP could only appeal to youtubers (with latest AMD drivers that allow both GPU to tandem as a single logic GPU) and Logic Pros, the same thing as the iMac, the problem still on ML/VR/AR simple, Apple still has nothing descent for ML/AR/VR workflows, while I think ML could be solved with external rigs (tensor processors managed thru TB3 connection), this is a single use case, and the work to adap the ML tools to this very optimal workflow is beyond Apple reach (even Google which is involven on this is on long way to have an integrated IDE working with external TPU rigs), with the time TPU rigs will dominate is the next step following GPUs in ML, but this do not cover AR/VR, even 8K video (I consider 8K video will never go mainstream and hold only as Master Copy format).
ALL THIS FAILURES HAS A TWO COMMON FACTOR: LEADERSHIP AND OVERSIGHT. (include the issues with MBP keyboard, SSD, and macOS naive bugs)
Exactly, the tcMP with updated CPUs (180W) is very short in TDP even for RX570 GPUs, the only way to Rise the tcMP form factor is with upcoming AMD PRO- APUs (16 zen cores + 32 vega/vega20 cores) which is aimed at HPC market, since this APU could be deployed as a 2P solutionn the tcMP could still hold DIMMs in the central board but GPU will be replaced by 2 APUs, it will deliver an awesome powerful and efficient tcMP, but I doubt Apple has this on sight (unless they plan to launch Two different Mac Pros, one mini MP (tcMP2) and a cheegrated revival with dual socket, multi gpu etc)
I bet Apple simple will release a MP with enough flexibility to host yearly updates w/o require more attention from R&D, likely having everithing propietary, maybe releasing the GPU interface as openSTD as much, so people could swap with non-apple parts with minimal issues, but I have no big hopes to see std PCIe slots for GPUs.