T2 doesn't have anything to do with protecting system files. That's a different class of technologies.
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.
MP 7,1 MPX module - a proprietary dead end?
- Thread starter Varmann
- Start date
- Sort by reaction score
T2 doesn't have anything to do with protecting system files. That's a different class of technologies.
What we know now about details of the MP7,1 amounts to little more than "elevator photos" from MacWorldSF2019. It will be months before those questions can be answered (although there may be many guesses along the way).
These encryptions don't integrate with Filevault though. There's no preboot security to ensure only an authorized user has access to the drive.
T2 machines move the login screen into firmware so only authorized users can mount the drive.
But isn‘t the T2 checking the OS checksums and if verified, you can boot from the SSD?
"Dongles" is more than a bit overblown. Type-C has been part of the DisplayPort standard since 2014
https://www.anandtech.com/show/8558/displayport-alternate-mode-for-usb-typec-announced
A DisplayPort cable with Type-C on one end and a "classic" DisplayPort connector on the other has DP standard defined connectors on both sides. How how merits the 'dongle' adjuective is huge stretch. It is like a USB Type A to mini-B being some kind of "dongle". It is a cable, not a dongle.
Thunderbolt from the beginning has had a DP pass-through, backwards compatibility mode for DisplayPort streams. You've always been able to hook a DP standards cable up to a Thunderbolt port and get just DP output. When TB moved to version 3 and standardized on Type-C then that basically matched DP's adoption of Type-C also as a standard compliant port.
For the new Mac Pro any one of the four standard TBv3 ports can deliver DP if you plug in a DP cable. It wouldn't be a direct plug to the MPX 580X card though.
Just need a different cable. If coming from TBv2 (and just prior era) where Apple used mini-DP ports then can't re-use that cable, but a new standards compliant cable will work.
I wouldn't bet on that. Putting at least one instance Vega 7 in the eGPU certified list would help a number of other Macs. And it is basically the same hardware so 99% of the driver work is done. ( Apple would have to remove the Infinity Fabric support and a few other things but the core of the whole set up is exactly the same base die. It simply wouldn't be that hard and getting other Macs to leverage the MPX modules would likely be substantively harder. )
Apple knee-capping the whole rest of the Mac line up to goose a cards for a super low run rate Mac Pro doesn't make much sense at all. ( Loosing money on sales of the rest of the line ups value prop just to goose MPX modules that even fewer can afford isn't going to work well long term. Gallons of Cupertino Kool-aid isn't going to change that. )
There is a substantive Metal difference between the Vega II MPX options and Vega 7. That's the support for Infinity Fabric. Apple has a functionality gap to do product separation/segmentation.
Technically, no. The T2 is coupled in a transitive way.
T2 checks and protects the Firmware. ( folks won't be able to hack the firmware anymore )
Signatures checked and only a copy of the firmware presented to the CPU for boot.
The EFI Firmware checks the OS.
( page 8 here in the "Secure Boot" section https://www.apple.com/mac/docs/Apple_T2_Security_Chip_Overview.pdf )
So to say the T2 is completely decoupled from the OS check is technically wrong. It just isn't a direct connection but it is part of the secure chain of the boot. IF folks don't like what EFI ( not the bootstrap file that ends
'.efi'. ) is doing then there are only fixed parameters that can be changed.
With SIP on , an unmodified EFI then the system files are pragmatically protected from several attack vectors from below. If boot onto a foundation that is insecure you usually can't fix it on top.
Apple does security using layers of defense. There is no singular magic security mechanism protecting the system files. The T2 is involved in just one of those defense layers during boot.
It also protects at rest (with encryption). And again transitively involved in securing against rogue external boots.
I like it, but I've been with Apple long enough to know there will be a few surprises; some good while some not so much. Now that I'm older,' I use productivity software that addresses my medical needs and research. Most of these apps are semi-AI based, e.g., voice-driven, document management, teleconferencing, and others' that are more robust on Microsoft Windows vs. macOS.
Also, some hobby software that I like is just not available on the Mac. Currently, I'm debating between the new MacPro rack-mounted option and a Trenton Systems rack-mounted workstation. Both hover at the $10k~ price point.
If I were ordering today, I would opt for a Windows based Trenton. No, I never thought I would say that. However, once more information becomes available on the MacPro, I may take another gamble with Apple. My main concern with the MacPro is the proprietary system(s) that Apple likes to employ.
Finally, Apple seldom provided meaningful updates to the current MacPro. Hopefully, that will change. I get the impression Apple is more interested in wearables, and consumer-based gadgets for the younger demographics, than anything business-oriented. There has been a few companies and government agencies that tried Apple products, only to return to Windows e.g. FoxNews, Airline carriers, and the military.
Also, some hobby software that I like is just not available on the Mac. Currently, I'm debating between the new MacPro rack-mounted option and a Trenton Systems rack-mounted workstation. Both hover at the $10k~ price point.
If I were ordering today, I would opt for a Windows based Trenton. No, I never thought I would say that. However, once more information becomes available on the MacPro, I may take another gamble with Apple. My main concern with the MacPro is the proprietary system(s) that Apple likes to employ.
Finally, Apple seldom provided meaningful updates to the current MacPro. Hopefully, that will change. I get the impression Apple is more interested in wearables, and consumer-based gadgets for the younger demographics, than anything business-oriented. There has been a few companies and government agencies that tried Apple products, only to return to Windows e.g. FoxNews, Airline carriers, and the military.
Last edited:
Can I use the MPX modules with the 580, buy a different GPU and plug that into the MPX module? Do I get TB3 support like that? Why is the 580 in a MPX module at all?
Because the MPX slot replaces pin power connectors off of the PSU.
Yet another solution in search of a problem.
The 580X GPU chip is in a MPX module most likely because AMD didn't have anything better in 2018 when the work for the module was being finished off. The (NAVI) 5700 series was late ( slid out of early 2019 release) and slightly higher power. (if went XT). The 580 worked. A variant already worked in a another Mac so it should be a rock solid solution. That also extremely likely makes it affordable.
For a decent number of workloads ( audio focused) it will perform just fine.
The GPU MPX modules are just like a standard GPU card. They not modular themselves. The "duo" is more sexy so there are pictures floating around of that.
https://www.tomshardware.com/news/amd-radeon-pro-vega-ii-7nm-gpus-apple-specs,39571.html
The 580X module is just more simpler version of that ( one smaller GPU chip. GDDDR5 memory , only two HDMI ports on the card. no PCI switch. Lots fewer capacitors. etc. ). Still has a MPX connector and a x16 standard connector. The cowling that goes on top of the 580X is only half as tall as the one for this DUO.
It is not a wrapper around a complete card. The 580 is a GPU chip soldered to the board as GPU chips are normally solder to some board.
You can buy another GPU card and put it into a MPX bay. The 'bay' is not the 'module'. Each MPX bay has PCI-e connectors and some molex power sources that can be optionally used. But that doesn't involve a MPX module.
If Apple comes up with another 'entry' MPX module later, you'd be able tp pull the 580X one out and put in the new one. Or place the new one in that 2nd MPX bay.
Chuckle. And how many threads on this forum about getting 6-pin to 8-pin hacks for the Mac Pro 5,1 to put upper range Metal qualified cards for Mojave upgrades.
MPX module that is 75W or 400+W ... all just plug in and go. No hacks needed. No need cables needed. No adapters needed.
It is a different solution. But the "problem" is more so widely tolerated rather than not being an issue.
If there is no MPX module in the MPX bay then there are alternative power source for regular cards. So there is no "because the MPX slot" there at all. The GPU chip isn't replaceable on normal GPU cards any more than it is on the GPU MPX modules. They come on whole cards. If get a new one, then can pull the old one out and insert new. There is nothing materially different here in terms of power.
Provisioning DisplayPort feeds for the standard Thunderbolt sockets on a Mac Pro is different. But that is a different dimension than power. There yes, need to buy another MPX GPU module to keep provisioning the TB socket(s).
The hacks are necessary because Apple put the mini 6-pin connectors on the main board instead of directly from PSU. That was itself a solution in search of a problem.
Does anyone remember the additional proprietary power connector slot on the video cards in G4 Macs?
MPX module could have been easily designed to accept power from normal 6/8 pin plugs.
Doing different just for different’s sake.
Just like MP6,1 was a redo of the G4 Cube that was itself a complete commercial failure.
The 580X GPU chip is in a MPX module most likely because AMD didn't have anything better in 2018 when the work for the module was being finished off. The (NAVI) 5700 series was late ( slid out of early 2019 release) and slightly higher power. (if went XT). The 580 worked. A variant already worked in a another Mac so it should be a rock solid solution. That also extremely likely makes it affordable.
For a decent number of workloads ( audio focused) it will perform just fine.
The GPU MPX modules are just like a standard GPU card. They not modular themselves. The "duo" is more sexy so there are pictures floating around of that.
https://www.tomshardware.com/news/amd-radeon-pro-vega-ii-7nm-gpus-apple-specs,39571.html
The 580X module is just more simpler version of that ( one smaller GPU chip. GDDDR5 memory , only two HDMI ports on the card. no PCI switch. Lots fewer capacitors. etc. ). Still has a MPX connector and a x16 standard connector. The cowling that goes on top of the 580X is only half as tall as the one for this DUO.
It is not a wrapper around a complete card. The 580 is a GPU chip soldered to the board as GPU chips are normally solder to some board.
You can buy another GPU card and put it into a MPX bay. The 'bay' is not the 'module'. Each MPX bay has PCI-e connectors and some molex power sources that can be optionally used. But that doesn't involve a MPX module.
If Apple comes up with another 'entry' MPX module later, you'd be able tp pull the 580X one out and put in the new one. Or place the new one in that 2nd MPX bay.
Thanks for your detailed reply.
There probably will not be a slew of GPU MPX modules. There were not a slew of "Mac" cards in the 2006-2012 Mac Pro when the form factor was closer and all primarily needed was specific mac firmware. The likelihood that someone is going to make something different is slim. ( only if Apple prices themselves too far out and goes back into Rip van Winkle mode for a long time). Unless AMD ( or Nvidia is that door cracked open again) steps up and does a reference card for them, it is highly doubtful any of the GPU makers will step up and do a card that isn't relatively simple derivative of the reference card.
So it isn't just Apple. If the GPU chip vendor isn't proactively participating, then it is a mostly dead market. But while Apple didn't "build from scratch" the cMP GPUs cards , they were in the loop in getting the adjust firmware made and certified. They did put resources/money into them.
The MPX bays have 8 pin power supplies and standard PCI-e slots in them. The MPX socket can probably be ignored by regular cards. So if there is close to zero R&D to sell a 'standard' card and substantive R&D to do a MPX specific one .... almost all of the larger GPU card vendors are going to choose the former. Lower risk and lower cost. Even if there is some minor firmware tweaks to make to the standard cards, that "hacked" card market is already exists. ( that "hacked" card market also keep the "Mac" cards from getting a slew of vendors participating. It will likely continue to suppress direct OEM solutions going forward too to a lower degrees as mainstream GPUs get easier to holistically boot. )
That cross compatible status is a further suppressant to lots of vendors jumping in.
Thirdly, the number of these new Mac Pro sold is probably going to be substantially lower than the old 2006-2012 modules sold.... and that older one wasn't a viable enough market to bring a slew of Mac GPU card makers. This is one will be even smaller over the first 2-3 years. Eventually there will be a base of 'older' systems to sell into, but the base will be relatively very small starting off. [ The lower volume doesn't impinge Apple as they get a shot a BTO additions for every purchase for the Vega 2 and primarily basically charge for at least a portion of the 580X in every Mac Pro sale. Apple won't directly sell any GPU up against these on same purchase order. The issue is how much is left over at that which can only be filled by a MPX solution. ]
For MPX to be a usable solution it can't just be GPU cards and there just can't just this one Mac Pro instance for 4-6 years. A baseline to build multiple systems with over time will bring diversity over the long term. Apple needs to clearly demonstrate that they aren't going to 'flake' on the design to bring a broad market of solution to life.
Storage there will probably be multiple competitors ( I highly doubt it goes past three though). But Apple isn't sitting there with a first party product at all. [ the "screw to frame and plug into SATA sockets will probably have several competitors. Zero electronics there (or very close to zero if simple SATA daisy chain for more than two drives with zero RAID electronics. ). ]
Something like Afterburner could be another. ( e.g., a ASIC or FPGA card that can be relatively seamlessly tapped into ). That would take some outreach by Apple into some new spaces. (and 'open' way to tap into the thermal control feedback system) , Future MPX bays with faster PCI-e "backplane" would open up more spaces for different modules. But low volume and expensive card would be a better match to supply/demand.
MPX can grow to be a usable solution. It extremely likely won't be a short-intermediate term winner. Growth here will be slower than Thunderbolt. Probably at least 3 years until see if there is traction.
The other uptick would be some vendor like Blackmagic adopting it for their eGPU enclosure. that would help bump the base market size a bit more. Same with another system vendor adopting it (and Apple being open to its use. )
The main problem with T2 is that you can't upgrade your main internal boot SSD. Ever. That's what people hate about it. You're stuck with what you got, and that's not what Mac Pro users want. The whole point of having a modular machine is to have the flexibility to upgrade anything.
For the iMac Pro and new Mac Pro that isn't technically true.
If you have a NAND daughter card failure, they can be replaced. That replacement path will negate the "ever" aspect there over the extended long term. Eventually there will be boneyard parts off of decommissioned systems. Even an Apple tool to reinitialize the driver (if required ) will eventually "fall off the back of a truck" and show up in some markets.
There won't be commodity , local electronics store driven upgrades, but they'll be some upgrades ( no universal, external application )
It is more so commoditization (and 'control' ) cloaked as modularity that what is being grumbled about. It is modular. It has subcomponents that detach and attach.
Nothing is going to meet "anything". The power management and system management ICs on the all the previous Mac Pros were soldered to the motherboard. Nobody was having a heart attack over that. Practically, none of the systems have that feature either.
I'm not worried about the SSD. But it has been noted that the tool requires a GSX account to operate, and does the authorization through Apple's backend. That would preclude that tool from "falling off the back of a truck." Or if it does, it won't be useful.
I meant a few in the relative space of the pro creative work, i.e. sound cards, video capture etc. It's a small space and will always be one. As you said, the PCI-E cards will already fit, but my guess would be, that some manufactorers will seize the opportunity to use the head room of MPX and potential ease-of-use of just pluging the card in without worrying about the correct cables (Macrumors is full of them threads).
That's all I saying. It will most certainly not have the size of the consumer / gamer market, I grant you that.
The problem is that you're ignoring that it's also routing DisplayPort for the TB3 ports as well. There's no real good commodity solution for that right now. About the best you get is to run a TB3 PCIe card with external DisplayPort inputs and use external cables to route the signals to the card. No internal routing.
Considering it attacks both issues, both the 75W power limit, and routing of DisplayPort/PCIe to support TB3 ports (either on the card or available elsewhere), it makes sense they'd provide some solution for it. I would have been floored if they let themselves ship an answer that required loopback cables from outside the case.
I'm not that annoyed by the MPX bays, with the dual 8-pin connectors also available in each bay. It's more that because of the need to route DisplayPort around to drive something like the 6K or 5K displays via TB3, that I'd probably need to keep the 580X installed to drive a 5K while using an off-the-shelf GPU for better compute when the Vega II is overkill for my needs.
To be fair though, the Mac Pro power supplies are designed to handle a stuffed case. So the monster PSUs they've been using in the 2019 and 2006-2012 models range from "overkill" to "comfortable headroom" depending on what you are running. It's not like someone needs to choose between a PSU that constrains them and one that doesn't.
As Apple expands the number of repair locations ( which need to do because user base number of Macs is getting bigger also ) ( Phones now but they'll be more later https://www.macrumors.com/2019/08/29/apple-independent-repair-program/ )
Account userid and passwords are even easier to replicate than hardware. It isn't like there aren't tons of folks sharing Netflix etc accounts. Similar mechanism only smaller will happen over time as employees move on from certified repair shops to others. Over time it probably will happen because Apple puts in close to minimal effort into repair depot footprint and logistics at the end of the lifecycle. They outsource most of it.
https://www.macrumors.com/2019/06/19/best-buy-nationwide-apple-service-repairs/
And over time the opportunity of doing iMac Pro and Mac Pro T2 daughter card upgrades will pick up as more parts to do those with pick up and there is a growing population who have the systems on a relatively inelastic (or fixed) budget.
but yeah some back alley repair shop in some 3rd world country probably won't be able to do it if the box fell off a truck in the middle of no where. Apple sales will be bad in those locations anyway. Unless Apple gets their act together though it is going to get worse in those locations as the T2 rolls out across the line up.
The thing I find interesting about MPX is the 500W-per-card power supply. Current cards at peak power are already pushing the limits of the usual PCIE + 2x 8-pin arrangement (the RTX 2080 Ti hits peak power of 360W for brief moments), and I find myself suspecting that more companies will start coming up with their own proprietary solutions like MPX to supply more than 375W. That may mean extended slots like MPX or just some Frankensteining of extra 8-pin connectors, but whatever ends up as the 'standard' solution, this seems like the first obvious move in that direction.
MPX takes up more space in addition to soak in more power. There is a problem trying to push more than 375W into single standard slot width spaces. ( of even just double with spaces). All the more so with "turn the flow 90 degrees" air cooling. Slapping liquid cooling on it to a ballon squeeze to another location in the case basically does away with card fits in any system dogma.
Better intercard communication bandwidth will also help. Along with process shrinks ( 7nm -> 5nm etc moves are going to lead to smaller "big" chips because just can't make 800mm chips anymore. ). For a large swath of embarrassing parallel stuff everything don't have to be done in a single small location. If can divide and conquer with "fast enough" comm links then that will work too with far less thermal 'drama'.
I think the that the 375W limit was part of PCI-e v4 that basically got rolled out early before official. I'd be surprised if PCI-e went much past that for reasons outlined above as it most likely will start to cause interaction impact with the rest of the slot card physical standards and host system relationship.
Instantaneous peaks aren't usually a problem.... You don't need to provision 360w continuous to a card that averages 200w with 360w peaks. Some power supplies are "nervous", and will shutdown when others of the same rating don't have a problem.
For Nvidia on Windows and Linux, the "nvidia-smi" tool can limit those peaks. The issue is that the boost clock on newer cards very quickly ramps up - and the power caps are a bit slower to take effect. A quick boost will overshoot the power cap for a while before the power cap kicks in.
I have a couple of systems with four Quadro RTX 6000 cards (Turing architecture with 24 GiB of VRAM). These cards are rated at 260 watts, but I've seen peaks of over 400 watts per card. These cards will boost to 2100 MHz at least - which causes a quick spike in power that triggered some power supplies. I used the "nvidia-smi --lock-gpu-clocks=300,1695" command to limit the boost to 1695 MHz - no more shutdowns. (Note that the systems in question have 6000 watt power supplies, for 400 watts per card should not have been an issue.)
Sure, but the connecting hardware still needs to be physically capable of the power draw, and 375W is the rated maximum of the PCIE + 2x molex arrangement. That's what I mean: we're hitting the point where graphics cards are butting right up against the maximum the hardware specs involved allow.