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

[ManuelGomes]

Their apology can be honest, but that doesn't mean they'll bend over and do what people want them to. They'll stick to their agenda. And with the passing of time, less and less people will need or want to keep changing cards (upgrading).
And for those who pretend to be eco-friendly (I'm very much so myself mind you) and say that they want to keep the computer more time and only change GPUs, come on!! Really? Don't buy that for a second.
But I digress.
More and more I believe the dual GPU will pay off, not quite there yet but by the time the mMP comes out maybe. And if by then it does matter, people will BTO with dual GPUs (the ones who needed only one) or will complain that the base model only has one.

 

[pat500000]

So.... by not expecting them to deliver something that we want/need, we will be happy - and buy that product? Sorry, not here... Either they deliver a flexible new system like the cMP or they will continue to loose the pro market and they will be history in that segment.

They wont meet your standard. Mac pro isn't the same mp from 08-12. Look how slim shady they have been since tim took over. Look how phill's attitude been with lame excuse on mbp 2016 issues. So, no. Not expecting much will ease some tension. mMp will still sell but not at the level during of steve's era. We still dont know what "modular" mean according to their definition.

 

[antonis]

People are the ones that are expected to buy their machines and for some reason apple tends to forget that, lately. Is it arrogance ? I don't know, but they seem to think they know better what people need, than people themselves. And if this is an attitude that can be passed on consumer products, it won't cut it on the professional market. After all, satisfying your customers is far from "bending over". For instance, I know I don't need apple to tell me what kind of laptop I need to do my work. And it is not a laptop that will be followed by a big pile of dongles.

Regarding user-replaceable GPUs, I really don't see where the prediction that "with the passing of time, less and less people will need or want to keep changing cards (upgrading)" is based on. PC market shows exactly the opposite. If apple thinks that they can 'train' their professional customers to need what apple want's them to, it's again too much arrogance from their side.

Regarding dual GPUs, apple should stop shooting themselves on the foot for a start and implement a crossfire after the tcMP fiasco unless they plan to keep producing machines that work better with windows. tcMP deserves a "designed for windows" sticker on it more than many PCs that have it. Then we can talk about dual GPUs.

 

[Mago]

We still dont know what "modular" mean according to their definition.

consider they asked for a full year to re-think it, it remotely means any comeback to the cheesegrater form factor, or even to adopt open stds, in case apple switchback to std PCIe slots on a moreless traditional Mobo, the all new mMP should be available at WWDC as most workstation based on Skylake-W.

They need all these time not to develope a Skylake-W mobo, nether to run burn-test on STD GPUs, I believe they are developing an all new GPU board design and need all this time tu run burn test to be sure no more thermal gotchas as with 6,1, they need it modular so they can introduce periodic updates w/o pain, with a lot of good luck we will see this GPU design as open source but gpu vendors sure shoud license Apple EFI if they want to sell in apple ecosystem.

also I dont expect any PCIe slot, neither M.2 storage, NVMe's sure but similar to those on non-tb MBP13'16. and maybe some TBv2 along a bunch of TBv3 and unlikely to see USB3, mayne HDMI2.1 will arrive, neither dual socket option, about GPUs we like to see nVidia on it, but for 3-D VR/Ar next AMD Vegas maybe ready and competitive by year's end, but htere are many signals about nVidia comeback to Apple, lets see.

 

[ManuelGomes]

Now if we got Power 9 and a couple of Volta GPUs all connected through nVLink... something like the DGX Station would do as well, Xeon based, no nVLink though.

 

[Mago]

Now if we got Power 9 and a couple of Volta GPUs all connected through nVLink... something like the DGX Station would do as well, Xeon based, no nVLink though.

nVidia's new DGX-1 looks alike the Cheesegrater... but 400x more powerfull (and 7x more expensive than the most expensive ever Mac Pro)

 

[Flint Ironstag]

Now if we got Power 9 and a couple of Volta GPUs all connected through nVLink... something like the DGX Station would do as well, Xeon based, no nVLink though.

I just glanced at the DGX brochure and it has NVLink now.

Apple will go high end with the internals; definitely something along the lines of what you described. The nMP was obviously selling in sufficient volume for them, so I don't see them crippling this box. Configure it cheap and upgrade yourself, or buy a monster at a monster price.

In fact, I can't shake the feeling that they've at least considered the following configuration - indulge my imagination for a moment...

Modular Mac Pro:

It'll be a compact object of desire, somewhere between the cheesegrater and cylinder in stature.

- configurable up to 2 CPUs, possibly water cooled on the high end. (enthusiast upgradeable)
- (2) 2.5" bays for SSD mass storage OR (2) dedicated NVME slots
- M.2 for primary storage (user upgradeable)
- (4) PCIe 4 x 16 slots. spaced sufficiently to accommodate 4 full sized GPUs
- 1500 watt PSU (or greater)
- definitely some type of high speed fabric like NVLink
- dual 10Gig ethernet
- smattering of Thunderbolt x ports front and rear
- dedicated x 16 PCIe expansion slot / or proprietary NVlink


Cook takes the stage:

"Oh, and one more thing. We love the Mac, and we know you do too. This is the most powerful generation of Mac Pro we've ever made, and we're really proud of it.

But powerful as Mac Pro is, some of our vocal users let us know it just wasn't powerful enough. We heard you! For all you Pros out there, allow us to introduce the Mac's little sister: Lisa."

Cue fanfare, blah blah blah...

Ok back to my speculation now. Lisa is a dedicated expansion chassis for the mMP:

- matching or very complementary enclosure
- massive PCIe backplane, dedicated daughter card with Thunderbolt x and PCIe x 16 adapters
- (6) PCIe x 16 slots sufficiently spaced to accommodate (6) full sized GPUs
- sufficient power
- silent cooling
- plug and play
- not cheap, but way more cost effective than buying a 2nd Mac Pro or Linux cluster
- let's get wild and say you can daisy chain a couple of these
- and when MP 8,1 hits, simply plug Lisa into that.

Yeah...

You know they could do some really cool stuff to differentiate the mMP if it comes standard with some sort of dedicated onboard AI.

 

[AidenShaw]

- 1500 watt PSU (or greater)

I hope that people hammer the Apple feedback button so that the mMP is something like this. Nobody wants Cube 3.0.

However, a 1500 watt power supply is not acceptable to US electrical codes on a dedicated 15 amp 120 volt circuit.

The biggest in my 120v servers is 1400 watts - above that I have to connect them to 208 volt or 240 volt power strips. (Actually, almost everything is on 208 volt 3Ø 180 amp power strips anyway.)

 

[vailr]

Whenever the finalized new MacPro hardware is introduced onstage as "available for sale today", they'll probably also want to include "Oculus Rift on macOS".
(Tim Cook shakes hands with Mark Zuckerburg.)
Which would also mean: HDMI & USB 3.0 ports.

I hope that people hammer the Apple feedback button so that the mMP is something like this.

Thanks for the link; here's my comment left for Apple's MacPro design team:
++++++++++++++++++++
Sir:
Please consider using only "off the shelf" standard PC form factor (& firmware) graphics cards for the MacPro currently being planned for release at some future point in time: 2018??. What we certainly don't want is: an Apple-only form factor GPU. Nor do we want to be required to use GPU's with a special Apple-only firmware.
Without those two characteristics, I see little chance of much MacPro sales going forward. HP's Z840 workstation design would be an excellent starting point for Apple's MacPro design team to start looking at.
As a matter of fact: why not simply contract with HP to sell the Z840 workstation (with a modified firmware, if required) with macOS Sierra installed at the factory?
Thanks.

 

[koyoot]

https://9to5mac.com/2017/05/10/hackintosh-amd-rx-480-vs-nvidia-1080-ti-final-cut-pro-x/

 

[flat five]

And it is not a laptop that will be followed by a big pile of dongles.

if you bought a 2016 MBP, what would your dongle situation actually look like?

as far as i can gather, this 'dongle hell' thing is way blown out of proportion.

-----
for example, if usb-A is required, you can get this:
(2 for 6bucks)

-------------

...are those to be considered 'dongles'? to me, they wouldn't be as they're just a small extension of the existing usb-A cable.. using them, nothing would be any different (well, except they're easier to physically use/plug-in since your usb-A plug would now be reversible).


--------
but if those are too dongle-ish for your tastes, then these:
(2 for 11 bucks)

-------
............would eliminate the 'dongle' entirely.. right?
and they're most certainly affordable, right?


----
idk, to me, anything that's being plugged in and not done wirelessly is a dongle.. but maybe that's just me.

 

[Mago]

Modular Mac Pro:

It'll be a compact object of desire, somewhere between the cheesegrater and cylinder in stature.

- configurable up to 2 CPUs, possibly water cooled on the high end. (enthusiast upgradeable)
- (2) 2.5" bays for SSD mass storage OR (2) dedicated NVME slots
- M.2 for primary storage (user upgradeable)
- (4) PCIe 4 x 16 slots. spaced sufficiently to accommodate 4 full sized GPUs
- 1500 watt PSU (or greater)
- definitely some type of high speed fabric like NVLink
- dual 10Gig ethernet
- smattering of Thunderbolt x ports front and rear
- dedicated x 16 PCIe expansion slot / or proprietary NVlink

I highligted the Only things i consider true Impossible to see on the next mMP, very few MP users need more than 1 GPU, what they want its an powerfull one, while Apple may allow multi-gpu configurations (with CUDA/nVlink some apps can see the all the GPUs as a single one, but this its only for CUDA), as GPU interconnect fabric all are proprietary solutions which in case become available should be provided by nVidia/AMD.

Personally I doubt Apple Will put any PCIe slot, as much they will sell an optimized TBv3<->PCIe cage, maybe modular stackable with the mMP, the big issue here is TDP flexibility on the tcMP more than actual modularity, I've found more conveniet to use DAS/NAS for storage and local storage only on data being processed.

I hope that people hammer the Apple feedback button so that the mMP is something like this. Nobody wants Cube 3.0.

However, a 1500 watt power supply is not acceptable to US electrical codes on a dedicated 15 amp 120 volt circuit.

An optimized dual GPU Mac doesnt need more than 900W (300x2gpu + 200 cpu/chipset + 100 RAM).

 

[vailr]

I highligted the Only things i consider true Impossible ...an powerfull one, .. I doubt Apple Will ... as much they will sell ... more conveniet t.. doesnt need ...

A little more spellchecking and proofreading, if you please.
That is: if you want anyone to pay attention to what you're saying.

 

[irked]

----
idk, to me, anything that's being plugged in and not done wirelessly is a dongle.. but maybe that's just me.

To me, the non-locking connector and irritating too-short-to-get-anywhere/too-long-to-cleanly-manage attached cabling are what really define dongles.

Displayport can lock, RJ45 can lock, moving them to a shallower non-locking connector that has a much larger chance of coming loose or partially unseated is what I hate about the dongle situation.

 

[ManuelGomes]

As much as I'd like to see it too I don't think dual CPU will come to the MP anymore.
What I would see possible (but unlikely) was a motherboard with 4 (or more but even the mMP will probably have a small footprint) general purpose slots (most likely not PCIe) hanging from a high speed fabric (NVLink would require Power 9 for the CPU, Xeon won't cut it - maybe the updated sheet from NVidia only considers NVLink for the GPUs) and you could slot in any number of cards of your liking, CPU, GPU or SSD. You could BTO with 2 CPUs, 1 GPU and 1 SSD, or 1 CPU, 2 GPUs and 1 SSD or whatever.
But that would be weird, since the cards would have to interface directly to the fabric. Those would have to be all custom cards (read expensive) with added circuitry. More complex design.
Come to think of it, as good as it sounds, not doable with currently used and available hardware.
[doublepost=1494517297][/doublepost]DGX says it's NVLink fully connected but it must mean Volta cards only, not the Xeon CPU.

Anyway, having dual CPU option on the mMP would require 2S CPUs (one of the Bronze or Silver Xeons) instead of -W, a more complex and bigger board. And you know what space means to Apple, even on the desktop.

Let's see if any more info comes along as we go.

 

[antonis]

if you bought a 2016 MBP, what would your dongle situation actually look like?

as far as i can gather, this 'dongle hell' thing is way blown out of proportion.

-----
for example, if usb-A is required, you can get this:
(2 for 6bucks)

View attachment 699314


-------------

...are those to be considered 'dongles'? to me, they wouldn't be as they're just a small extension of the existing usb-A cable.. using them, nothing would be any different (well, except they're easier to physically use/plug-in since your usb-A plug would now be reversible).


--------
but if those are too dongle-ish for your tastes, then these:
(2 for 11 bucks)

View attachment 699315


-------
............would eliminate the 'dongle' entirely.. right?
and they're most certainly affordable, right?


----
idk, to me, anything that's being plugged in and not done wirelessly is a dongle.. but maybe that's just me.

Well, I didn't buy a 2016 MBP, this machine is out of question for me. Thankfully, my 2015 MBP is in perfect shape. But if I had it, I would have to add to these USBs a dongle for hdmi and another one for ethernet. And since I don't go anywhere without my TB SSD drive with a few small but important VMs in it, I'd need a converter for this as well.

The bottom line is, apple tries so arrogantly to get their choices to look like they belong to the future and all they achieve is to be irrelevant for the present, where the rest of the real world do not share apple's opinions.

The most funny part of this is that by the time their machines will be relevant (e.g. total switch to USB-C ports) these very same machines will be considered old for the rest of their specs and will be candidates for replacement.

 

[mcnallym]

https://9to5mac.com/2017/05/10/hackintosh-amd-rx-480-vs-nvidia-1080-ti-final-cut-pro-x/

I really liked how in there was testing a Hackintosh with the iGPU enabled when have the AMD GPU so presumably Quicksync is also enabled in there, and then compares a single pass H.264 which will use Quicksync.

Yes you need the iGPU enabled as AMD card cannot be the first to init on a Hack whereas generally Nvidia GPU Hack has the iGPU disabled so no Quicksync for that.

 

[JMacHack]

The most funny part of this is that by the time their machines will be relevant (e.g. total switch to USB-C ports) these very same machines will be considered old for the rest of their specs and will be candidates for replacement.

Considering how many people keep their MBPs for 5-8 years or better I'd say a switch to USB-C was warranted. I've already got a total USB-C environment now, I doubt it will take more than 5 years to complete the transition.

And with most devices trying to integrate bluetooth now, who knows if we'll have ports at all in the future?

 

[koyoot]

I really liked how in there was testing a Hackintosh with the iGPU enabled when have the AMD GPU so presumably Quicksync is also enabled in there, and then compares a single pass H.264 which will use Quicksync.

Yes you need the iGPU enabled as AMD card cannot be the first to init on a Hack whereas generally Nvidia GPU Hack has the iGPU disabled so no Quicksync for that.

Where do you see it in settings? I am watching the video and do not see anything pointing to this.

 

[mcnallym]

Where do you see it in settings? I am watching the video and do not see anything pointing to this.

Not in the Video in the writeup

From Article

On Hackintosh machines, the latest AMD cards require the iGPU to be enabled on boot,

The real interesting story unfolds when comparing real world exports. For example, I exported a 4K version of the LaMetric Time hands-on that I posted yesterday. This video was shot using my Panasonic GH5 at 4K 150Mbps.

As I do with all of the videos that I eventually upload to YouTube, I used a Compressor preset to facilitate the export. The single-pass Compressor preset results in an H.264 file with a bitrate somewhere around 50,000 kbps.

End of From Article

Compressor/FCP X Presets with H.264 will use QuickSync if available.

Is why back in 2013 there was all the cries about how even the MacBook Pro could export quicker then the 12 Core Dual d700 nMP. All down to using quicksync which isn't in the E5 Xeons but is there in Core i5/i7 series.

Nvidia Hack don't have the iGPU enabled. Most writeup seen say to disable the iGPU.

The Hack has an i7 6700K so will have QuickSync if the iGPU is enabled.

I am starting to find cMP starting to feel the pinch encoding 1080p Films with Handbrake. Seriously looking at the iMac/iMac Pro or even MBP for next time as more Video Encoding software for OS X starting to use Quicksync now and my main thing for my Mac Pro is editing film recordings in FCP X and then encoding into iTunes so H.264 and could use QuickSync rather then CPU power for that. Mac Pro prices here in UK ridiculous still unless business purchase.

 

[deconstruct60]

Considering how many people keep their MBPs for 5-8 years or better I'd say a switch to USB-C was warranted. I've already got a total USB-C environment now, I doubt it will take more than 5 years to complete the transition.

The USB Type-C connector in and of itself doesn't guarantee any 'future proof" in terms of bandwidth. A type-A to Type-C cable is likely to cost just a much as a Type-C to Type-A cable. Or in long term Type-C to type-C cable.

The port not being USB 3.1 gen 2 is more of an "5-8" year horizon issue than the shape of the port. When it comes to USB usage mode.


Type-C is going to be end user odd when some of them are multiple Alternative mode capable and some are not. They could be separated on a Mac Pro ( e.g., " group of X number of USB only type-C , a couple of HDMI and/or miniDP , and then another group of type-C with several alternative modes. ). As a transition device it would probably go much smoother to make them distinct between shapes and protocols until train alot more users that shape is not an indicator of protocols accepted. The label on the port is more significant than it was in previous generations. ( i.e.,, there is no "one port that does everything, everywhere. " )

And with most devices trying to integrate bluetooth now, who knows if we'll have ports at all in the future?

Quite likely you will. When Thunderbolt was "LightPeak" it was not saddled with power distribution duties. USB ( and Thunderbolt ) has grown into a power bus now as much as a data bus. the Modern MB and MBP 2016 models depend upon it for power of the whole laptop. Easy, wireless transmission of 60-80W is a pain in butt in terms of footprint and efficiency And that will remain so well into the future.

A broader range of bluetooth devices ( clothes , data trackers , etc. ), but high core baseline usage of USB devices ( simple external bulk storage , cheap or exotic mouse/keyboard , audio capture , video capture, etc. ) that have modest to high power requirements will remain on USB/TB not because of data in additional to more bandwidth and far better Quality of Service (QoS) in the connections. ( full expect the bluetooth/wifi spectrum to get grossly polluted in a growing number of places over time. ) There is limitation of spectrum. Everyone can't use it all the time for everything; it is a fixed resource.

 

[flat five]

The most funny part of this is that by the time their machines will be relevant (e.g. total switch to USB-C ports) these very same machines will be considered old for the rest of their specs and will be candidates for replacement.

idk, if you're tech savvy, i think you just don't buy something like this on iteration #1...

even i (who is a little more lenient about this kind of stuff compared to others at this forum) didn't buy one though i was considering replacing my 2014mbp..

the next version (or two) will be the one to get.

as far as i see it, we don't replace computers every year or two or three.. and within those spans, there are generally some computers that aren't worth updating to anyway.. but so far for me and apple, there's usually always a gem that pops up and that's the one to buy.. you just have to get your timing right is all..

for others (probably most people) who aren't very tech savvy, it doesn't really matter if they have a 2016 mbp or hypothetical 2017 model.

 

[Mago]

A little more spellchecking and proofreading, if you please.
That is: if you want anyone to pay attention to what you're saying.

Take it easy, I don't wanna be next MacRumor's Trump...

 

[AidenShaw]

Take it easy, I don't wanna be next MacRumor's Trump...

You already are

 

[flat five]

( i.e.,, there is no "one port that does everything, everywhere. " )

you don't think it could be set up to where one port (and one cable) can transmit USB, Thunderbolt, HDMI, DisplayPort, audio, and/or power?

like, do the alternate modes need to be run through dedicated ports and not able to be fed through ports with other modes?