Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Another possible variation of the m³ Mac Pro concept is the Super-Trash Can, each GPU Cartrige being Half-Moon Shaped and the central frame in the middle, being barely bigger than the Fan on Top and resembling the tcMP but a bit taller and wider.
 
A couple comments.

First, the only way you see a dual-CPU "next Mac Pro" is if the "next Mac Pro" will start at a price above that of the iMac Pro. If that call gets made then the "low end" may get filled-in via a "Mac Mini (Pro?)"--picture a "MacBook Pro (in a box)"--which'd cover that 2999-and-below range. Good: "reasonable" options--for macs--across the entire price spectrum; Bad: "Mac Mini (Pro?)" or no, there would still nothing attractive to those who want to "pay 2999-3999 now, then do five years of incremental upgrades".

Second, about VR/AR: VR is very premature and something of a sideshow; AR has a nice near-term future. For both, however, what really matters is essentially latency (really it's "time-accuracy of the imagery" but latency is necessary). Desktop VR has a permanent handicap here. It works better than it should and has a higher ceiling for visual fidelity, but will never be latency-competitive against co-designed, tightly-integrated, highly-synchronized hardware/software stacks running literal centimeters away from each eye.
 
Second, about VR/AR: VR is very premature and something of a sideshow; AR has a nice near-term future. For both, however, what really matters is essentially latency (really it's "time-accuracy of the imagery" but latency is necessary). Desktop VR has a permanent handicap here. It works better than it should and has a higher ceiling for visual fidelity, but will never be latency-competitive against co-designed, tightly-integrated, highly-synchronized hardware/software stacks running literal centimeters away from each eye.
For a premature sideshow, I enjoy the hell out of my VR, and even make some change helping businesses with it :rolleyes:. By all accounts, it doesn't have to be latency competitive with whatever's faster. It's clearly good enough for immersive experiences today.
 
2019-2020 may be when we first see any cpu with pci-e 5.0 (intel and amd need skip 4.0)

Large scale CPU design and development pipelines are usually in the range of 2-4 years. PCI-e v4 wasn't done until recently. It just happened to synch up with IBM's Power9 schedule and has a super long gestation period so betting after the initial estimated arrival ( It was suppose to come in 2013 so IBM starting in 2013 to target what was suppose to come that year is not very risk at all). I'm sure IBM would like a two year window where they were only game in town ( https://www.ibm.com/us-en/marketplace/power-systems-ac922 ).

The PCI-e v5 spec is not suppose to be even done by 2019 ( a year faster than their previous 3 year tempo ) . Any prudent design manager would probably want to track it for a year to see if it too hiccups. PCI-e 5.0 is faster bandwidth than DDR3 memory. How that works with substantially longer trace lengths is in part folks just saying "oh yeah we should be able to do that" , not this is all pretty much done. PCI-e committee talks about how PCI-e v4 laid ground work for future, but doesn't mean the committees are necessarily move quicker or the problems upgrading get significantly easier. [ if they shoot for 2019 and land in late 2020 then back to 3 years. Stretch goal, 'happy' talk. ]
 
No doubt Apple has studied this and is borrowing some tricks - most likely their own version of NVlink. Be nice if the industry decided on a standard that all GPUs could use.
Actually, I find that post to be one of the most unlikely predictions of 2017.

First of all, intra-GPU com links are very dependent on the architecture of the GPU. Nvidia SLI only works with identical chips:
the graphics cards you are wanting to use in SLI need to share the same GPU (i.e. Geforce GTX 670 + Geforce GTX 660 will not work but Geforce GTX 670 + Geforce GTX 670 will work), Memory type (DDR3 vs DDR5), and same memory volume (1gb vs 2gb)

http://nvidia.custhelp.com/app/answ...-work-together-in-an-nvidia-sli-configuration

Do you think that Apple can come up with a proprietary interconnect with features that the vendors' own interconnects can't support.

If Apple did, would you want to throw money at an Apple proprietary interconnect?
 
For a premature sideshow, I enjoy the hell out of my VR, and even make some change helping businesses with it :rolleyes:. By all accounts, it doesn't have to be latency competitive with whatever's faster. It's clearly good enough for immersive experiences today.

Vast fortunes were made in the personal computing business even though phones wound up becoming the primary platform for personal computing.

If you’re having fun and making money with VR...good!
 
Actually, I find that post to be one of the most unlikely predictions of 2017.

First of all, intra-GPU com links are very dependent on the architecture of the GPU. Nvidia SLI only works with identical chips:


Do you think that Apple can come up with a proprietary interconnect with features that the vendors' own interconnects can't support.

If Apple did, would you want to throw money at an Apple proprietary interconnect?
I said it would be nice, not that I thought it likely. But yes, I think it's possible they might do so with AMD only cards. You could still install Nvidia, but that would only work at regular PCIe speeds. They favor AMD a lot these days, it could happen.
 
No doubt Apple has studied this and is borrowing some tricks - most likely their own version of NVlink. Be nice if the industry decided on a standard that all GPUs could use.

You mean like the single CPU package? They probably noticed that.

The 1.5k W power supply? Yes, but only in context not going to touch that.

That the OS here is Linux ( it isn't a Windows threat machine)? Yep.... probably noticed that too. ( Ubutu is on what iteration of desktop GUI du-jour ? ) A deskside staging for what lots of folks are racking in data center .

About 10db higher than a 2012 era Mac Pro ... probably saw that too.


Apple has probably already bought, let along just looked at, something like this. Primarily to do "big data" scraping of their cloud and devices customer data. That doesn't mean they need to make one for sale.

Industry standard NVink? PCI-e v5 is pretty close. Unless Apple is rolling their own GPU/AI engine from top to bottom they don't really get to pick any proprietary interconnect link.


It was odd Apple probably wasted time on the Crossfire for the 2013 Mac Pro in Windows mode. I highly doubt Apple is looking to do their own link that spans cards. ( they have an internal link to work on inside of there ARM SoC chips. That is where 99.9999% of their effort in that space will go. )
[doublepost=1513827066][/doublepost]
I said it would be nice, not that I thought it likely. But yes, I think it's possible they might do so with AMD only cards. You could still install Nvidia, but that would only work at regular PCIe speeds.

Eh? SLI/Crossfire/NVLInk are all back channel VRAM to VRAM data buses. It is so you have a bigger, virtual RAM footprint image of the VRAM that is actually located farther apart ( yeah for high frame rate games that's mostly buffer copying but it is still "big shared buffer". ).

None of these are replacing PCI-e in any meaning full sense ( there is corner case with NVLink being coupled to IBM's OpenCAPI that is far from being the norm at the moment. )

AMD has Infinity Fabric and they are talking much about "Crossfire" anymore.

https://www.pcgamesn.com/amd/amd-navi-gpu-specifications

Not the "last big" GPU made if they just do multiple modules on a card. Hooking modules doesn't have to be a multiple card thing. Just a extra large card thing.
 
Last edited:
Just build a freaking tower already.
tritable12_2017-Dec-21_04-22-47AM-000_CustomizedView15670581863_png.jpeg
 
  • Like
Reactions: OS6-OSX and Aldaris
Our thinnest Pro yet!

And in our amazing rose gold finish, diamond burnished for an impossible sheen!
lol.

the real thing is a little cigarette table.. i wanted to do solid concrete except still reasonable light.. this shape proved to maintain strength while shaving the pounds ;)

Screen Shot 2017-12-21 at 1.05.41 AM.png


-----
i thought it might look funny as a pink computer hence the earlier post.

i really don't think i did a triangle because of nMP thermal core but now that i'm talking about it in this forum, maybe it subconsciously creeped in. :)
 
lol.
the real thing is a little cigarette table.. i wanted to do solid concrete except still reasonable light.. this shape proved to maintain strength while shaving the pounds ;)

As a place to temporarily place my Martini while I dig around my pockets for contraband ... hell yes !!
As a case design, please no, god no, have mercy ( no offense meant, and thank you for sharing ).

But, if you were to somehow ... mc escher it; impossible edges, warped dimensions, time becoming space becoming time again ....

You know, something to espouse in a sculptural metaphor how a 5 year old GPU can somehow be 'new', 'magical', 'groundbreaking' and 'relevant' at the same time !!
 
one of the major apps they were demoing at the iMac Pro launch, Gravity Sketch, is Vega64 minimum in macOS - which rules out the lower config GPU iMac Pro.

The hardware requirements for Gravity Sketch on SteamVR have now been lowered to include the Vega56, which is interesting - I wonder if listing one of the iMac Pro configs as below minimum spec caused some discussions.
 
Let's See. Dec 1st 2016 - Dec 1st 2017 are pgs from 111-275!
Will it take from pgs 275 to 440 for Dec 1st 2017 - Dec 1st 2018?
That's a lot of time and posts! Maybe that is the month for the 7,1!
Anyway it goes I'll bring the doughnuts! :p
Doughnuts.png
 
SSD in the iMac Pro is paired with the logic board: https://de.ifixit.com/Antworten/Ansehen/447686/Number+of+RAM+slots,+SSD+modules+and+Thunderbolt+data+paths

Quote: The SSD is comprised of two modules. They are removable, but are paired to the original logic board and data won't be accessible if moved to another iMac Pro logic board. If either SSD card fails or if the logic board needs replacement, the data on the drives will be lost. I suspect this pairing aspect will make third party replacement difficult or even impossible.

https://imgur.com/a/RjqwA
 
  • Like
Reactions: singhs.apps
Synchro3, it's even on Apple's own iMac Pro page, spatial and temporal dithering. But it's supposed to foul 99.999999% of people, only the real eagle eyed Pros will see banding. There was an article mentioning it too, can't remember where I read it though. But it's true alright, Apple doesn't even hide it. Pixels switch colors and all.
Doing this in such an expensive machine is a bad move.
I can see the TB3 display using the same panel with the same trick.
Too bad. I had such high hopes. I was hoping for a 10b HDR panel.
Now any card can output 10b and using 8b panels seems poor at best. Probably to save on bandwidth.
 
  • Like
Reactions: Synchro3
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.