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

[Mago]

specially in since some cards in markets they want to support push more than Thunderbolt 3 can support.

Which ones? Even 8K raw capture device hypothetically don't require more than PCIe 3.0 x4 ... So theoretically a capture card plugged to TB3 has enough bandwidth (even consider most capture card still PCIe 1.0/2.0 soo ...)


Having capture cards with a Full PCIe slot form factor doesn't means it actually uses more than even PCIe 3.0 x2

There are PCIe x8 SSDs NVMe Apple could opt in a proprietary FF, but should be very good if the allow PCIe x4 m.2 FF.

Meanwhile, the group that controls the NVMe STD seems wont double the bus lines, instead wait for PCIe 5.0 with same FF with minor updates.

Personally I think NO STD PCIe slot to come inside the mMP, Apple will stick with a semi-custom GPU as the tcMP but on independent optimized modules with own coolers and maybe even own PSU, also to stand firmly on TB3 as for 3rd party peripheral support.

Take note on this, I had a Vision, a Mac Pro made from 4 small towers(bricks) stick together , the two towers at back hold the CPU, Memory, I/O and the Main PSU, the Two Towers at the front each holds a GPU with Integrated Optimal Cooler and Independent PSUs, the whole setup barely bigger than the tcMP but squared and a bit taller.

 

[Biped]

Well, there is only so much one can talk about - a lot of us have been waiting 5+ years for a replacement. At some point, every pro and con is recycled.

Not sure if between the banter about pcie vs mezzanine vs proprietary GPU and the whole TB3 kerfuffle about a freaking port, if anyone has brought up srvio hardware, and efforts like looking glass and other technologies that use nDruidiaOrgone™ and ATiChopra™ quantum teleportation to manipulate framebuffers onto native gpu chipsets to output over TB3.

 

[deconstruct60]

The more I think about this argument, the more I go back to Apple probably using PCIe GPUs.

I think it's a given that Apple will want to support PCIe I/O cards, especially in since some cards in markets they want to support push more than Thunderbolt 3 can support.

But if you start dividing your PCIe lanes between proprietary GPU slots and normal PCIe slots this becomes a real problem. You might even have 16 PCIe lanes reserved for a second GPU on single GPU configs that become entirely wasted.

I think you are thinking monolithic proprietary custom card connector. It doesn't have to be. The card could have two connectors at different heights. One would be your basic x16 PCIe supply via a PCI-e standards. Another could raise an inch higher from the floor of the logic board than the standard socket and located toward the 'outer edge" of card on the half the card the DP signals are normally routed to on reference standard designs . That second connector would carry the DP signals out and could also bring in power so don't need a extra power cable. That would be a very Apple like plug in and play solution. No extra wires; just properly seat the card and it is ready to go. Done.

The Apple boards would have raised notch so that second connector fit, but generic PCI-e cards would not seat properly. So no plug and play for them ( if supplying power no power cables either). The Apple cards would have no outside edge sockets at all. The end plate would be 100% devoted to hot air evacuation.

The second x16 slot would have no 2nd raised connector. So standard x4-x16 cards would fit. The notched Apple cards could fit as a "Compute" card but supplying an cable adaptor to fit from external power cord lying around for that open slot to fit the power on the bottom. . If the card works on just "PCI bus " power than simply a $0.20 plastic cap to put over the unused connector for DP out. The notched Apple card could fit in the "open" slot. Done. the outer edge is still 100% to hot air evac. Same general usage slot could hold 3rd party "compute' x16 cards just fine to if provide reasonable power and spacing.

The "Compute" GPU card in the 2013 Mac Pro really didn't have to same connector needs as the other card at all and yet they were weaved into the same connector. Uniformity wise so that buying more of the same connector, sure but there was no electrical/data need.

There is simply one and only one slot for "video coupling to TB". There really is no huge split here. 1-2 general purpose slots here is very doable. Even if Apple wants to occasionally sell a "Compute" GPU in the BTO line up if simply use some reasonable design priorities. However, those card are not the crux of what Apple is missing out on that the 2013 Mac Pro didn't serve well. That can't be possibly be a driving design issue. Make the Apple Compute Card adapt to what the other cards need, not vice versa.

It is also likely that Apple's targeted for single GPU usage cards are going to line up well as being good pure compute cards. I can see Apple hoping to get some "extra gravy" money by selling a few more custom cards as compute cards but if the cards can't make it economically primarily just being display cards then that points to a deeper volume problem issue.

If we know Apple wants to support both dual and single GPU configs (which is something they've hinted at) the proprietary, dedicated GPU slots just seem to make no sense.

One of the primary laments in the April pow wow that Apple missed out on the trend toward single even more powerful GPU rather than dual. I don't think the iMac Pro hits that market. That is about power supply budget and air flow. The 2013 Mac Pro had a power supply budget of 450W. The iMac Pro raises that to 500W. That 50W was not the whole story there missing out on the super duper single card market. The 2013 Mac Pro is missing by about 150W at least. A new Mac Pro should be in the historical range. 900-1000W.

If the system is going to have a 140-180W range CPU , 290-800W GPU card(s) , 100W USB/TB power supply duties and other stuff that 500W budget doesn't really cut it.

The trend of single GPUs that are much hotter than the CPUs makes it so if want to compete in the dual GPU market then overall system power budget is going to have trouble going down by half.

Even if you go "But AMD has so many spare lanes!", do you think Apple would want to just leave a giant double wide wasted slot in systems that will never go to dual GPU that can't be used for anything else? I mean just think of the space alone.

It isn't just space it is power creep too. The power creep drives even more space. (e.g., 3 double wide cards. the cards may be spaced close but then need even more wide diameter fans to feed cool air flow at low noise rates).

If the iMac Pro satisfies Apple's desire for a literal desktop pro solution then that should open the door to a more classic deskside Mac Pro solution. Apple could shrink the 2009 era case enough to make it less hostile to rack deployment ( both the mini and 2013 Mac Pro got aftermarket rack solutions. ). Enough space for two double wides should not be a problem for a deskside system.

 

[namethisfile]

Which ones? Even 8K raw capture device hypothetically don't require more than PCIe 3.0 x4 ... So theoretically a capture card plugged to TB3 has enough bandwidth (even consider most capture card still PCIe 1.0/2.0 soo ...)


Having capture cards with a Full PCIe slot form factor doesn't means it actually uses more than even PCIe 3.0 x2

There are PCIe x8 SSDs NVMe Apple could opt in a proprietary FF, but should be very good if the allow PCIe x4 m.2 FF.

Meanwhile, the group that controls the NVMe STD seems wont double the bus lines, instead wait for PCIe 5.0 with same FF with minor updates.

Personally I think NO STD PCIe slot to come inside the mMP, Apple will stick with a semi-custom GPU as the tcMP but on independent optimized modules with own coolers and maybe even own PSU, also to stand firmly on TB3 as for 3rd party peripheral support.

Take note on this, I had a Vision, a Mac Pro made from 4 small towers(bricks) stick together , the two towers at back hold the CPU, Memory, I/O and the Main PSU, the Two Towers at the front each holds a GPU with Integrated Optimal Cooler and Independent PSUs, the whole setup barely bigger than the tcMP but squared and a bit taller.

While, the idea of a self-enclosed/self-powered GPU sounds like a cool idea. Albeit, reminiscent of existing eGPU enclosures, I think that the idea of a cartridge-style GPU as proposed by another member is a better simpler idea.

The connecter of this GPU-cartridge can send more than 75 watts that is typical of current PCIE standard. Since it is created from the ground up, it can send however much power through it to power a GPU without needing PCIE cables, thus, the cartridge-like ease of use with it.

I picture the connecter being where the ports are on a typical dGPU. You sort of just push that part into the modular-Mac Pro and you're good to go.

Sort of like this:

 

[mattspace]

Which ones? Even 8K raw capture device hypothetically don't require more than PCIe 3.0 x4 ... So theoretically a capture card plugged to TB3 has enough bandwidth (even consider most capture card still PCIe 1.0/2.0 soo ....

8K capture cards already exist - that's the Blackmagic card I've posted about here before. TB3 does not have sufficient bandwidth for them.

 

[namethisfile]

Ok. Another quick, very dirty sketch.

 

[Peace]

Ok. Another quick, very dirty sketch.

View attachment 742368

Where's the cooling for all that power hungry stuff ?

 

[mattspace]

Apropos of our complaints about the philosophy of Apple's most recent Mac Pro design:

https://misfitsarchitecture.com/2013/09/29/architectural-myths-8-clean-lines/

For people who don't want to read the article - the summary at the end (copied and pasted):
________________

1. Good design has no need of clean lines. They are a form of decorative ornament.
2. Good design is backwards-compatible, retro-fittable. A modular design with components that can be repaired, replaced or upgraded will have a longer life.
3. Good design takes into account the economic footprint for manufacture, as well as (from 2., above) the economic circumstances of the market.
4. Good design takes a successful prototype and improves upon it. If the result is not a clear improvement, then the prototype IS RETURNED for a re-think.
5. THE HISTORY OF GOOD DESIGN IS THE CHRONOLOGY OF INCREMENTAL AND CUMULATIVE IMPROVEMENTS TO SUCCESSFUL PROTOTYPES. The history of one-offs that led no-where is the history of one-offs that led nowhere.

_______________

That first one is IMHO a gutpunch to the design team at Apple, because they've always presented themselves as inheritors to Modernist ideals of design, but the spare, neatness of Modernist design was always a byproduct of its focus on functionality, not a goal in itself. Apple's current design language is closer to Memphis furniture than Braun appliances in philosophy.

 

[namethisfile]

Where's the cooling for all that power hungry stuff ?

The footprint of the base is roughly the size of a 220mm fan. It's cooled the same was as trashcan mac pro, via natural convection and one bottom fan pulling cool air from the bottom. I guess, the design is inspired by the same thing it is replacing but making it more modular.

The structure/core is the same as trashcan mac pro, except this one has two triangular bases acting as heatsink, which is similar to the trashcan Pro. Not sure if the drawing shows it. But, the blue lines is where the cpu and ram and m.2 boards are situated. The red area is where the GPU-cartridges are inserted. The GPU-cart is cooled also by the system fan and it's own blower style fan. There are no ports to it. Everything is routed into the i/o port.

The yellow area is where there is space for storage, ala PCIE-cartridge style carts. There are two yellow areas and there are two cartridge-style things that can fit in there depending on how the user wants it. Either one 3.5" HDD or 3 2.5" SSD's or 4 m.2 SSD's can fit in one cartridge!

 

[deconstruct60]

...
Not right as you only need a single DP signal for each TB3 Header (2 TB3 ports) so only 3 DP internal connectors are enough to feed the beast (as the tcMP configuration), and are something reasonable, those using DP/HDMI display setup could ditch Mac Pro exclusive GPUs for more mundane GPUs (still the question about the GPU bios).

Incorrect. TBv3 controllers take two DP v1.2 streams. That is how external 5K Thunderbolt Display Docking stations work. Two DPv1.2 streams are sent to the 5K display. That is one of the major feature upgrades to TBv3; the bandwidth to carry two full streams plus more for some modest addition I/O ports to complete a docking station. A system with four TBv3 ports pretty much should have 4 DP streams available. If go to 6 ports then 6 streams. ( even if don't have 6 DP stream fed through TB ports having the last two fed into HDMIv2 (or 2.1 ) and/or classic mini-DP with DP v1.3-1.4 wouldn't hurt on an updated Mac Pro. That would avoid adapters for long term Mac users with sunk costs in the 3rd party monitor market. ]

Conceptually a single TB "v3.5" header with DPv1.4 pass-thru could drive an external monitor that took pure DPv1.4. Those monitors are far from mainstream at the moment and it would be a pure monitor ( there is no data for the I/O ports or downstream devices. ).

I'm on the side Apple will offer their very own
GPUs, maybe something like giant game carthidges integratin liquid cooling and maybe a dedicated PSU.

Game cartridges completely ignores the power supply and power dissipation issues that are very present for most of what would be competitive GPU cards. Liquid cooling isn't a panacea to get out of that. Pragmatically that would be a separate subsystem that attaches to the card as opposed to be a property of the card.

Apple would likely need a range of cards some of which liquid cooling would be too much of an overhead ( e.g., lower end card for folks 100% purely focused on 3rd party GPUs in secondary slot or just don't really need a GPU all that much. Data capture , data serving mac that doesn't have high level 3D video issues. )

 

[deconstruct60]

...

Personally I think NO STD PCIe slot to come inside the mMP, Apple will stick with a semi-custom GPU as the tcMP but on independent optimized modules with own coolers and maybe even own PSU, also to stand firmly on TB3 as for 3rd party peripheral support. ...

GPUs aren't the sole reason for having a PCI slot. The AMD vs Nvidia fan boy hoohah is large but it isn't the principal issue that the current Mac Pro is missing on. Simply pointing at external TB PCI-e card enclosures for x8 and x16 cards is hugely missing the point that they are x8 and x16 cards.

In the "what is in your slots" survey that was done in these forums one of the dominate configurations there were few folks who generally had this:

1. a "boot" GPU ( GT120 or original GPU )
2. newer GPU
3. PCI-e SSD ( because no native PCI-e SSD present) [non outdated SATA kind of fits this also]
4. some sort of x2-x4 card for outdated I/O. (e.g., USB 3.0 or 3.1 )

The 4th can easily be covered by TBv3. The third is more so why when are not stuck with old design just put in a PCI-e SSD socket ( either a M.2 socket or card with M.2 slots on it). If only have 1-2 PCI-e SSD drives and want more .... yeah TBv3 can cover it. Those aren't really the main issue.


The boot + newer GPU drives the issue of why second slot would relieve a lot of unnecessary grief. Some folks beat the farm on sunk cost Nvidia infrastructure. That second slot doesn't have to be used for a GPU. any x2 , x4, x8 card will fit in a x16 slot also. That single slot can cover lots of edge cases to the market (much more numerous than simple GPU selection). That's is what is missing and a sore point. If the iMac Pro is going to eat into the current Mac Pro market (the some customers the current model worked for) then newer Mac Pro needs to expand outside the zone that the 2013 Mac Pro operated in. At least one slot would do that. Zero slots won't.


The boot card though is illustrative as to the pragmatically problem with folks who heap stacks of hate on any Apple GPU solution. The folks who want to eject every possible Apple GPU from their system is yelped a lot here in this forum but it is really a narrow corner case. Apple taking one x16 to cover the "boot" GPU will likely solve more problems than it causes. (always going to have complainers no matter what. ) Seamless plug-in integration with TB just magnifies how much the general mainstream market cards just don't solve the problem. However, that doesn't extend past this one specific function, baseline video output. There is zero need to make other slots disappear just for one sub function. Even on the older configurations those other slot fillers were in the primary GPU role.

Take note on this, I had a Vision, a Mac Pro made from 4 small towers(bricks) stick together , the two towers at back hold the CPU, Memory, I/O and the Main PSU, the Two Towers at the front each holds a GPU with Integrated Optimal Cooler and Independent PSUs, the whole setup barely bigger than the tcMP but squared and a bit taller.

That's is a lot of hocus pocus for not very much. Again the iMac Pro has the desktop "Pro" system relatively well covered. Something that has a bigger power/thermal envelope, more I/O configuration options ( data capture , large scale i/O > 10GbE , x8 storage , x16 compute , etc. ) enabled by at least a single standard slot , and more data capacity than just one drive. The same Intel W, but the primary GPU unleashed from narrow thermal constraint to hit the "big single GPU" case plus the optional configuration to hit some of the older Mac Pro market missed by the iMac Pro.


A single 900-1000W PSU unit can power the whole thing. Multiple PSUs just presents more failure points, not benefits. ( redundant failover PSUs are different. However, don't see that likely at all for a Mac Pro. Even when Apple did 1U server pizza boxes it wasn't. )


Something like 3/4 the size of the 2006-2014 would be a more sane target than trying to pack 800+ watts into approximately the same dimensions of the current Mac Pro. There is going to be more than just one fan and not pinned under 500W. The more you make the power levels overlap the iMac Pro the less differentiate you are going to be able to do between the two products. The new Mac Pro has to cover some different ground than the iMac Pro. ( the CPUs can overlap if let the other elements of the system diverge significantly. )

 

[William Payne]

We are officially in "crazy ideas" mode. For a thread that started off as not wanting radical or proprietary or whatever but wanting standard available parts and stuff we have officially gone completely into crazy idea zone.

 

[Mago]

8K capture cards already exist - that's the Blackmagic card I've posted about here before. TB3 does not have sufficient bandwidth for them.

https://www.blackmagicdesign.com/products/decklink
12G SDI => 11.8 Gbps
TBv3 == PCIe3 x4 => 4GBps => 32Gbps
I think their x8 PCIe is PCIe v1 (eqv x2 PCIe v3)
[doublepost=1513486415][/doublepost]

Incorrect. TBv3 controllers take two DP v1.2 streams.

DP 1.4 signal can be split into 2 x 1.2.

Conceptually a single TB "v3.5" header with DPv1.4

USB-C's DP alt-mode is independent of TBv3, you can feed USB-C header with DP1.4 while the TBv3 still can be feed with DP1.2

1. a "boot" GPU ( GT120 or original GPU )

as tcMP's Boot GPU Apple uses a propietary PCIe form factor

2. newer GPU

Apple as they barely mods STD Reference design, they can offer newer GPUs as soon AMD or nVidia releases reference designs to AVRs,

3. PCI-e SSD ( because no native PCI-e SSD present) [non outdated SATA kind of fits this also]

iMac Pro uses PCIe x4 NVMe in proprietary form factor, unlikely the mMP to use a different storage solution.

4. some sort of x2-x4 card for outdated I/O. (e.g., USB 3.0 or 3.1 )

A TBv3 covers 16x PCIe v1, 8x PCIe v2, 4x PCIe v3

more I/O configuration options ( data capture , large scale i/O > 10GbE , x8 storage , x16 compute , etc.

NBaseT comes integral to the mMP as the iMac Pro, Data Capture is covered by TBv3, Apple wont provide more than 8TB Internal storage (2x NVMe x4), more than 8 TB are covered by DAS either TBv3 or 10G networked, capable upto 96TB (LaCie 12BiG). Even Apple could Offer Compute (FPGA/Xeon-Phi alike) or even license the GPU's cartridge design for such purposes, or an "OpenCartdrige" where you can Install whatever std PCIe card.

A single 900-1000W PSU unit can power the whole thing. Multiple PSUs just presents more failure points, not benefits. ( redundant failover PSUs are different. However, don't see that likely at all for a Mac Pro. Even when Apple did 1U server pizza boxes it wasn't. )

Segregation of The CPU/main syst PSU from GPU PSU allow Apple to offer optimal Power Configurations, withou having to ship a 1000W PSU for people using a single RX570, while allows Apple to offer GPUs withe diverse power demands nor foresee at mMP launch.

PSU failures are Gimmick, a PSU motly fails from overload or unbalanced loads, segregated PSUs are a design safeguard for unbalanced loads.

Cartridge Full Contained GPU/PSU/Cooling solutions with customized PCIe+DP interface delivers Safe Quik KISS updates/upgrades to the mMP.

Ok. Another quick, very dirty sketch.

View attachment 742368

I think on something like this

Ressembling This:

 

[mattspace]

https://www.blackmagicdesign.com/products/decklink

The Decklink 8k Pro should not work over a TB connection of any description. It's a PCI 3.0 8x card.

 

[deconstruct60]

Great post, joema2! I think the point Apple was making is that they want to be able refresh motherboards, CPU and GPU with more of-the-shelf parts than in the current Mac Pro.

Not sure how you get that at all from what Apple said. Pulling something from the post #75 joema2 pointed to that is part of the quotes directly from Apple

"... the Mac has always been about that, it’s been about not doing conventional thinking, not ‘me too’ stuff.. ..."

primary focus on "me too" stuff is off the shelf. Build and use all exactly the parts our competitors are using.... that is exactly what Apple does not want to do. How do we know, because they said it. A primary focus on the same shapes and sizes as everyone else is form over function. The primary focus is the physical dimensions and properties (i.e., form) not function.

Both the motherboard and the GPUs in that were very much custom designs that would just be to costly to re-engineer each year.

There is nothing about the 2013 Mac Pro motherboard (all these subcomponent CPU board , i/O board , or PCH board ) that was prohibitively difficult. They would have had to switch to a new socket for Xeon E5 v3 and possibly bump the USB controller to 3.1. Once switched to v3 bumping to v4 would be even less of an issue ( the PCH board would be the same (or bump to 3.1 if skipped). The i/O board would likely be the same too. The socket on the CPU board would be the same (that is simply just a firmware update. ). The primary issue there is Apple simply not doing it; There was no major technical or financial challenge for this subsystem at all. None.

intel didn't deliver 'yearly' updates to Xeon E5 series. Apple could have done an update in 2015 and then 2017 for a two year cadence.

Apple hasn't been doing yearly updates on most of the Mac line up. It isn't just a Mac Pro issue. it is where Apple is putting effort .... and it is just plainly not yearly tweaks to every Mac product. ( Mac Mini comatose, MBA somewhat ignored or regular basis, iMac went of pause for a year. etc. etc. )

The GPU is murkier because AMD went off into the weeds for a while. At the top end of the scale the major updates that they did just pushed power higher. That was probably a cost thing was an issue the had cut too close with. Questionable whether they did get the D700 right in terms of power envelope. Apple's hard left turn out of OpenCL into Metal shot a giant hole in the dual GPU focus.

However, by mid 2015 again there were some AMD Tonga and maybe Fiji moves they could have made that could have lowered the thermal boundary limits they had without sacrificing any performance. (wouldn't be super bragging rights change, but would have functioned better with the general design limits.)

I honestly think that was the main issue with the nMP; that they didn’t upgrade the CPU, mobo, GPU and ram each year.

Looking for that large conjunction. Have to do CPU , GPU , SSD , and TB all at the same time couldn't possible happen yearly because all of those don't update yearly. The whole notion " it has be some kind of big bang for it to be viable release " is actually part of the problem. Pretty good chance that was the thinking and that Apple was going to shoot for every 3 years perhaps (at very minimal every 2). Not knowing what the cadence was suppose to be was and still is a major problem.



Something like odd years CPU focus and even years GPU focus would give a yearly cadence but it wouldn't be "top to bottom" refresh every year. That is not where the product is for Apple ( or any of the other major workstation vendors either). That is in part driven by the user base (people are sitting on their larger computers longer; not shorter these days. ). What Apple needs to rebuild is trust that they aren't yet again going to hibernate for 2-3 years doing nothing ( 2010 -> 2013 pragmatically nothing. 2014 -> 2017 pragmatically nothing. That's the is the established pattern and it is a major problem. Especially for a company that has a standard practice of not talking about future products. )

So i’m going to assume that the next Mac Pro will be based on components that is more plug and play, especially GPUs.

Plug and play doesn't necessarily mean off the shelf. The only thing in the "especially" category is probably just RAM DiMMs and that the case comes off. That's what you had in 2013 Mac Pro and hasn't every gone away.

The primary GPU probably won't be free market plug and play. Being able to plug in another GPU has a decent chance. I don't think that is what Apple was necessarily talking about though about updating on more regular schedule though.

That will probably also mean easier to upgrade for the user, but that is just a side effect, not a goal in itself as far as Apple is concerned.

Just unbinding the CPU and GPU boards from being bolted to a central core will make access to those boards easier. They were already plug in boards, but the process of bolting/unbolting them to the core is what made access more cumbersome. That decoupling is highly likely. that isn't necessarily a move to off the shelf.

 

[Macintosh IIcx]

*snip*

Looking for that large conjunction. Have to do CPU , GPU , SSD , and TB all at the same time couldn't possible happen yearly because all of those don't update yearly. The whole notion " it has be some kind of big bang for it to be viable release " is actually part of the problem. Pretty good chance that was the thinking and that Apple was going to shoot for every 3 years perhaps (at very minimal every 2). Not knowing what the cadence was suppose to be was and still is a major problem.

Something like odd years CPU focus and even years GPU focus would give a yearly cadence but it wouldn't be "top to bottom" refresh every year. That is not where the product is for Apple ( or any of the other major workstation vendors either). That is in part driven by the user base (people are sitting on their larger computers longer; not shorter these days. ). What Apple needs to rebuild is trust that they aren't yet again going to hibernate for 2-3 years doing nothing ( 2010 -> 2013 pragmatically nothing. 2014 -> 2017 pragmatically nothing. That's the is the established pattern and it is a major problem. Especially for a company that has a standard practice of not talking about future products. )

*snip*

I totally agree on this, and it is just my gut feeling that Apple is implying that they want to be able to refresh parts of the Mac Pro more often. Maybe it is just the GPU options one year, but the next it could be the CPU or the whole mobo/CPU/ram package if a new chipset is needed to support a new CPU generation.

Regarding GPUs and custom designs: I think that the market has very clearly stated that it wants freedom to chose which GPU to use. Some needs nVidia for CUDA, end of story. And I don’t see Apple being committed to make a lot of custom GPUs for different needs, neither do I think the market will see any advantages to paying an Apple premium for those custom designs when they are used to just buy a PCI card and plug it in.

That would only make sense if Apple came up with a new, superior connection for GPUs ala nVidia’s NVLink, but now I’m dreaming. That is what an innovative Apple would do, IMO.

 

[William Payne]

Haha first people were saying about the Mac Pro that Apple needs to realise they don't need to innovate all the time. Now people are asking for extreme innovation.

 

[singhs.apps]

Haha first people were saying about the Mac Pro that Apple needs to realise they don't need to innovate all the time. Now people are asking for extreme innovation.

I think it’s mostly due to exasperation that they may not follow the successful cMP form factors. In which case, what are the existing/upcoming standards that Apple may implement in the nMP and it’s benefits + challenges to satisfy the Mac Pro customers ?

That’s what’s being discussed.

 

[deconstruct60]

...
Regarding GPUs and custom designs: I think that the market has very clearly stated that it wants freedom to chose which GPU to use. Some needs nVidia for CUDA, end of story.

First I don't think that is the whole market saying that. There is a highly vocal subset, they don't totally comprise the market. However, if Apple leaves an empty x16 slot in the upcoming Mac Pro then the freedom could be fullfilled. Fill the empty slot with an Nvidia card. Plug your 3rd party monitors into that card and you are done. If only wanted the Nvidia card for purely for compute ( ML inferencing , data calculations ) then can still use Apple's GPU just fine.

The major problem with the market "clearly stating" is that it completely ignores the Thunderbolt integration design constraints. If the solely PCI-e socket standard cards solved the problem that would be fine. They do not. It is largely a form over function argument since that don't serve the function.

There are two parts to the tension that somewhat flow out of the 2013 Mac Pro design. One is Apple's push to control all of video output of every GPU card that is in the system. 100% of the video out flow has to go through TB or the world will end. Goofy ( in 2013 seem based on 'tail wags dog' notion that Mac Pro exist to drive Thunderbolt display docking station sales. Even more goofy now that Apple doesn't even have a display docking station product of their own. ). About as equally delusional is some user insistence that Apple has zero control over where at least some of the video output goes ( and boot screen compatibility ... ha only girly men look at boot screens ) . Both position are more about "control" than solving problem.


If a user wants to put a CUDA compute card in then simply providing a standards compliant slot settles that issue. Done. If Google started selling Tensor inference cards that fit in a standard slot and had macOS drivers. Again plop that compute card in and off to the races. Data capture card? Again into the slot and done. Could you do all three if there is just one x16 slot? No. However, just one slot will highly likely cover more than decent amount of the market. Can cover a bit more of that market with compute cards in TBv3 external enclosures (not everyone missing but some more. ) which the current Mac Pro can't really cover well (since restricted to TBv2).

And I don’t see Apple being committed to make a lot of custom GPUs for different needs, neither do I think the market will see any advantages to paying an Apple premium for those custom designs when they are used to just buy a PCI card and plug it in.

To the second part of the sentence, if the card is bundled with the system then really don't have a choice. Want to buy an Mac without buying macOS ... can't because it is bundled. The GPU bundled. Apple isn't going to sell GPU-less Macs any more than they are going to sell CPU-less Mac. You get to buy a complete system. Apple isn't a barebones shop or computer parts store. ( Why Nvidia can't seem to win any design bakeoffs of late has a good chance of being what Nvidia is doing at least as much as what Apple is doing. ). Apple's standard configuration GPU solution is going to be an engineered to Apple specification component. That isn't new; that is the way it has been even before the 2013 Mac Pro.

It is only really custom because the standard cards don't solve the problem (integration with TB). There is a somewhat defacto standard/guidelines for upper end laptop GPU cards (MXM ). Works great with TB integration (at least one iteration of the iMacs had them and thunderbolt before Apple fully integrated the GPUs. ). If the GPU and system vendors sat down and did a TB graphics card standard for "large" cards and Apple didn't follow it then yelping Apple for being custom would have some merit. There isn't so custom is what you get when there is no standard. (e.g., Apple moving the notch slightly on M.2 for custom SSDs is worth complaining about. Apple doing a custom SSD card before there was an finalized and approved M.2 standard wasn't/isn't.)



To the first part of the sentence, if Apple provides an empty slot then they don't have to make everything for everybody. As for diversity in the custom cards Apple could make for the upcoming Mac Pro that doesn't have to be all that narrow. The 27" iMac and the iMac Pro have GPU designs that could be relatively easily lifted and modified for the Mac Pro. So could have a sequence of minimal , good , better , and best ( typical apple standard config range)

tweaked top iMac GPU ("minimal") , tweaked iMac Pro GPU still down clocked ( "good" ) , up clocked iMac Pro (better) , and something unique ("best")

Only the last two would be a 100% Mac Pro R&D spend. The first two probably wouldn't cost as much as either one of those independent ones. If those are all AMD GPUs then the "AMD haters" can just buy the first option and fill the empty slot with something else ( more than likely cheaper than buying a TB external PCI-e enclosure box.)

There is a delusion that if simply nuked Apple from doing GPU cards that there would be this missive inrush of 3rd party Mac GPU card vendors. There likely won't be. There was no massive inrush of back in 2006-2009 time period. That was before the MBP, iMac, and now iMac Pro skimmed off a decent number of Mac Pro user base. This new Mac Pro is going to have an even smaller user base. Mostly likely what is going to happen is that there are going to be some "happen to work" PC cards that work better in the post boot status.

That would only make sense if Apple came up with a new, superior connection for GPUs ala nVidia’s NVLink, but now I’m dreaming. That is what an innovative Apple would do, IMO.

NVLink has nothing to do with the boot or TB interface issue. "Superior" isn't the hurdle here. One that actually works is the first step. Don't pump the DisplayPort output out of the box when it is needed in the box. That's it; simply function over form.


P.S. note that Apple is now a full fledged GPU vendor. the iOS devices are moving to fully custom Apple GPUs. With time this too could come to the Macs also. It is more a matter of how much additional resources Apple has to assign to more silicon design work and how well there GPU scales up. It would just be Apple is selecting AMD over Nvidia. It could get to put Apple is selecting Apple over all of them. Telling Apple they can't use Apple because it doesn't look like the form factor from 10 years ago is going to be tough. As long as can put a 2nd GPU into the system though it isn't a dead end or "painting into a corner" issue.

 

[Mago]

The Decklink 8k Pro should not work over a TB connection of any description. It's a PCI 3.0 8x card.

Where are the specifications that stand is a PCIe v3 -a link, not just your mouth-? I have doubts since it uses barely 12Gbps while PCIe3 8x provides 64 Gbps, please support your argument.

This link considers a TB3 version imminent.https://nofilmschool.com/2017/11/blackmagic-8k-cards-are-here

 

[mattspace]

Where are the specifications that stand is a PCIe v3 -a link, not just your mouth-? I have doubts since it uses barely 12Gbps while PCIe3 8x provides 64 Gbps, please support your argument.

This link considers a TB3 version imminent.https://nofilmschool.com/2017/11/blackmagic-8k-cards-are-here

The product's technical specifications sheet, from what it would seem is a reseller's site.

http://www.mediateknik.se/dokument/BM-BDLKHCPRO8K12.pdf

Or on this reseller's site:

https://www.thestreamingguys.com.au/blackmagic-decklink-8k-pro/#product_tabs_tech_specs

"PCI Express 8 lane generation 3, compatible with 8 and 16 lane PCI Express slots."

Or, from Blackmagic's press release posted here:

https://www.dpreview.com/news/30975...k-pro-capture-card-for-real-time-8k-workflows

"The advanced broadcast technology in DeckLink 8K Pro is built into an easy to install 8 lane generation 3 PCI Express for Mac, Windows, and Linux workstations."

So, you know, pretty much the entire internet after 5 minutes googling, as well as my mouth.

 

[Mago]

The product's technical specifications sheet, from what it would seem is a reseller's site.

http://www.mediateknik.se/dokument/BM-BDLKHCPRO8K12.pdf

Or on this reseller's site:

https://www.thestreamingguys.com.au/blackmagic-decklink-8k-pro/#product_tabs_tech_specs

"PCI Express 8 lane generation 3, compatible with 8 and 16 lane PCI Express slots."

Or, from Blackmagic's press release posted here:

https://www.dpreview.com/news/30975...k-pro-capture-card-for-real-time-8k-workflows

"The advanced broadcast technology in DeckLink 8K Pro is built into an easy to install 8 lane generation 3 PCI Express for Mac, Windows, and Linux workstations."

So, you know, pretty much the entire internet after 5 minutes googling, as well as my mouth.

Sneaky words, "PCIe v3 compatible" is not the same as "requires PCIe v3", a PCIe v1 card still compatible with PCIe v3.

Sorry, I'm still not convinced, the main argument against is the bandwidth they declare is 1/4th the bandwidth provided by PCIe v3 x8.

Itś very common for peripherals manufacturers not to declare which technology they are actually using, if they produce a PCIe v1 card (indeed compatible with PCIe v3) they wont say "its a PCIe v1 x8 compatible with PCIe v2 and v3", even given that they stand its mac compatible it suggest it should be MP 5,1 (pcie v2) compatible or thru external PCIe-TB3/TB2 enclosure.

 

[beaker7]

Sneaky words, "PCIe v3 compatible" is not the same as "requires PCIe v3", a PCIe v1 card still compatible with PCIe v3.

Sorry, I'm still not convinced, the main argument against is the bandwidth they declare is 1/4th the bandwidth provided by PCIe v3 x8.

Itś very common for peripherals manufacturers not to declare which technology they are actually using, if they produce a PCIe v1 card (indeed compatible with PCIe v3) they wont say "its a PCIe v1 x8 compatible with PCIe v2 and v3", even given that they stand its mac compatible it suggest it should be MP 5,1 (pcie v2) compatible or thru external PCIe-TB3/TB2 enclosure.

Commas can be confusing.

It says ‘PCI Express 8 lane generation 3, compatible with 8 and 16 lane PCI Express slots.’

That means the card is built with PCI-e 3.0 x8 (as shown in picture) but could also be used in a 3.0 x16 slot.

 

[Mago]

Commas can be confusing.

It says ‘PCI Express 8 lane generation 3, compatible with 8 and 16 lane PCI Express slots.’

That means the card is built with PCI-e 3.0 x8 (as shown in picture) but could also be used in a 3.0 x16 slot.

So I believe in facts, they declare a bandwidth capture 1/4 the bus capacity, even if they are actually using PCIe v3 x8, they dont need more than 2 pcie lines.

 

[beaker7]

So I believe in facts, they declare a bandwidth capture 1/4 the bus capacity, even if they are actually using PCIe v3 x8, they dont need more than 2 pcie lines.

The card has a PCI-e 3.0 x8 connector.

Maybe you should apply for a job there so you can teach them how to design their cards correctly.