Yep. It'll be DOA for most of us if the user can't swap the cards.
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 Waiting for Mac Pro 7,1 A1991 (no more)
- Thread starter Mago
- Start date
-
- Tags
- macpro 2016
- Sort by reaction score
Yep. It'll be DOA for most of us if the user can't swap the cards.
People keep wanting Apple to turn the mMP into a PC, and I don't think that will ever happen. For that they might as well make the Hackintosh legal. Maybe call it nHP (new HackPro) 
[doublepost=1494344537][/doublepost]Again I see a trusted source selling here cMPs for 689,99€ and I feel tented.
These are of course refurbished, quad core 2.8GHzXeon E5462 Nehalem with 10GB RAM and 250GB SSD and a Radeon HD 2600 XT.
A good base for an upgrade. But still, a lot of stuff would have to be upgraded and added so...
[doublepost=1494344537][/doublepost]Again I see a trusted source selling here cMPs for 689,99€ and I feel tented.
These are of course refurbished, quad core 2.8GHzXeon E5462 Nehalem with 10GB RAM and 250GB SSD and a Radeon HD 2600 XT.
A good base for an upgrade. But still, a lot of stuff would have to be upgraded and added so...
When I describe what it would take to get me back--that's not a prediction, it's just what I want. What we want and what we predict are two different things.
I think we may see a return to a dual CPU Mac Pro. My main thinking for this is because there just aren't enough PCIE lanes from a single HEDT/SP CPU if you go with a modular approach which provides the option to go with dual graphics. IIRC HEDT/SP Intel CPUs only come with 40 PCIE lanes. Typically, each graphics card will take 16 PCIE lanes, which means you'll only have about 8 lanes left for things like multiple NVMe/m.2 drives, USB C 3.1 controllers, and Thunderbolt 3, if you use a single HEDT/SP CPU. Recall that a single NVMe drive or Thunderbolt 3 port requires 4 PCIE lanes each -- the required number of PCIE increases rapidly with these buses. Thus, going with a dual CPU platform should provide the Mac Pro platform with plenty of PCIE expandability.
If you're asking me "why dual graphics?", then I would be inclined to reason that Apple will continue with dual graphics because they foresee VR and GPU compute as growing markets -- however, unlike the MP6,1, I think Apple will allow dual high-end (250W+) GPU options.
Now, for the tricky part: GPU upgrade options. Apple have got themselves stuck into a corner here. If they allow users to insert generic PCIE graphics card, then there's no way they can continue to rely upon TB3 as the main connector port of the future Mac Pro. They essentially have to choose between providing user upgradeable GPUs or TB3.
As many of you know, TB3 can carry both PCIE and DisplayPort traffic (as well as power when run in USB C alternative mode) over a single cable. However, in order for both PCIE and DP traffic to be transmitted over the cable, both the input and output device must be able to send/receive PCIE and DP traffic (via TB3 controllers). Normal consumer PCIE graphics cards cannot do this. They typically only output DP traffic. Conversely, HEDT/DP Intel CPUs, can output PCIE traffic (when connected to a TB3 controller), but they cannot output DP traffic as they do not have integrated graphics functionality.
Apple made TB2 work on the MP6,1 because they used customised graphics cards which sent their DP outputs back to the TB2 controller chips. The TB2 controllers then combined the GPU DP output with PCIE traffic from the single Ivy Bridge-E CPU, and then output both PCIE and DP traffic via the 6 TB2 ports.
If Apple chooses to go with TB3 as the main choice of port for the MP7,1, then they cannot offer customers the option of inserting generic PCIE graphics cards. They will either have to sell user upgradable custom graphics cards (basically the cards need to have a TB3 controller on the same PCB as the GPU), or they need to pray that Intel will release HEDT/DP class Xeons with integrated graphics capable of outputting DP1.2/1.4 -- but I just don't see this happening.
On the other hand, it could be possible for Apple to abandon TB3 on the Mac Pro platform and simply stick to USB C. USB C can still carry a lot of bandwidth, provides power to laptops (MacBook), and can even carry audio. I think fitting the next Mac Pro with something like 10+ USB C 3.1 ports (easily doable with dual CPUs), while also allowing users to install their own graphics cards/PCIE add-in cards, is a good balance between performance, functionality, and engineering simplicity. Apple can then keep TB3 as the default one-port-for-all connector for the MacBook/Pro.
If you're asking me "why dual graphics?", then I would be inclined to reason that Apple will continue with dual graphics because they foresee VR and GPU compute as growing markets -- however, unlike the MP6,1, I think Apple will allow dual high-end (250W+) GPU options.
Now, for the tricky part: GPU upgrade options. Apple have got themselves stuck into a corner here. If they allow users to insert generic PCIE graphics card, then there's no way they can continue to rely upon TB3 as the main connector port of the future Mac Pro. They essentially have to choose between providing user upgradeable GPUs or TB3.
As many of you know, TB3 can carry both PCIE and DisplayPort traffic (as well as power when run in USB C alternative mode) over a single cable. However, in order for both PCIE and DP traffic to be transmitted over the cable, both the input and output device must be able to send/receive PCIE and DP traffic (via TB3 controllers). Normal consumer PCIE graphics cards cannot do this. They typically only output DP traffic. Conversely, HEDT/DP Intel CPUs, can output PCIE traffic (when connected to a TB3 controller), but they cannot output DP traffic as they do not have integrated graphics functionality.
Apple made TB2 work on the MP6,1 because they used customised graphics cards which sent their DP outputs back to the TB2 controller chips. The TB2 controllers then combined the GPU DP output with PCIE traffic from the single Ivy Bridge-E CPU, and then output both PCIE and DP traffic via the 6 TB2 ports.
If Apple chooses to go with TB3 as the main choice of port for the MP7,1, then they cannot offer customers the option of inserting generic PCIE graphics cards. They will either have to sell user upgradable custom graphics cards (basically the cards need to have a TB3 controller on the same PCB as the GPU), or they need to pray that Intel will release HEDT/DP class Xeons with integrated graphics capable of outputting DP1.2/1.4 -- but I just don't see this happening.
On the other hand, it could be possible for Apple to abandon TB3 on the Mac Pro platform and simply stick to USB C. USB C can still carry a lot of bandwidth, provides power to laptops (MacBook), and can even carry audio. I think fitting the next Mac Pro with something like 10+ USB C 3.1 ports (easily doable with dual CPUs), while also allowing users to install their own graphics cards/PCIE add-in cards, is a good balance between performance, functionality, and engineering simplicity. Apple can then keep TB3 as the default one-port-for-all connector for the MacBook/Pro.
People keep wanting Apple to turn the mMP into a PC, and I don't think that will ever happen. For that they might as well make the Hackintosh legal. Maybe call it nHP (new HackPro)
Not True! We would like it to be something like the cMP that can use industry standard parts. Plain & Simple. If I wanted a PC Tower or a Hackintosh I would buy for build one. But, what I want is a modern, real Mac Pro, not the current closed trash can.
Lou
I can't imagine Apple doing a 38" display, seems too big. Maybe 32", or keep 27" and Retina it is.
If it's intended to replace the TBD, a dock for mostly laptops, it's the wrong size. If not, the target audience is quite limited.
If it's intended to replace the TBD, a dock for mostly laptops, it's the wrong size. If not, the target audience is quite limited.
Agreed.
The biggest display Apple would do is a 32" display. I actually think they will stick to the 27" 5K form factor.
"Apple 27" 5K Thunderbolt Retina Display"
Remember, Apple wants to sell these displays to "normal consumers" and prosumers alike. Pros need amazing colour accuracy things like HDR and DCI-P3 support, while normal want something that looks amazing on their home office desks. Jobs and his Mac team understood this "two birds one stone" product appeal. A 38" display is simply too big and cumbersome for normal consumers. Heck, you couldn't even carry one home after buying it from an Apple Store.
SP Xeons will have 48 lanes.
Dual GPUs I believe is still their target. 16 lanes to spare for dual SSDs and dual TB3 controllers (4 ports). Most likely the 2nd SSD (BTO) will be on the PCH so that 3 TB3 controllers will be on the CPU.
Dual GPUs I believe is still their target. 16 lanes to spare for dual SSDs and dual TB3 controllers (4 ports). Most likely the 2nd SSD (BTO) will be on the PCH so that 3 TB3 controllers will be on the CPU.
I don't see Apple abandoning TB on the Mac Pro - if any Mac leverages TB, it would be the Pro. But they could just make it straight PCI TB3 and have dedicated DisplayPort and/or HDMI ports for the video. If Apple is truly planning an 8K monitor, they would need dedicated DisplayPort because TB3 can't do it as it is limited to DP1.2 and 8K requires DP1.3 or DP1.4.
I don't know: maybe during the extended delay, they're waiting on Intel to provide an updated chipset that would provide another dozen or so PCI lanes? Enough for raided NVMe SSD's, TB3 ports, 3 or more PCIe x16 slots, additional PCIe slots for a professional sound card, a video transcoding card, 10 Gb LAN card, etc. Maybe even a Lightning port for plugging in iPhone headphones.
Same idea for 8k video: waiting on AMD and/or nVidia to release updated GPU's (standard PCIe form factor) that can natively output 8K video over DP 1.4 or HDMI 2.1.
Same idea for 8k video: waiting on AMD and/or nVidia to release updated GPU's (standard PCIe form factor) that can natively output 8K video over DP 1.4 or HDMI 2.1.
Last edited:
You can get 8k presently if you want via DSC with the Dell 8k monitor, and Windows 10 does support 8k with the right software that scale beyond 4k. The GTX 1080 and up and the AMD WX7100 support it. But unlike 4k, the public demand isn't quite there yet. It is said that the next important milestone for 8K will be in 2020, with the Tokyo Olympics is expected to be broadcast at that resolution. At CES, Chinese TV maker Changhong showed off an 8K TV, and chipmaker Ambarella touted its H3 chip for drones and video cameras that can capture 8K video at 30 frames per second.
http://www.pcworld.com/article/3187...are-now-compatible-with-dells-8k-monitor.html
This is fundamentally flawed on two fronts. The next generation workstation chip from Intel has more PCI-e lanes on both the CPU and the PCH. There is hardly any shortage for a normal single user workstation context.
Xeon E5 v4 CPU -- 40 PCi-e v3 lanes + PCH -- 8 Lanes PCI-e v2 = 48 lanes of mixed speed.
"Skylake-W" CPU -- 48 PCI-e v3 + PCH -- "up to" 24 PCI-e v3 = 72 lanes of PCI-e v3 ( although total aggregate bandwidth not quite as high. Closer to 52 in aggregate bisection throughput. )
Intel is going to tweak a subset of the desktop 200 Series PCH controllers for the Skylake-W (and Skylake-X) line up.
http://www.anandtech.com/show/10959...ration-kaby-lake-i7-7700k-i5-7600k-i3-7350k/7
More than likely this is a tweaked Z270. There may be a Skylake-W specific on that based on the Q270 ( the C610 and C600 chipsets both had vPro features. So it likely that the X299 is a tweaked Z270 for the overclocked crowd and some 'C620' (or whatever name Intel slaps on it) will be a tweaked Q270.
The x8/x4/x4 PCI-e config allows the x8 to be mapped to the old C600 duties, and still have two x4 lanes for NVMe SSDs. Or one NVMe SSD and one 10GbE controller.
So the following is more that practical with the shortly upcoming chips from Intel.
"custom Video GPU" x16 ( custom socket PCI-e + DisplayPort + maybe power. )
Compute GPU x16 ( standard PCI-e slot )
TBv3 controller x4 (Thunderbolt bus 1 )
TBv3 bus/controller x4 (TB bus 2 )
SSD VRAM cache x4 ( or second SSD ... cache example http://www.anandtech.com/show/11210/the-intel-optane-memory-ssd-review-32gb-of-kaby-lake-caching )
SSD boot x4 (primary boot )
subtotal 44 [ if back to desk side 'large' case , then add a x4 standard slot and at 48. ]
PCH
Wifi/Bluetooth 5 x1 ( Wifi/BT from rest of mac line )
Audio upgrade x1
( built in around x3-4 of USB 3.0 gen 1 ) [ if the chipset has been bumped to gen 2 then great. ]
Dual 10GbE x4
M.2 SSD socket x4
subtotal 10
Total 54. Slightly over subscribed above the bisection bandwidth, but lots of folks may not use 3rd SSD and/or two Ethernet ports.
If put out video to two miniDisplay Port ports then don't need to consume any PCI-e bandwidth for another TB bus. ( two mDP ports + one HDMI (sharing with a mDP ) port. That would keep the 6 video streams out capability.
Personally, I'd find a way to add two USB 3.0 Type-A ports to 'front' and keep the 4 Type-A in back, so a total of six. That would probably not to piss off a large faction of the folks left.
And done. If it is a much bigger box use one (maybe ) two of the SATA ports for a internal spinner. That is a bit more oversubscribed. The primary usage SSDs are segregated off the PCH to lower the bandwidth pressure between the CPU and PCH.
Sure Apple could swap that VRAM cache SSD for a third TB bus, but if tracking 8K displays over next 2-3 years two miniDisplay ports is better ( those are all dual cable displays now and likely will remain so at least a couple of years. ). Apple needs more storage capacity inside the system. 2-3 mainstream M.2 SSDs is likely a bigger demand than two "wasted" TB ports covered with purely "Alternative mode DisplayPort" duties. That is just a waste of PCI-e bandwidth.
Likewise they could throw the x4 budget at a standard slot. ( which could be filled with a SSD card. ), but that would likely be an increase in volume. ( going Desk side that wouldn't be a problem. )
Not really. They only need one of the PCI-e slots to also feed DisplayPort to the logic streams to the logic back. The "Compute" card has no need one way or the other for a edge "exit" ( venting hot air could be used positively with no connectors. ).
TB is only a constraint if want two random pair GPUs. The begging question if really need two.
As for the "custom" display GPU it could be quite close to a regular slot. Two major connectors. One with PCI-e . One with the 6 DisplayPort streams. It isn't like other workstations don't have a "Rube Goldberg" hackery problem also. If the sub assembly can be pulled out and replaced by simple tools then Apple could sell ( and take back the old) as parts. The Primary display would be Apple specitific but they don't have to mandate that for a 2nd card.
Probably not. The MacBook is likley not a pre-cursor. Apple was 'stuck' with getting TBv3 rolled out so it went out the door with USB-C only. But a one-socket wonder just screams for TBv3. If only going to have one socket it better do everything.
Detaching the Mac Pro from the rest of the ecosystem doesn't many any strategic sense at all. Apple wants folks to be able to move up or down from Mac Pro to rest of the line up without loosing major peripherals. Yes, they periodically "blow up" the transitions with these port shape changes, but after the major transition to Type-C is done.... that is the ecosytem they are likely targeting.
10 USB 3.1 gen 1 ports could easily be covered by the upcoming PCH. ( dump the 10GbE controller above and just go regular two 1 GbE ) and can easily cover.
PCH
Wifi/Bluetooth x1
Audio x1
Ethernet 1 x1
Ethernet 2 x1
USB 3.1 gen 2 x2 ( four )
USB 3.1 gen 2 x2 ( four )
USB 3.1 gen 2 x2 (four )
M2. SSD slot x4
total 14 lanes. That is still 10 less than the PCH's top end of 24. And still have not touched a single CPU's lanes at all. There is about zero justification for another CPU socket to fill some PCI-e lane "shortage gap".
[doublepost=1494362903][/doublepost]
Since they aren't shipping in 2017 that may not be an issue.
I won't be surprised if 'TBv3.1' comes out next year with a DP1.3/1.4 legacy pass through mode. ( in essence a tweaked controller+firmware combination. Not a new version of TB. )Intel did it between TBv1 and TBv2.
"... Redwood Ridge maintains feature compatibility, but you get official support for DisplayPort 1.2 (and 4K resolution) if you're using a DisplayPort monitor. This extension of DP 1.2 support does not apply to Thunderbolt displays or DP 1.2 displays connected to a Thunderbolt chain however. The explanation is simple. Redwood Ridge won't be sold for use in devices, only host computers, ... "
http://www.anandtech.com/show/7049/intel-thunderbolt-2-everything-you-need-to-know
Apple (and many others) avoided this (or at least didn't hype it in system with Redwood ... perhaps cheaper to buy), but I think this time a TBv4 is not in the short or intermediate term future. The port shape change and the merge with USB is causing a ton of ruckus. That last thing TB needs now is more change drama. Pure "Alliterative Mode Display Port " can do DPv1.3/v1.4 (with right support chips). It is really an USB Alternative mode that the TB controller is implements for the legacy, pass-through mode now. So either the current controllers will "happen to work" or a slight feature bump controller should handle this. The main DisplayPort connector got no changes for DPv1.3/1.4 so same set of wires in the current Type-C plugs should be good to go. The only minor hurdle is the handshake negotiation into DP mode. Should not need major silicon to do that.
That said deploying TBv3 (or v3.1) ports to be pragmatically "permanently" covered by two cables is a significant waste. The use case where the desktop monitor is always connected to the system means those two ports are in a 'primary' mode not really an "Alternative" one.
Last edited:
That's true
[doublepost=1494364731][/doublepost]I copied here the rumour about that 8K Pro monitor from apple, but personally I dont buy Apple to engage on a hard to sell 8K monitor very few users considers, I bet Apple will come with a more refined 5K external Display to Address it own natural market not satisfied with LG branded solution.
I think a TBv3 revision is neeed to allow DP 1.4 "alt mode" or pass throug, not a big challenge, since there are now USB-C with DP1.4 alt mode available now.
Also i dont see sense on Apple to re-introduce a PCIe x4/x8 Slot, the few users actually needing this should buy an TB3-PCIe x4 cage.
As for a theorethical mMP based on Skylake-W, Apple should put both GPUs and 2xSSD on the CPU PCIe Lines, also 2 TBv3 headers, and put on the PCH all the other peripherals: 2x10GBe Lan (x4) 2x extra TBv3 header (x8) 3x TBv2 Headers (x6), and the remaining 4 lines for WiFI, BT, and Whatever else (the mMP dont need more than 8 USB-C ports provided by TBv3 controllers).
About its GPU modules, A dual Powerfull GPU solution being seen as a single GPU by the system its possible with nVidia NVLink, also AMD seems is working on something similar,installing a 2nd GPU for compute its an waste in a mac, today still very few applications using the tcMP 2nd GPU.
Last edited:
I recognise that it is....apples and pears but still, it is an interesting comparison.
but still, it is an interesting comparison.
"Straight PCIE TB3" isn't TB3 -- that was my point. You can't have TB support without either PCIE or DP. You need both in order to support the TB standard.
For example if you have the following TB chain:
MP7,1 -> RAID NAS -> Display
The MP7,1 needs to output PCIE and DP on the same cable in order for the display and NAS drives to receive DP signals and PCIE data respectively. You need both in order to support different combinations of TB chains too.
Woah, ease up there buddy. Nothing I posted is "fundamentally flawed". You're simply advocating that there is enough PCIE bandwidth coming from a SP Xeon CPU to support the future MacPro7,1 platform.
All your assumptions get thrown out the window if Apple simply decides to offer four x16 PCIE slots (with full x16 electrical support) like the old cheesegrater Mac Pros.
Then they would need two Xeons for the four PCIE slots, as well as all the extra PCIE lanes for NVMe/m.2 drives and TB3 ports.
Then we come full circle to the Mac Pro and non upgradeable PCIE cards... Which was my original point.
Yes, I'm well aware of the MacBook's TB constraints.
But my point was that Apple could roll out TB as the standard port for its MacBook and iMac platforms, while keeping the Mac Pro on USB C if it decides to let users upgrade their own graphics cards.
The only thing you lose by using USB C instead of TB3 on a MP7,1 platform is the ability to create TB daisy chains. Bandwidth is also lost, but because you're not daisy chaining it's not a big deal anyway. Apple could easily make up for this by providing a **** tonne of USB C 3.1 ports, especially if they use a DP Xeon platform.
What you do gain by dropping TB3 in favour of USB C is fully customisable PCIE graphics cards.
Compatibility with TB3 isn't a big deal since TB3 devices can run in USB C mode anyway.
Again, I want to stress that I don't think Apple will go down this path, but that it must make a choice between customisable PCIE graphics and TB3 support.
Apple being Apple, they will probably choose to go with the TB3 option, forsaking user upgradable, standard PCIE graphics cards. So you may see a situation where Apple will fit the MP7,1 with custom graphics cards with onboard TB controllers (and 5K+ support), and allow users to purchase new graphics card from the Apple Store (very expensive though).
Actually, if the entire GPU industry integrated TB support on their graphics cards (Intel would need to allow this too), then Apple could support both upgradeable GPUs and TB3 support. But alas, we're in this situation, and part of the dilemma isn't really Apple's fault.
I assume it is the same for TB2 devices? I don't intend to get a Mac Pro, but if I did, I'd want to be sure my Drobo 5D could still connect.
You just keep digging a deeper hole.... The workstation solution that Apple will likely use is not a SP processor. The 2 (or more ) solution is being forked off from the workstation solution (-W and so far nameless product name). The notion that the a single SP processor is short on PCI-e lanes is even more ridiculous than saying that the -W version is coming up short. The bandwidth coming off the SP processors is even higher. That is one reason why it is called "Scalable Processor".
If Apple wanted to do something like two custom GPUs and then allow users to put in two more "off the shelf" GPUs then yeah, it would warrant two CPU sockets. But there hasn't really be a Mac Pro like that in a decade. And 4 GT 120 cards really doesn't count because that was far more so that limitation that low end GPUs couldn't drive multiple streams back then. The current Mac Pro can drive 6 off of one card. Scaling cards to cover more screens is a solved problem that doesn't involve CPU PCIe lanes at all.
If Apple is saying the current Mac Pro has "problems" because a large enough set of customers couldn't take advantage of 2 GPU why is 3-4 GPUs going to get a better market share. It won't. So Apple it is extremely unlikely Apple is going there. Yes there are folks who stuff 2-3 cards into Mac Pros. Apparently they are not the bulk of the Mac Pro market. ( There is a current poll on this very same forum that quite clearly demonstrate that it is. Strip out the "keep the boot screen card" numbers from that poll and it is even lower. )
It isn't really an assumption when Apple, after looking at their customer marketing data, says that single GPU is there market weak point right now. They made an explict and overt statement about that. I'm assuming nothing. I'm actually listening to what the market research is showing.
I am assuming that Apple is not going to completely walk away from two GPU solutions. I think the next Mac Pro will optionally do that. However, from what Apple is saying it isn't likely anymore to be the default configuration.
You have left completely unjustified why there is a significantly high market demand for 4 x16 systems. Only a corner case of Dells, HP, Lenovos workstation line up covers that. There is nothing to indicate that is the "missing piece" that Apple needs to have a viable Mac Pro.
The circiular logic here is you. Assume 4 x16 so therefore need two CPUs. That isn't what vast majority of customers are saying.
You are going in circles. The real core issue is some folks being fanatical on either side. That either all the GPU video needs to go into the logic board or that all of the GPU video needs to go out the card edge. TB doesn't mandate that at all even in the slightest. You just need one. Once you go past one GPU in the system then that requirement that the 2nd ( or 3rd , 4th, etc) need to go through the TB is zero.
There was nothing about the current Mac Pro that made the GPUs "non upgradable" other than Apple lazniess. There is no technical blockage there at all. They could make it "cheaper" ( low cost end user service) but TB inherently blocking? No.
The Macbook doesn't have TB yet. They have stripped out so much internal volume it is going to be a tight squeeze to get it in even with the current TB controllers. I suspect they will solve that ( because largely once again it is Apple painting themselves into a corner with self imposed constraints rather than the technology's constraints demanding it).
It makes no sense for Apple to 'fork' the Mac ecosystem like that. Mac share of the market is at best 5-7%. The Mac Pro is single digit of that. That makes the Mac Pro pretty close to approximately zero percent of the market.
USB Type C sockets solve none of the Mac Pro's current core issue. Zip. Nothing. In fact, the Mac Pro probably has more 'hard core', legacy Type-A socket users than any of the other Mac products. ( software key locks with Type-A. expensive peripherals with Type-A ports and cables. , 4-5 USB ports filled all the time, etc. ) For just pure USB Type-A goes just as fast as Type-C. Changing shape buys no bandwidth speed gain in the slightest. None.
I think should "fork" the Mac Pro off from the rest of the Mac products, but only in that they should "slow down" the Type-C revolution on the next Mac Pro. A Type-C only Mac Pro is a bozo move and more than a little tone deaf. Some Type-C via TBv3 .... sure. But having just four is plenty. It works for the MBP 15" and the Mac Pro can actually use all four since will have a separate power cable. So pragmatically, it is more than any other Mac. ( user can chest beat on bragging rights on that if that gets their rocks off. )
So piss port bandwidth down the drain by not using TB..... but need bandwidth so put more ports on than would have had before just to make up for you just pissed down the drain by not using it?
Sigh. USB Type C is just a port shape that has been also coupled to some Alternative modes. From a strict USB (only) versus Thunderbolt perspective, USB do not solve the problem that TB solves. It simply doesn't. If going to implement Alternativ modes on a Type-C port, then TB just does it better. It is a prepackaged solution for that in addition to all of the things that TB v1 and v2 did. If you just want to have USB protocol only then Type-C buys relatively little from Type-A.
No. If there is a second open slot users can put anything they want in there. It is customizable. Done.
Those other cards aren't going to give the user's boot screens ( corner case hackery exceptions excluded. In general, they won't. ) Dropping TB still doesn't solve the non-Mac specific boot problem in and of itself. It is not a holistic solution.
Customizable and getting into a pissing match with Apple about removing everything that Apple put in the system are really two different issues.
TBv3 is a superset of the modes that a plain USB port can do. It is a big deal if have a TB device because it won't work if the Alternative mode isn't there for either TB or DP devices. Whereas if you have a TBv3 port you have all three of the major modes covered. Apple is going to want a road where users can go up or down depending upon how their workload needs change over time. If you make so that only the down product line systems have the full superset then over an extended period of time that is where most of the migration is going to go. Down..... smaller Mac Pro sales. When you get to the point that as Jobs put it, "(practically) nobody is buying them" .... it will extremely likely get canceled.
So the decision is really about do really care about long term viability of the product. From Apples comments the migration trends are already toward the other products ( laptops/all-in-ones growing, Mac Pro not nearly at the same rate.). Tilting the playing field even more so toward the others is only going to speed up the process.
This is all myopically about alot of hot air. Apple needs a PCI-e slot. Not so much for GPUs only, but also for other usages. A x16 slot can hold a x1, x4 , x8 , or x16 card. The Mac Pro needs an add a card capability to incrementally broaded its scope from where it is now. Making this solely about GPUs is hooey. It was hooey back in 2012 when Apple did it. And it is hooey now when muckrakers try to do the same thing from the opposite direction.
TB needs three inputs. GPIO, PCIe , and DisplayPort. Putting TB on a GPU card is highly dubious. It was so in 2011-2012 when folks hand waved about the concept and it is even more dubious now.
Nvidia and AMD don't really like Intel so telling them they have to buy an Intel part for their cards isn't going to fly. Somewhat petty? Yes. But it is basically a reality. The PCI-e bandwidth encoded for transport bleeds bandwidth away from the GPU chip ( hence getting in the way and incrementally reducing their performance). Even if the GPU vendors didn't dislike Intel you really think they are going to be happy with the bandwidth bleed? Nope ( you are only deluding yourself if you do. ).
Even if you get the DP on the board to TB you are still missing the GPIO connection. So doesn't work.
The mainstream PC "box with slots" market has already settled on their "Rube Goldberg" solution of external loop back cables and custom card/cables solution. It is a hack, but it is a cheap. Cheap and tight margins is primarily what they care about. It is not real well engineered solution.
If TBv3 gets to critical mass perhaps they will come along. But it is still the case some vendors want it to die because it is just not a "lowest common denominator" solution. That isn't Apple. Since APple designs the Mac Pro it probably isn't the Mac Pro either. TB enables Apple to make the Mac Pro into a literal desktop. ( as opposed to a desk side box that is thrown into the "desktop" category. ). It has much higher synergy with the rest of the Mac systems.
Last edited:
Could you turn down the d***ish tone, and calm the f*** down?
All we're trying to do is forecast what Apple's next Mac Pro will look like. We're not trying to solve world poverty, debating politics, or trying to find peace in the Middle East -- there's no need to get so hot under the collar.
Are you aware that I was using SP and DP acronyms for "Single Processor" and "Dual Processor"? I made no reference to Intel's new confounded "Scalable Processor" naming scheme.
Whether or not it's Broadwell-E or Skylake-X, a single processor Xeon platform only features 40-48 PCIE lanes, which simply isn't enough to power 4 PCIE x16 (electrical slots), NVMe ports, and multiple TB3 controllers. That's all I'm saying. I'm not going to bother retorting your entire message because we're getting off point.
You say that Apple doesn't need a DP Xeon platform because a single Xeon should suffice PCIE lane needs. I simply disagreed with that point. Move on.
My original post was all about the dilemma that Apple faced: customisable, bog-standard PCIE graphics cards or TB3.
Yes, it's common knowledge that Apple was unable to integrate a TB3 controller on both the Broadwell and Skylake based MacBooks. Fingers crossed that they managed to get fit it in for the upcoming Kaby Lake MacBook.
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.
It is not very complicated really. Just make a tower similar to HP Z840 or Dell 7910 and likes, with standard off-the-shelf products. Similar to the old Mac Pro, but even more standardized than what it used to be. For me the ideal solution would be to run osx (or what the h*** its called nowadays, mac os?) on a z840.
Last edited:
I suppose Mac Pro 5.1 also had off the shelf, standard hardware. You could put every GPU in it, without any problem.
Yeah, right.
With nVidia GPUs has more sense to offer Multiple GPU as with CUDA its easy to support multiple GPU context solving the same solution (even better if those GPU use nVlink -Pascal GP100 only by now-).
But doing the same on AMD/OpenCL its too steep -Metal?-.
People keep wanting Apple to turn the mMP into a PC, and I don't think that will ever happen. For that they might as well make the Hackintosh legal. Maybe call it nHP (new HackPro)
Or people could wait from apple to just prove that their apology was honest.