That is a lovely dream.
I am afraid that the reality is going to be more locked down, out of date components and a further increase of PR folks on here telling us that we are holding it wrong.
But would be awesome if you are right.
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.
2016 nMP
- Thread starter ManuelGomes
- Start date
-
- Tags
- nmp
- Sort by reaction score
- Status
That is a lovely dream.
I am afraid that the reality is going to be more locked down, out of date components and a further increase of PR folks on here telling us that we are holding it wrong.
But would be awesome if you are right.
During the new Mac Pro reveal, Phil Schiller said this would be the form factor for the next 10 years. So I'm guessing the latter.
We can only hope that "Phil My Ass" made a general comment about a series of cylindrical systems with a central heat sink.
It would be tragic to wait ten years for the design flaws of the 6,1 to be corrected (tiny power supply, dual GPUs whether useful or not, single GPU vendor, single CPU, four DIMM slots, a tightly packed mess of external connectors without latching mechanisms...).
Last edited:
At this rate it seems he meant to say the 2013 model is all we'll get for 10 years.
Only if Apple buys up 7 years worth of obsolete and out-of-production chips to keep making the MP6,1 after all of its components go end-of-life.
And that's not something that bean-counter-in-chief-Cook would do.
Apple is run by bean-counters, not by engineers or computer people.
On the other hand, maybe "Phil My Ass" was looking at his retirement date, and didn't plan on coming into the office for 10 years.
Last edited:
DMI is pci-e some dual or more cpu boards link the 2th 3rd 4rd cpu DMI right to an pci-e slot.
The impact will be the same if the code is badly written and not optimized for the API and hardware being used.
I wonder if it would be possible for them to retain the cylinder configuration and thermal cooling approach, but simply use two internal PCI connectors for stock GPU boards. Although I doubt you could fit something the size of two Titan cards along with their fans or heat sinks into the current case.
The more you look at this the more it becomes evident that at the root of many of the performance limitations of the nMP is the design of the case and it's cramped size. Basically we got one CPU and down clocked GPU boards so it would fit into the design of the case, which is ass backwards.
It's just another example of Apple's new direction where they are prioritizing design and looks over functionality.
This is the same approach that compromised the iMac. They got rid of the vents at the top of the case and now the machine is constantly running at it's thermal limit the moment you do any kind of processing.
The more you look at this the more it becomes evident that at the root of many of the performance limitations of the nMP is the design of the case and it's cramped size. Basically we got one CPU and down clocked GPU boards so it would fit into the design of the case, which is ass backwards.
It's just another example of Apple's new direction where they are prioritizing design and looks over functionality.
This is the same approach that compromised the iMac. They got rid of the vents at the top of the case and now the machine is constantly running at it's thermal limit the moment you do any kind of processing.
Another plausible configuration for the nnMP w/o pcie 3 bridges not considering the "leaks" about tb3/usb-c:
Both gpu retaining 32 pie 3 lines (a gold rule for a workstation).
One Falcón Ridge Thunderbolt 3 header (2 tb3 ports) using 4 pcie e lines
One NVMe using 4 pcie 3 lines
Two Thunderbolt 2 headers connected to the PCH 's 8 pcie 2 lines.
Single GbE from PCH.
4 usb 3.0 from PCH
USB 3 Wifi adapter attached to one usb3 port from PCH.
USB 3 bluetooth adapter attached to the remaining USB3 Port.
The little having only 2 Thunderbolt 3, but still having more common Thunderbolt 2.
The single GbE port would be backed with Thunderbolt 3' on board 10GbE, still having 1 port for an external gpu or co processor, and still having a DP 1.2 from the port shared with 10GbE.
Another variations could be to use an Pcie 3 switch to allow 6 Thunderbolt 3 and an 4 lines NVMe, allowing the 8 PCIE 2 lines from PCH to bt/wifi/GbE/10GbE this don't increase too much the board cost but maybe not compatible with Falcon Ridge, on the other hand given Falcon Ridge chips are cheaper than Thunderbolt 2 this board still cheaper than previous generation.
Both gpu retaining 32 pie 3 lines (a gold rule for a workstation).
One Falcón Ridge Thunderbolt 3 header (2 tb3 ports) using 4 pcie e lines
One NVMe using 4 pcie 3 lines
Two Thunderbolt 2 headers connected to the PCH 's 8 pcie 2 lines.
Single GbE from PCH.
4 usb 3.0 from PCH
USB 3 Wifi adapter attached to one usb3 port from PCH.
USB 3 bluetooth adapter attached to the remaining USB3 Port.
The little having only 2 Thunderbolt 3, but still having more common Thunderbolt 2.
The single GbE port would be backed with Thunderbolt 3' on board 10GbE, still having 1 port for an external gpu or co processor, and still having a DP 1.2 from the port shared with 10GbE.
Another variations could be to use an Pcie 3 switch to allow 6 Thunderbolt 3 and an 4 lines NVMe, allowing the 8 PCIE 2 lines from PCH to bt/wifi/GbE/10GbE this don't increase too much the board cost but maybe not compatible with Falcon Ridge, on the other hand given Falcon Ridge chips are cheaper than Thunderbolt 2 this board still cheaper than previous generation.
I really like your ideas of the dual CPU option and Apple designed external GPU enclosure. I imagine they could offer the Mac Pro in two configurations 2 GPUs and 1 CPU or 2 CPUs and 1 GPU and let buyers pick their preference depending on their work load. They could definitely seriously innovate with a super small 22 core xeon v4 + R9 nano combo with external GPU options for the future if they wanted to...
Mago, I advanced something like that quite some time ago but I wouldn't count on it. The GbE port uses a lane on the PCH, I also thought it didn't but it does. And WiFi and BT are a one chip package.
You could have the same controllers on the PCH and have only one TB2 controller also, but that would mean less TB ports overall.
You could have the same controllers on the PCH and have only one TB2 controller also, but that would mean less TB ports overall.
Peripherals provided by PCH don't count on PCH's pcie lines unless you are plugging some other controller for such peripheral (as could be an alternative NIC).
I'm still thinking Apple will offer 4 tb3 and 2 tb2, using or not an Pcie 3 bridge, it depends on the actual NVMe performance on Pcie 2 x4 or Pcie 3 x4, yet there is no NVMe faster than 2.5 Gbps which is on the pcie 2 x4 bandwidth, maybe Apple can get something faster it's very possible this would require pcie 3 and a maybe a bridge.
About 10GbE on the nMP the only solution we will ever see is on Thunderbolt 3's capability it's more likely Apple to offer an dual socket cpu option than they replace the NICs with 2x 10.000/1000/100mbps NICs.
If I'm Mac project manager at Apple what I would do is to put an eSATA port on the back instead the 2nd nic, this will never happen, it's more likely an frontal panel with usb and power switch than that (the other changes I'll do).
This gets said a lot, but I don't think it's true. I think they started with the goal of eliminating PCIe expandability for support and manufacturing reasons and the small form factor was just a happy secondary benefit. I personally think the eGPU idea is a great one, but I'm not sure Apple will embrace it for the same support reasons. Apple doesn't really test against third-party hardware - not against 4K displays with the nMP, not even against non-Apple wifi routers across the entire product line. The forums are full of this stuff, people are still having trouble hooking up their Apple TVs. Apple has no intention of doing something similar to Microsoft's WHQL (Apple MFi notwithstanding), and going through the expense of doing a full open architecture but then saying it's only supported for one or two Apple-blessed cards never made a lot of sense when they did it last time. I'd really like to see an Apple eGPU chassis but I don't know how that works in a way that doesn't turn into the type of burden that they seem intent in avoiding.
http://www.anandtech.com/show/10054/amd-and-ap-team-up-for-experimental-360-video-vr-news-channel
AMD is indeed targeting for VR content. I belive Zen & new GPU's will be the key players.
Again, the integrated GbE does require one of the lanes.
I was also under the impression that it didn't, even Intels' own diagrams lead to believe so, but if you look at past posts, and I believe it was deconstruct that corrected me on that same issue, pointing to an Intel doc, it does use a lane.
The same is true with the newer versions of the FlexIO on Z170 for example. You have 20 lanes to use but some are tied to SATA, others have optional functionalities, but both possible GbE ports do use a lane each. Check it out and see for yourself.
I was also under the impression that it didn't, even Intels' own diagrams lead to believe so, but if you look at past posts, and I believe it was deconstruct that corrected me on that same issue, pointing to an Intel doc, it does use a lane.
The same is true with the newer versions of the FlexIO on Z170 for example. You have 20 lanes to use but some are tied to SATA, others have optional functionalities, but both possible GbE ports do use a lane each. Check it out and see for yourself.
There are going to be no changes to take the nMP back to single GPU. Not happening. An Apple external GPU enclosure is also not happening, especially as the only GPU upgrade path.
Stock GPU boards is also not happening. No easy way to route video back to Thunderbolt internally. Yes there are all sorts of crazy solutions for doing it on Windows. No Apple isn't going to do it. Even if they did they're not fitting a dual height GPU in there so good luck with your entry level GPUs.
Stock GPU boards is also not happening. No easy way to route video back to Thunderbolt internally. Yes there are all sorts of crazy solutions for doing it on Windows. No Apple isn't going to do it. Even if they did they're not fitting a dual height GPU in there so good luck with your entry level GPUs.
..and AMD is going to release something at Game Developer Conference next month... The conference will be held between 14th and 18th of March.
https://twitter.com/GChip/status/700076659546796035?ref_src=twsrc^tfw
More VR content or VR hardware.. or news about new GPU's?
https://twitter.com/GChip/status/700420476070301697
Are these next GPU's names? "Anaheim or jalapeño or cayenne or habanero or bhut jolokia or all the way up to *pure capsaicin*?" - as Raja Koduri teases there..? Bye bye islands..? Self-irony from AMD (hot, hotter and volcano hot chips)?
Last edited:
The Game Developer Conference happens same time with Apple's predicted launch of new products. Sure rumors have been talking about iPad/iPhone only, but...
Time for nnMP? With VR glasses?
Time for nnMP? With VR glasses?
Last edited:
I do not know what to think about it. It is possible, that we will see new Macs with Polaris GPUs, and Skylake CPUs and TB3. For VR, tho. Im absolutely not sure. Thats all.
- Status