No, if you look detained you'll see the SSD connected to the PCH pcie 2.0 x2 lines
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.
Is the new Mac Pro a failure?
- Thread starter ZombiePhysicist
- Start date
- Sort by reaction score
- Status
It seems there are a significant number of people whose love of nVidia far exceeds their love of Apple in this thread (though they obviously love both). For them, MSI may have just announced the ideal machine. It's an i7 with twin GTX 980 cards (MXM) and is about the same size/shape as the Mac Pro.
Not the first trashcan nMP lookalike but actually the first with performance components.
Good lord.
How someone manages to take such a simple concept and make it so ugly/tacky is beyond my comprehension.
Retaking the Apple nnMP speculation, there is also another possibility, Apple to ditch TB2 and provide only 4 TB3 and 6 Usb3 (plus 4 usb-c on tb3) plus dual NVMe on Pcie 2, this architecture will be very similar to the current nMP except don't need the pcie bridge for TB3, also considering TB3 chipset are much cheaper than Tb2 makes some sense, and while NVMe on Pcie 2 aren as faster as on Pcie 3, the possibility to have 2 of them also more than twice faster as the current one on the nMP would be enough compensation. Only issue here is the loss of TB2 and only having 4 tb3 may look as an loss (while in the practice there is only 3 tb2 headers on the nMP, indeed it's an win for the nnMP).
Apple may provide also tb3 to tb2 adapters both as stand alone plug device and as part of the upcoming Tb3 Cinema Display 5K (also rumoured to have an internal gpu).
Look at the diagram more closely - the SSD is connected to 4 lanes from the PCH. It doesn't use the GPU B PCIe 3.0 lanes.
The Anandtech analysis has the USB3 controller off the PCH. The block diagram clearly shows one unused lane on the PCH, and the USB3 controller off the PCIe switch.
I'm not sure which is right, not that it matters much, the only way to be sure is someone with a MacPro6,1 to dump the IOReg.
http://www.osx86.net/file/4251-ioregistryexplorer/
That's what the Anandtech reviewer was doing.
If the USB3 is on the PCH - that means that there's an used PCIe 2.0 x4 sitting near the T-Bolt controllers....
The bandwidth of 4x PCIE 3.0 is 3.94GB/s, a single Thunderbolt 3 port has a bandwidth of 40 Gbps or 5.0 GB/s.
[doublepost=1452182141][/doublepost]
But it's on a switch.
There's not data sheet available for the PEX 8723 so we don't know, but there's only 7.8 GB/s of Bandwidth up stream from the 8x PCIE 3.0 that feeds the PEX.
As the 3 TB2 hubs each need 2,5 GB/s that only leaves 300MB/s.
I'm not sure, if the TB2 ports are really 2.5GB/s, if the PEX switches to 4x PCIE 2.0 the max bandwidth is 2GB/s, so that would leave 1.8 GB/s for the other 4x connection.
However, we're still talking about a switch, in order for something to be on, something else must be off.
Last edited:
Looks very nice, just like the nMP is a very nice design. But for me personally, i really dont care about the biger size of a "normal" computer case. There are some really nice computer cases on the market that can eat all your harddrives, ssds, dual gpu, etc etc etc. I am going for the Define R5, maybe with a window for fun
40Gbps is the medium bandwidth including overhead (tb3 requires less overhead than tb2), each Tb3 dual port controller (aka Alpine Ridge) is connected to 4 pcie 3 lines (check thunderbolttechnology.net official Thunderbolt tech data source). So each pair of Tb3 ports will use that 3.94MBps as maximum concurrent data throughput.
The Thunderbolt controller is which determines how many pcie lines to use, tb2 Falcón Ridge uses 4 pcie 2 at each controller (driving 2 ports), as TB3 uses same 4 lines for double bandwidth but on PCIe3 which indeed provides double bandwith.
worth to read this article: http://www.tested.com/tech/457440-theoretical-vs-actual-bandwidth-pci-express-and-thunderbolt/
Ok, so TB2, post encoding, is really 2GB/s, and TB3, post encoding, is really 3.94GB/s?
You should consider Thundebolt just as the mean to teleport that pcie x4 connection (which actually is what it does, at the other end you can connect an Pcie peripheral as is), so will never give you more than the underlying pcie bridged by it.
Hmmmmm..............
So 40 Gbps is just smoke and marketing speak?
Are all Thunderbolt devices PCIE devices, other than displays?
Is that why Thunderbolt devices cost so much?
On tb2 you can connect either a displayport display or am Tb device which actually is an Pcie expansion card re-arranged, so you can plug an sata controller (using an existent sata pcie chip), the same with an capture card, the only consideration is the drivers being capable to handle plug unplug events (notwithstanding bring foresee on the pcie specification, rarely vendors implement it on drivers before Thunderbolt).
Note that PCIe 2.0 uses 8/10 encoding - so 20% of the T-Bolt bits are created in the PEX, they don't come down the PCIe 3.0 lanes.
Look at Mago's link in https://forums.macrumors.com/threads/is-the-new-mac-pro-a-failure.1939541/page-56#post-22440661
A crossbar switch doesn't have to "turn anything off" until the cross-sectional bandwidth is reached. Eight PCIe 3.0 lanes to sixteen PCIe 2.0 lanes is balanced.
Everybody loves to bash Apple and end up copying their ideas.
Vortex looks great spec wise. Some here will love it, too bad it won't run OS X.
Also the new laptops from HP, Lenovo and LG, even the adds are copies. Man, is this low.
Samsung had an internal manual how to copy the iPhone UI.
[doublepost=1452194543][/doublepost]the encoding overhead is not created in the PEX, but on the controller. The switch does not append additional data, just switches.
Vortex looks great spec wise. Some here will love it, too bad it won't run OS X.
Also the new laptops from HP, Lenovo and LG, even the adds are copies. Man, is this low.
Samsung had an internal manual how to copy the iPhone UI.
[doublepost=1452194543][/doublepost]the encoding overhead is not created in the PEX, but on the controller. The switch does not append additional data, just switches.
Innovation? While Apple may copy they tend to innovate & improve upon the design.
The PCIe 2.0 x4 outputs wouldn't be PCIe 2.0 without the 8/10 expansion in the switch - the PEX has to do a protocol conversion. The T-Bolt controllers expect PCIe 2.0.
Not defending anyone here but who seems to always get the best product out, and others feel the need to follow?
We might say it's expensive and that it won't exactly always meet our needs but credit is due to Apple to be a trend setter.
They're not afraid to innovate.
We might say it's expensive and that it won't exactly always meet our needs but credit is due to Apple to be a trend setter.
They're not afraid to innovate.
Still doesn't excuse the original copying.
[doublepost=1452198909][/doublepost]
"best" is totally arbitrary here. Best at what? And follow who?
If talking about their computer line I would argue that it is entirelly debatable. They aren't the best performing computers or even with the best internal hardware. OSX isn't the best OS either as none are. They all have their limitations and bugs. Their phones aren't the best either.
I think you are confusing "best" with "most popular" or "most glamorous".
The MSI Vortex it's just an iMac retina w/o the retina screen and with a couple of desktop cards instead everything fitted into a exotic bucket, for 3000$ (i7 6700K+GTX980 x2) seems expensive also for a fanboy budget.
- Status