Except that TB is only capable of 2GBps and there are only 6 of them on the new Mac Pro.
Flawed. It 2GBps of PCI-e data per controller. As there are no 6 physical port controllers it is extremely doubtful there is just one here in the Mac Pro 2013 model. Likely there are three. The 6 ports represent access of over 6GBps which is hardly limiting.
That means any LGA2011 with a few PCIe 3.0 slots technically has more expandability than the New Mac Pro.
Bandwidth wise, just the v3.0 lanes of the E5 Xeon means more bandwidth. But not necessarily an increase in devices. Firmware and reconfiguration allows some bandwidth reallocated. That is flexibility.
On a metric of expandability where talking about more devices then the Mac Pro 6 ports allows to add 36 additional PCI-e switches to the system. A given mainboard has a fixed number of switches. The bandwidth is diluted as add more devices but the device count goes up.
We are comparing thunderbolt to PCIe, since it was expressed to me (by you, among others) that it was an adequate substitution.
Where it makes sense. You'll be hard pressed to find a post where I'm cheerleading moving x16 or x8 cards to Thunderbolt.
The question with the MP 2013 is far more so is whether trading two x4 PCI-e v3.0 slots is better than 4 Thunderbolt ports based on the same two lane bundles. I think that more so depends upon what folks were going to toss into those two x4 slots.
If those cards are primarily current v2.0 cards then the downshift to v2.0 with Thunderbolt isn't a big loss. Users were going to do that anyway. If this is mainly lower-mid range eSATA , audio, FW , USB , etc. cards there is probably as much need in the rest of the Mac line up as there is with the Mac Pro's. So going to format that whole Mac line up can share tips the scales relative to some distant future x4 v3.0 cards that might show up.
It's reasonable to say that a board with 5 slots can each run at 4 times TB2 without being throttled is superior in terms of bandwidth. ......a board like that allows for more total bandwidth than the new Mac Pro's TB ports all combined.
The MP 2013 has no need to ship
all the PCIe bandwidth out since there are two x32 worth of PCIe bandwidth that is needed for the GPU cards ( and likely the PCIe SSD also ).
It is just an Apples to Oranges comparison to snarf up all the lanes when all the lanes aren't in question.
As far as PCIe 3.0 16x, my opinion is that it is currently an overkill for everything up to and including GPU--8x (8GBps) seems adequate for the cards on the market. ...--the benchmarks of the 7970 (the "first PCIe 3.0 16x card) running at PCIE 2.0 16x (8GBps) prove that.
You are benchmarking with software optimized for x8 v3.0 worth of bandwidth it should not be too surprising to find out that it doesn't really take advantage of much more. Most of these benchmarks themselves are deeply flawed in evaluating the utility of x16 of v3.0 bandwidth because they have no concept of that kind of bandwidth.
Plenty unless you want more than one SAS port on your new Mac Pro--then you have to buy another $900 SAS controller instead of a single PCIe card with 2-4 ports.
There no good reason a single ( or even dual ) port SAS card would cost $900. In the current Mac Pro, there is no place to put another $900 SAS controller card (presuming don't evict the GPU). The MP 2013 isn't try to expand into markets the current Mac Pro isn't even in.
There are some folks would probably would have preferred the current Mac Pro looked more like the HP z800 , but it doesn't. The MP 2013 moves even further away but example that outstrip the current Mac Pro aren't really very relvant about the MP 2012 - MP 2013 transition.
Yes, it allows me to be a jackass and make the motherboard share lanes, but only if I start utilizing more than 40 lanes. I contend that this is a better option than having two non-replaceable proprietary video cards and six thunderbolt ports with no PCIe expansion--that's been my fundamental point this whole time.
4 x8 slots is a different market than even the current Mac Pro. It isn't going to be suprising if this is a mismatch to the MP 2013.
Also, I like how you were just saying 8GBps PCIe would "throttle" a four port SAS card... I guess we now agree that's not the case?
You are throttling GPGPUs far more than SAS. Still doesn't get past where you labeled these as x16 slots which was only physical.
I think we agree, I'm not sure where either of us got the idea we didn't. There are many on this board (goMac included, IIRC) that stated that Apple had no choice but to remove PCIe due to the addition of thunderbolt.
Apple didn't remove PCIe. They removed standard PCIe sockets and a standard card cage, but PCIe is still there. Don't confuse form for function. The function is still there. The question is whether the form is required.
While Apple could have used just one TB controllers the dominos falls like this. TB means needs at least one embedded GPU. If Apple wants the best possible DisplayPort signal out the TB port then it is a very large embedded GPU.
Next domino fall is need to sell more embedded GPUs to get volume ( not just shifting an iMac GPU over ). A single W7000-W9000 class could supply DisplayPort input for two TB controllers. Another could definitely do two controllers. So that is one x4 slot on the chopping block. Another embedded GPU and could cover both and the whole card cage disappears.
----------
Take a look at that cable cross section. It clearly shows a 20Gb/s capability in one cable.
It is a logical data allocation not a physical one. On the left side there are actually two physical wires in each of those 10Gb's strands. This
not a physical wire diagram on either side.
Try looking up link aggregation.
https://en.wikipedia.org/wiki/Link_aggregation That is really the only major change to TB v2.
A single device connected on a Thunderbolt port is a daisy chain network. With no devices Thunderbolt serves no purpose. With two Thunderbolt devices you have a network. Hence you have a switch involved. You can hand wave till the cows come home, but it won't change the truth.