Not particularly. It is the basically the same task done 3 times. That may need to dilute one of the x4 bundles into two x4's isn't that big of an increment.
None of that appears to be particularly accurate at all.
1. The dSL4510 is Thunderbolt 1.0.
No change.
2. Sincethere is no change it has exactly the same PCI-e v2.0 links.
"... These two are replacements for Intel's current DSL3510 and DSL3310, with 4/2 and 2/1 (channels/ports) respectively. There are no performance changes other than official support for DisplayPort 1.2 (and thus 4K displays) ... "
" ... There's still a PCIe gen 2 x4 interface on the other end of these controllers ... "
http://www.anandtech.com/show/6885/...e-20gbps-falcon-ridge-tb-controller-announced
Intel has said a long while back that Thunderbolt v2.0 wasn't coming until about 2014. It is middle of 2013 which means anything new is likely still TB v.10.
Apple is claming the the Mac Pro will get Thunderbolt v2.0 chips. Those aren't on the market yet. There is an outside chance that PCI-e v3.0 could be added but none of the bandwidth improvements apply.
Outside of the Display Port 1.2 input streams changes in these interium TB chips and followed up on with Falcon Ridge (TB v2.0 ) the changes are more so in shuffling Thunderbolt's backbone bandwidth
not the on/off ramp network edge connections.
" ... Named Thunderbolt 2, this next generation of the technology enables 4K video file transfer and display simultaneously thats a lot of eye-popping video and data capability. It is achieved by combining the two previously independent 10Gbs channels into one 20Gbs bi-directional channel that supports data and/or display. ..."
http://blogs.intel.com/technology/2...ndwidth-enabling-4k-video-transfer-display-2/
Intel reshuffled the
same aggregate network bandwidth that was already there. Just rearrange the deck chairs. Not tweaking the PCI-e edge. That would be a horrible network upgrade to not increase overall bandwidth at all but increase the edges substantially. That is exactly backwards of the fat tree network (
http://en.wikipedia.org/wiki/Fat_tree) they should be trying to setup where the backbone is like 2x (or at least greater than ) the edges.
x1 to the cannonical controllers "tree leave nodes"
x4 from host to TB network "next highest "
> x4 for node-to-node TB traffic " highest "
Cranking up the bottom and leaving the top the same is only going to lead to network congestion and craptasic QoS on data delivery.
The DSL4510 is a stopap to all folks to put the TB socket into DP v1.2mode so it can take over and run 4K video
TB v2.0 is somewhat of a kludge so that don't have to drop back to DP v1.2 mod to get 4K video. They do that by "robbing Peter to pay Paul" and merge the two separate channels into one and then doling out more bandwidth to the 4K video traffic. It is obvious which has priority here. It isn't PCI-e data traffic.
If there is no 4K traffic (or any video) then will get better network QoS, but there is no indication at all that host side throughput is going to go up.
In that alternative world sure, but in this one it has got problems.
Apple has pretty mastered this. The problem they have in the > $2000 zone though is that there are lots of "pretty good" computing alternatives under that price point that are "good enough" for increasingly largely number of people.
There is a point at which you can be too much of a Scrooge McDuck. Too much of cost-cutter and fewer folks will by the high end goods . The Mac Pro needs to very clearly provide more perceived value to justify the higher price. Apple can blowing smoke up folks butt by saying the new Mac Pro is super magically delicious, revolutionary and cures world hunger but after a while that Joe Isuzu stuff really don't work so well. They actually have to add something with value to the system.
The system is smaller. Probably easier to make after tool-up for the new set of parts. Certiainly easier and cheaper to ship and inventory. Apple can take those savings and put higher value added parts into the system. The margins stay exactly the same, ( i.e., they tread water.)
Dual GPUs ... well guess what the other CPU is gone in half the line up. Trading X for Y and the mark-up is the same leads to same margins. The lowest end of the line up had even bigger margins. Goosing the entry box's margins into the stratosphere certainly did
NOT help the overall product. In fact, it appears to have almost gotten the whole product category cancelled.
If gen 2 can stick on same switch at TB controller. Not a big deal.
None of that is in Texas ( an old Dell plant) where reported the new USA Apple line up is to be made. So still nothing here screaming Foxconn.
Honestly Apple needs other working relationships than just Foxconn. Jobs was good at some things but sometimes his OCD got in the way. Einstein had quote. "Make things as simple as possible, but not simpler."
Having multiple vendor relationships is necessarily for a global mutlinational company. Trying to feed everything to exactly one supplier is not a good idea long term.
Yes, but I'm under the impression it still uses one of the gen 2 PCIe lanes to transfer that data.
I'll dig a bit later, as I am curious.
Doubling the the "on ramp" isn't going to change the freeway speedlimit. Those actually should be operating at two different speeds. Chaning the native thunderbolt ( not native PCI-e ) data throughput speeds will change Thunderbolt bandwidth. The transported protocols data operate their own, slower speed and bandwidth.
If Transporting 4K video there isn't 20Gb/s available for PCI-e data traffic.
If not transporting 4K video 20Gb/s improves QoS.
that is basically what Intel is saying above. Bonding two separate channels into one channel is reshuffling the exact same deck.
The PCI-e v3 protocol are completely oblivious to what is happening on the Thunderbolt network. Changing them isn't going to make TB backbone network faster. In fact the drama for the TB network just increase because not has to do more isochronous traffic management than before when the traffic was more hard segmented.
Right which what it would dubious to choke faster by putting a faster "on ramp" for the PCI-e data ( i.e., a v3 ramp.)
Until Intel can actually move the aggregate bandwidth forward, the PCI-e data "on/off ramp" is likely going to be stuck at v2.0. Couple that it will be hard for x1 v2 controllers to go to 0.5 v3 controllers.... ( they don't need any more speed) and the TB will also more inclined to stay at v2.0
The notion that Thunderbolt was primarily creating for ultra super duper DAS and/r utltra Gamer eGPU is just way disconnected from the reality.
Apple's Thunderbolt docking station/display. That's what Thunderbolt is aimed at.