I get what you mean, but from an upgradability perspective it has effectively replaced PCIe slots as an option, which is what slughead has been getting at;
PCIe slots are not PCIe. What he is getting at is muddle the issues and perspective.
instead of popping a PCIe card into the new Mac Pro to upgrade it we now have to find either a suitable USB 3 or Thunderbolt alternative,
You can still pop in a wide variety of PCIe cards into an Expansion devices. Cards are still "popped in". The 'where' has changed but the action of 'popping' has
not. The aspect of being internal and requiring a larger open rectangular hole be cut into the outer surface of the case has been replaced. The underlying technology has
not been.
Replaced means it would disappear from the system. PCIe data is still being pumped in/out of the mechanism being used.
The physical mechanism for expansion has changed. But it is really form rather than function that has changed. It is not the fundalmental PCIe function that has change. It far more a physical format change.
so in that respect it has replaced PCIe, whether PCIe as a technology still exists inside the machine or not.
Either it is removed or not to effect a replacement. If the technology is
still there it hasn't been replaced. PCIe technology is still there whether Thunderbolt is in the solution or not. That does not change.
Specific physical slots? Yes. The technology? No.
They're using a proprietary connector though, so it's possible the connector can handle say 20 lanes, giving 16 to the GPU and 4 to the SSD, with the latter just not being connected/used on the other card. So I'm not sure the SSD will share with a GPU, at least I certainly hope not.
SATA Express only needs x2 lanes. x2 PCIe v3 lanes provisions SATA Express for 2GB/s. That is enough to cover the 1.2GB/s the Mac Pro's PCIe SSD needs.
x4 lanes would only be required if using some SFF8639 derived solution.
http://www.anandtech.com/show/6294/breaking-the-sata-barrier-sata-express-and-sff8639-connectors. That doesn't particularly sound like something Apple would do, given using SATA express solutions elsewhere in the line up. The difference between rest of line up is more likely using PCIe v3 as opposed v2 for the x2 lanes. That Flash controller will cost more now, but in future generations not quite as much so could be trickled down to lower priced systems. However, if vendors are leaving PCI v3 SATA express as a future product category it may be only option now.
If Apple found a x4 v3 -> ( x4 v2 , x4 v2 ) switch/splitter that is non-blocking (e.g., can multiplex those two v2 streams into one v3 stream without dropping all three ports/bundles back to v2 speeds ) then perhaps two of the TB controllers are sitting on just one set of x4 PCIe v3 lanes. x2-4 lanes sent off to the other x4 PCIe v3 lanes would work. I'm not sure the TB controller is going to be 'happy' sitting behind another switch though.
You're right that the maximum bandwidth does look strained, but I doubt many people are going to be pushing the limit so hard that they couldn't have a second PCIe SSD.
If for this generation the SSDs is squeezed into the PCIe v2 space then don't really have good options right now. With limited v3 lanes should be looking for v3 implementations to make more effective use of the lanes.
A better support chipset to pull discrete USB 3.0 inside and perhaps also allocate some more PCIe v2 lanes to relieve pressure on the v3 lanes would help long term.
It just seems weird not to at least leave the choice up to the customer; especially if adding a second 1tb SSD meant you didn't need external storage.
If storage requirements were for 2-3 TB of storage you do. But yes, that the mini and iMac have two storage drives and the Mac Pro just one is extremely odd. I suspect it is a "version 1" quirk will go away eventually but seems this specific instantiation is stuck with.
Of course I agree, and I've supported Apple's decision to do this kind of thing in the past. But it's not quite the same in this particular case; when SCSI was abandoned for Firewire and USB, it was still easy (and reasonably affordable) to just pop a SCIS card into a PowerMac G3 and carry on using your old devices. Plus Firewire was an obvious upgrade to SCSI thanks to being hot-pluggable, faster etc.
Not so sure that is useful way to frame the context. Apple isn't going Thunderbolt to improve the Mac Pro only. Apple is going Thunderbolt to improve the
whole Mac product line up.
Is the Mac Pro being pulled back into alignment with the rest of its siblings? Yes. The tail doesn't wag the dog. The issue was really whether the Mac Pro survived at all. Was there going to be any Mac Pro in the line up. If the Mac Pro isn't contributing to Mac growth or significantly the revenue it has problems standing all by its lonesome. The fact that Apple was willing to let the classic Mac Pro product get cut off from the whole EU market for almost a year is illustrative that being a highly strategic or tactically significant product in the previous form factor was not an issue.
For the whole Mac line up Thunderbolt is a win. Faster ( either better than nothing or completely stomps on ExpressCard ) and can attach the previous technology if needed at reasonable prices. The Mac Pro aligning with Thunderbolt will assist in jumpstarting more growth in the TB device market. That will have overall good impact on overall Mac market. Overall healthier Mac market means Apple probably has motivations to allocate resources on Mac Pro. So dog wags tail.
Thunderbolt isn't quite as obvious a transition from internal PCIe slots, as it's not as fast, and keeping your legacy PCIe cards means buying extremely expensive PCIe enclosures.
Thunderbolt isn't a transition from all internal slots. That is a swamp that folks keep spinning, but it is not what Thunderbolt was designed for. Nor is that the usage as instantiated in this new Mac Pro design. So there is no transition inherent in the new Mac Pro design on some sweeping generalization broad front.
In the x4 card slot space, which are legacy cards keeping? The extra FW ports? The USB 3.0 ports? The eSATA ports? Those that need the ODD and a HDD sled there are other options than to keep push the legacy card forward.
http://www.sonnettech.com/product/echo15thunderboltdock.html
Thunderbolt has far more traction when there is aggregation than in chasing after 1-to-1 replacements. Swapping a collection of a eSATA card, USB 3.0 card , DVD drive for the above is far more cost effective than attacking the collective problem with a external PCIe enclosure and a discrete DVD drive.
Yes there are classes of audio cards, a/v capture cards , etc. that will require PCIe expansion enclosures, but not all of the transitions have to a enclosure.
Want your own full bandwidth x16 PCIe card? That isn't "replaced". That is in the same discontinued bucket as dual CPU packages, XServe, XRaid , laser printers , Cube , six slot Quadras , MacBook , etc.
I'm not saying that they should consider eSATA or Mini SAS as standard options, but they should have either added them to the Mac Pro or provided suitable Thunderbolt cables/devices to ease the transition.
Apple doesn't provide/sell a Thunderbolt cable ? Or you mean enclosed in the box with the system with every system? Certainly this isn't Apple enclosing TB devices with every system sold. Apple could bundle TB dockstation/displays with every Mac Pro sold, but I bet there would be even more howls of protest over the price and suitability than the non-bundle any device option. Which device a users needs is highly dependent upon that user's configuration needs. There is not likely going to be one TB device that makes everyone happy.
Furthermore, In those two narrow cases they aren't really needed or particularly desirable as those ports. If just aggregating multiple SATA or SAS channels then don't really need external SATA/SAS ports. The requirement is actually really for an additional SATA/SAS controller; not the ports. Moving the controller to the external box solves the problem. It doesn't re-use the card but core issue is connectivity to the actual storage devices.
Should Apple be in the business of selling JBOD and RAID boxes? Probably not. Apple and its $100B cash horde is far more likely to serve as an additional barrier to entry than any help here. First, Apple is already the largest, by far, buyer of TB controller chips than anyone out there. Apple's discounts and economies of scale are far bigger, so more than a few lower margin competitors are going to be scared off by that. If Apple wakes up one morning and decides to crush them, they can. That is far less likely to happen if Apple takes the position that those devices are primarily a 3rd party opportunity.
Second issue exactly as alluded to above. It is a diversity of TB devices that are need because the user needs are diverse. Several end product vendors is what is required so that a healthy economic market can spring up. Apple isn't going to build everything for everybody. Apple is going to deliver TB devices that are needed for a broader spectrum of Mac users to either cover Apple specific issues (aggressive socket retirement) or the need to complete the user interaction system (monitor). The three they have now ( TB-> FW , TB -> Ethernet , TB docking station) cover needs by some user's need on each of the Mac product lines, not just on any single Mac product.
But, but., but the GPU card market needs huge diversity of vendors. That is kind of funny because over the last 6-8 years it hasn't demonstrated that very much at all as a large viable market. Strip out the cards vendors created during that time period that Apple commissioned as part of their configs. Now strip out all the hacked non mac cards that were enabled by those cards (there is no vendor working with Apple to bring those to market). You have what maybe 1 card/year if you're lucky. Contrast to the somewhat slow paced, but now growing, TB device market and there are stark differences. There are 3-4 PCIe expansion vendors now. There are 3-4 external storage vendors. 3-4 docking station vendors now. When was there ever 3-4 GPU card building vendors in a single year over that previous 6-8 span? Or 3-4 vendors in the same GPU card subcategory. Even when had multiple vendors it typically was spread around as to not compete much head to head with anyone expect perhaps Apple.
If Apple creates a 3rd party vendor market opportunity and it doesn't take off over a protracted amount of time, then eventually it will probably get canceled.
