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.
What's the word on thunderbolt via PCIe cards?
- Thread starter Ravich
- Start date
- Sort by reaction score
I don't this is gonna happen. TB needs a direct link to chipset afaik. Plus I really don't know how video to TB port is gonna work with PCI-Ex video cards.
Not going to happen. Intel themselves said TB requires a hardware connection to the GPU. This is clearly evident since TB steals half the GPU's bandwidth in the MBP.
So the only possible way there might be any add-on cards is built into a video card.
So the only possible way there might be any add-on cards is built into a video card.
It's just odd that all through development no one had any doubt that we'd be seeing them, and all of a sudden upon release there is this "needs to be on the motherboard" business.
I just cant help but wonder why you'd need thunderbolt on a mac pro, or anything other than laptops (and of course the imac).
For me its a great idea, but, well. We have internal storage space for desktops, with up to SATAIII now, plus esata for external, plus many other I/O options.
I just dont see the need for thunderbolt on desktops, other than to enable you to use the peripherals.
Unless the thunderbolt port could be used as some crazy fast data and video network connection.... to a thunderbolt switch..... with fibre storage hanging out of it.... Now that would make it interesting.
For me its a great idea, but, well. We have internal storage space for desktops, with up to SATAIII now, plus esata for external, plus many other I/O options.
I just dont see the need for thunderbolt on desktops, other than to enable you to use the peripherals.
Unless the thunderbolt port could be used as some crazy fast data and video network connection.... to a thunderbolt switch..... with fibre storage hanging out of it.... Now that would make it interesting.
(from another post)
Comparing USB 3 to TB is like comparing VHS to Blu-ray. TB is REMARKABLY faster than USB 3. Couple that with the fact that TB allows for simutaneous transfer in BOTH DIRECTIONS at consistently close to theoretical speeds of 10 Gb/s. USB however as always, fluctuates in speed and almost NEVER reaches close to theoretical speeds and is only one way transfer. TB can daisy chain up to FOUR 1080p HD streams simultaneously (if I recall correctly) while I don't think USB 3 can do even 1.
If you do video editing, you might be working with cameras or external RAIDs which can push much much more data than FW800 can handle. For cameras you're currently stuck, while external RAIDs usually are done over fiber channel right now.
Internally you can only put in 4 drives, which isn't very much for real pros.
USB3 can stream 1080p. So can Firewire 800. I've streamed 1080i MPEG2 over FW400 before from my TV.
If you're importing from a camera, you're going to want faster than real time though.
It doesn't technically need to talk to the video chipset. Also, you could just put a video chipset on the same card to solve the problem.
TB definitely does not need direct access to the chipset chipset.
It wouldn't be that hard.
There's two primary methods that could be employed IMO (doable, and keeps costs to a reasonable level):
- Use an edge connector (flexible PCB similar to an SLI or Crossfire bridge) that connects a compliant graphics card (needs the edge connector, so existing cards wouldn't work) and TB card. Ideally, such a connector would be established via an open standard (i.e. agreed upon by Intel and their development partners and graphics card industry).
- Place the TB chip directly on the graphics card (this presents more technical challenges due to PCB real estate and cooling issues, but it's possible). Of course, there's more of a cost increase this way, but it's a guarantee the TB port will carry DisplayPort information as well as data sent over the PCIe bus.
Yeah, I know.
There's no technical reason it must be soldered to the main board, but there are issues with adoption to consider if implemented on desktops (systems with slots; no GPU in the CPU <IGP> or an embedded part soldered on the main board). Specifically, the confusion that would be created if some implementations are data only and others data + video.
As it's primarily aimed at portable devices (would be a viable port tech in AIO's as well), it's easy to just add the TB chip to the main board as the GPU is either in the CPU or embedded (separate chip that's soldered to the board), and wire it up (also guarantees that the TB port will contain both data and video information = no confusion).
Combine these two facts, and that accounts for the confusion IMO (i.e. by keeping quiet about a PCIe version, they keep their options open while allowing the portable market to drive adoption).
This is the most realistic reason to have a TB card in a desktop ATM.
As the speed is increased however, this could change (i.e. additional features added, such as adapting it to the enterprise market as an inexpensive competitor to 10G Ethernet or Fibre Channel networking).
Considering the first devices being put out on the market are medium size RAID arrays, I think it's already being positioned as a fibre channel alternative.
There are ways to get more disks internally, or external enclosures can also be used. So using a desktop can allow for faster throughputs for storage, faster than TB in fact.
There are faster solutions (i.e. fast CF card readers), but they're expensive, and lugging a MP around isn't all that convenient.If you're importing from a camera, you're going to want faster than real time though.
Thus using a laptop with a camera via a TB port to pull in the data, then edit on a workstation makes sense. Ideally, using an external storage systems, such as the Promise Pegasus R4 or R6, would allow for editing to be performed immediately rather than transferring data, then getting started.
But this is a relatively rare user in the entire scheme of things (including consumer users).
I don't, as it can't be used for networking as it currently exists.
What it would be good for, is when a laptop and Promise R4 or R6 are used to record camera data (i.e. location shooting), then bring the data back to the office, attach the enclosure to the workstation, and begin editing (could shave time by not having to transfer data before editing can begin, or if it's done for say data security reasons, it's much faster assuming the pool it's going to is faster than a single disk, such as another array).
It would be very useful for video professionals, and that's where TB is being targeted at from what I get from Intel's site (audio as well, but they can potentially use a laptop as their primary machine with fast data access, which can simplify matters as well as reduce costs).
Both incorrect.
The TB spec requires a GPU and TB itself is nothing more than an external 4x PCIe2.0, which means it needs direct access to the logic board to steal the PCIe lanes.
External enclosures linked by eSata, which doesn't come on the Mac Pros, and depending on the drive count, may not be suitable...
I don't see how using a card reader solves the speed problem. You're still pulling the data off at the same speed, just off a card reader instead of a camera.
Again, not sure how this solves the importing from camera problem... Unless you're suggesting I buy a dedicated MBP to import footage into my Mac Pro.
Right, but I think given that we've already maxed out the speeds of USB3, pigeonholing it as a "good for portable Macs" port is kind of disingenuous. Any situation in which one could use it for a portable Mac is also useful for a Mac Pro.
Networking would be nice. But we've certainly maxed out the ports we already have.
I was thinking more in terms of hardware RAID controllers, as they can exceed the throughput limit of TB.
In the case of eSATA, you do have to add a card to a MP, as Apple has never seen fit to include an eSATA port. But they're also cheap, and using a Port Multiplier enclosure will be cheaper than a TB alternative (for now anyway, but this may not change either, as to create a PM enclosure on TB, means adding another chip to the PM board - TB to eSATA bridge).
It comes down to the interface (figuring an existing camera that uses FW800 or USB 2.0).
There are eSATA and USB3.0 CF readers out now, and they're not expensive (either one can get CF to read faster; at or near it's limit). Faster cards on a faster interface will speed things up. It's not ideal, but it's due to the limitations of the technology being used (i.e. too expensive to dump the existing camera for a newer unit that has TB when they arrive).
Now as per using TB in the camera, I agree that's the way to go (also presumes the storage media is fast enough TB is warranted in terms of bandwidth BTW, or it's a waste). But I'm not aware of them being out yet.
No, the example was based on location shooting (easier to haul a laptop and drive enclosure than MP and any accessories needed, as well as the camera itself). With airline fees lately, there's further reasoning to go for the laptop method.Again, not sure how this solves the importing from camera problem... Unless you're suggesting I buy a dedicated MBP to import footage into my Mac Pro.
As per importing data off the camera directly to the desktop, see above (need a TB camera to do what you're thinking + TB port on the MP). Neither TB cards or cameras are out yet, and the cameras will be expensive. So finding a faster interface for the CF cards (i.e. eSATA or USB 3.0) could be a cheaper way to go vs. buying a new camera and selling off the existing one.
In terms of maxed out, do you mean in terms of single device or daisy chaining?
For single devices, such as HDD's, can't max it out yet. SSD's at 6.0Gb/s are the only things that come to mind in terms of a single device that can. As you daisy chain/add devices via hubs, the odds of maxing it out do increase (you'd certainly throttle with a pair of fast 6.0Gb/s SSD's running on USB 3.0 for example, and it's not that hard to do with USB 2.0 or FW800 either).
As per TB in the laptop/portable device based systems, that's the primary market, and has been clearly indicated by Intel. It's also being followed suit by system vendors (laptops/devices are the first products to get it, and I expect AIO's won't be too far behind).
Now I'm not saying it doesn't have it's uses for desktops with PCIe slots (niche market that actually would benefit from using it), but there's better ways to utilize the PCIe bandwidth than TB. Existing port technology may also be cheaper down the road (definitely the case now, as there's little TB gear out).
I'd love to see networking capabilities added (makes clusters more accessible for example due to the lowered cost).
In terms of a single common port that's being maxed out, there are instances where they are (namely do to SSD's in terms of a single device or attaching enough devices on ports that can use either hubs or Daisy Chaining). Which is why TB has serious merit in the laptop/device market. Particularly with systems that don't offer ExpressSlots.
But I don't see it as a 'Holy Grail' in terms of desktops. There are certainly limited uses, but I don't see it taking over for existing ports, when it's already possible to make better use of bandwidth by using the existing ports (not just data, but video - existing monitor specifications can exceed what TB can transfer, which is particularly important for larger displays).
I would be happy with a TB card on the Mac Pro that doesn't handle video. It would be worth it JUST to connect to external peripherals such as a RAID SSD. Yes it is quicker to stick more drives inside the Pro but I want to work on my MBP then stick the drive onto the MP then back to the MBP. With TB I can have the application and all the data files on the external drive running at full speed and move it between computers at will. Even if the application will not live on an external drive keeping large amounts of application data and libraries on the ext drive means I can work on whichever machine is most convenient. I appreciate that they could both work on FW800 but we are seeing drives getting to that limit in an array now.
Matrox has announced a PCIe Thunderbolt card, albeit on the expensive side (MSRP of $299).
Not unless the cost of the TB chip is significantly reduced (currently, it's ~$90 per unit in quantity
). Expensive little devils, especially for something that was hyped as a low cost, high speed interconnect. 
