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.
Thunderbolt Card for 2010 Mac Pro
- Thread starter odinsride
- Start date
- Sort by reaction score
You could by a 27" apple display right now and hook your ethernet/USB/Firewire to the respective ports on your Mac Pro.
What exactly is the TB version suppose to be giving you that you don't have now?
Can't find the link at the moment, but I did read on the Intel site that it won't be possible to make a drop-in TB card. The logic board needs to be "enabled" somehow. That said, a friend of mine with a new pre-TB Mac Pro is in touch with an outfit in Los Angeles that use a lot of Mac Pros, and they say that they will taking delivery of drop-in TB cards (into the PCI slots) in the near future. (My friend was talking to them because he wanted to know how long he'll be waiting for a TB Mac Pro - either as a trade-in or a drop-in).
So - no real useful info, but now we have more speculation....
So - no real useful info, but now we have more speculation....
Not quite. Some Intel folks have suggested would need a new motherboard to support TB as a PCI-e card
" There won't be add-in TB adapters, you'll need a new computer/motherboard that supports TB."
https://forums.macrumors.com/threads/1102818/
which quotes
http://news.cnet.com/8301-11386_3-20035571-76.html
It is possible, it just takes an unconventional design approach.
For example could leverage PCI-e v3.0 so that the GPU and TB don't compete for bandwidth on the same card. Since, the bandwidth is double a 16x card could use 8x for GPU (same bandwidth as old v2.0 16) and still get the TB's 10Gps traffic out in the "unused" portion.
I don't see the video card vendors being excited to create a reference design along these lines though since it means putting more money into Intel's pocket for every card sold. They are already behind Intel in the graphics market in terms of sales. That will just give Intel more money to beat them up with.
Similarly, another workaround could use a short internal DP cable to run the data down to the motherboard and the TB controller chip. You'd need a new motherboard with a DP input cable socket.
Apple has enough money to fund a non-reference design if they wish. Not going to make the video card options on the Mac Pro get less expensive though. Back to non-mainstream video cards.
What Intel is trying to avoid are "PCI-e data out only" ports on computers. Obviously on some peripherals (e.g., a FW-to-TB dongle with one port) that will be the case so it is possible to do. However, it is going to cause some user confusion if sometimes the TB port on a computer can pump out display video and sometimes it can't. The directions can't be "plug your display into the socket with the lightning bolt".
Some of the upcoming, external box, AV capture solutions have TB ports. Previously, , those vendors often sold custom PCI-e cards to connect their external boxes to a Mac Pro. It would not be surprising if they have built new cards for the new boxes. Perhaps they will do like Sony and just not label it as being officially Thunderbolt. Something like "PCI-e Data only over DP socket" connector tech which just happens to work with Thunderbolt. So no lighting bolt symbol on the plug but it works. Along the lines of how USB/eSATA sockets are not part of any standard but they appear on many boxes.
What they kind of need is a variant of Thunderbolt. TB-D which only does data which is an optional standard with a different symbol so user expections aren't the same.
Of the different approaches, this makes the most sense IMO, as it's simple and cost effective.
Hopefully, a standard for such an implementation will be agreed upon, solving this issue for users on any platform and keep costs at a minimum (i.e specifications for an edge connector).
Definitely. Both Data + Video and Data only implementations would have a negative impact on adoption due to the resulting confusion, so Intel is trying to prevent this.
Maybe, but it's not idiot-proof in terms of users spending sufficient time to understand the differences IMO and still end up in a mess.
For example, lets say they use a different symbol. Users that don't pay sufficient attention won't realize they only have a data signal until they've plugged it all up and the monitor doesn't work. Despite it's their own fault, they're probably going to be upset (return costs such as restocking fees and shipping would likely do it, even if they do realize it's their own fault).
I just don't see this ending well, even if they create a different TB connector for the computer side on a Data only implementation (i.e. add a Key to the existing port). The peripheral end can't change, so it could still be connected to a TB monitor which won't work.
USB and eSATA have survived USB+eSATA combo sockets pretty well so it could go OK. There are couple of key factors. One, the standard socket has to be the dominantly deployed one. Second, they keep these non-standard sockets away from the populations that have higher requirements for "idiot proof" connectors.
For the first, if they can get a large number of laptops and all-in-one design to adopt it then that will solve the dominate usage factor. This tech more than any proceeding technology really should be the industry standard docking port connector tech. Nothing better has ever been trotted out and it works well ( Apple's TB display/docking port and Sony's Z docking port ).
For the second, not that many folks buy high end PCI-e Audio or AV capture cards. Capture is inherently in the opposite direction as video out so the "idiot" factor should be low but non zero. If all of the inputs are on one side of the card and the outputs on the other, along with "In" and "Out" labels, that will help.
If someone tries to come out with a "Bubba Joe's TB data only" card for generic PCs, then I think Intel will make their life pretty miserable if they can.
Not sure if the active transmitters inside the TB cables are tuned so that they do DisplayPort also so TB may have another , bigger "idiot proof" problem.
My comments were based on mass consumption, not specialty products. So Idiot Proofing the port as best as possible is in Intel's best interest for fast adoption rates (hence the comments they've made so far).
Granted, a specialty port could be used (or standard TB ports without video) for specialty products that won't be purchased by the vast majority of users.
I don't disagree here.
My point is that Intel's public comments indicate that the initial TB implementations will follow the full standard, as it's in their own best interest as a means of encouraging adoption. Specialty ports on consumer systems this early in the product's introduction could have a serious negative impact on adoption, and thus long term sales/profits.
Once this is done (market dominance), then allowing a mass marketed variant such as a generic Data only solutions won't be an issue compared to allowing it to happen now.
I see the timing of what gets released as critical in order to facilitate adoption. That's all. Sorry if this wasn't clear.
This was the main point of my post.
It's not in Intel's interests for such products to release right now, so they would resort to whatever means available to them to prevent it.
Which is why I suspect they'd only be interested in cards that either have TB chips attached to the GPU card, or are sold with an interconnect between separate TB and GPU cards (needs a standard).
I would expect they are (I'd be shocked if Intel made this big a blunder, given how they're marketing it currently).
You do know you can use the display through the mini-display port on your video card right? So storage can be with a eSATA 6Gb PCI card and mini-display to TB 27" display. Storage and video solved without thunderbolt. Imagine that.
+1
I still don't pretend to understand why so many people are hung up on having the current implementation of TB on a desktop workstation.
It's easy enough to add an eSATA card to your Mac Pro if you need to connect fast external storage. Right now, I think TB has a lot more practicality with mobile and all-in-one computing, where adding drive controller and/or RAID cards are not options.
For displays and storage disks yes.
A little murkier for devices like :
http://www.matrox.com/video/en/thunderbolt/ (MX02 )
http://www.blackmagic-design.com/products/ultrastudio3d/
At one point Avid had a Lightpeak box. Pictured here :
http://www.anandtech.com/show/3930/intel-light-peak-lighting-up-in-2011
&
http://www.engadget.com/2010/09/14/intels-light-peak-optical-interconnect-shrinks-slightly-while/
in one of the earlier show demos. I guess they had to re-engineer it a bit for Thunderbolt since flip-flopped from USB styled ports.
AJA announced both a PCI-e card and a TB solution:
http://www.aja.com/news/index_article.php?id=141
However, there are PCI-e (and sometimes less expensive) alternatives to these TB solutions. The smaller market would be those that needed something connected to Mac Pro sometimes while at other times going out into the field with a TB MBP at other times. For folks with 100% fixed studio work, yeah it is lots of protest for no substantive increase in functionality. Even more so, if Apple would finally leverage the hardware RAID in the chipset with the C600's. For many folks that would free up a PCI-e slot so have plenty of room for AV capture and other things.
P.S. In previous years Apple didn't add ExpressCard slots to a Mac Pro ( for very similar reasons: real PCI-e cards available), so not sure that split-work market is really all that big. But I can see the folks in it making a racket about it.
You might remember this last year we had the same problem with mini display-port, The 06, 07, 08 Mac Pro's where not compatible with the 27-inch LED Cinema Display but then bare feats did a test with a 5870 on a 2008 Mac Pro and it worked and once everyone saw that it work users with 06, 07 08 Mac Pro's upgrading to 5770 or 5870 and using the 27-inch LED with late model mac's I think we just need to wait for 2011 Mac Pro TB GPU upgrade kit becomes available and then it might work
I think you are right about this being a mobile feature worth having. Road warriors who use their laptop as their desktop too can have a one cord connection to an external monitor, wired network, back-up disc... Thunderbolt becomes the elegant docking station.
If you aren't going to connect/reconnect you don't need Thunderbolt. There are existing solutions.
Physically impossible. Will never happen.
Thunderbolt communicates directly with the CPU and on-CPU GPU. The lack of on-die GPU is also why the current MacPro has not been updated with TB.
TB is fantastic for laptops and closed desktops like the mini/iMac because it allows for port expansion. Real desktops like Mac Pro don't need it, it has PCIe slots that can accommodate cards for everything TB is capable of.
Well, until you factor in the other "one cord" needed to connect the other device to the device connected to the Mac. Oh, and the power cord to elegantly connect to the device connected to the Mac and the other power cord connected to the other device connected to the device connected to the Mac. Etc., etc.
Physically possible. Will never happen because Intel and Apple don't care.
FTFY.
Thunderbolt is nothing more then a bridge for PCI-e signals and a clever multiplexer to allow DisplayPort through the same cable as well. It is somewhat analogous to Magma's PCI-e expansion system, which consists of an external chassis with a bunch of PCI-e slots in it and a host interface card that goes in your Mac. You connect the two together with a cable, and suddenly one PCI-e slot in your Mac has become 4 or 6 or 8 or whatever (though you are limited by the available bandwidth to the slot the host interface card is installed in).
Thunderbolt does the same thing, just differently. DisplayPort is not required in any way shape or form. They *could* make a Thunderbolt <-> PCI-e adapter without the GPU bits (which, again, are totally optional)- it'd support everything except monitors.
The problem is that they won't, because they've already established Thunderbolt as a product that supports displays. If they make a version that doesn't, it'll lead to consumer confusion and anger when people inevitably try to plug a display into the card and nothing happens. That's why they don't want to do it- because it'd be 80% of the same thing, but not totally the same, and they don't want to fragment the way the public views a Thunderbolt port.
-SC
It isn't impossible. A sufficiently complicated Rube Goldberg solution could work. But it is not very likely.
Yes, there are really three inputs to a Thunderbolt controller. One is PCI-e (a x4 connection). Second is one or two Display Port inputs. The third a switching fabric commands for Thunderbolt.
Board makers could build a card that takes all three inputs. In fact ASUS already has.
It should be possible attach a microcontroller to the TB controllers switching input that mimics a standard/fix config of a Mac 2010. It wouldn't pass Intel's TB certification tests but it is a giant kluge that would work. If there was enough demand someone might start sells kludges that kind-of-sort-of work most of the time.
It doesn't have to be on-die GPU. That makes implementation easier but iMacs and MBP 15" work just fine with GPUs embedded on the motherboard. The issue is having access to PCI-e controller, boot controller , and GPU display port outputs directly on the logic board along with the TB controller.
"Real" is debatable, but any Personal Computer with more than a couple of PCI-e v2 (or v3 ) slots really doesn't have the significant problem that Thunderbolt addresses.
That is false information.
False.
False.
The ASUS is designed only for their specific logicboard. Its actually just a feature expansion card.
The same thing as Apple used to do when they included an empty math co-processor slot on 68k machines.
Its a cost cutting move that allows people who want the feature to add it while those that don't to save money while ASUS only has to make one board to fill both markets.
False. It does, thats the primary fact preventing the MacPro from getting TB.
False.
A Real desktop is one that is owner serviceable and has internal hardware expansion beyond ram and storage.
False.
The issue is that it needs direct communication with the CPU.
Isn't it funny/sad that this thread is a year and a half old, and we still don't have any idea if/when we willl ever see this?
No, it doesn't...
It needs direct communication with system memory, not the CPU. However, you can do this on PCI Express. Firewire has a very similar requirement, and Firewire PCI cards can pull this off just fine.
Thunderbolt has it's own CPU. There is no reason it needs to talk to the system CPU.
The big difference between Thunderbolt and something like Firewire is the GPU requirement.
http://en.wikipedia.org/wiki/File:Thunderbolt_Technology_model_1_E.png
Where is the direct link to the CPU, apart from the GPU DisplayPort data?
http://guide-images.ifixit.net/igi/QWKKPpAh4AHJm1bf.huge
The blue chip is the TB controller. Red is the PCH (platform controller hub). Source is from here: http://www.ifixit.com/Teardown/MacBook+Pro+15-Inch+Unibody+Early+2011+Teardown/4990/2
If TB required a direct connection to the CPU, the chip would have been positioned in-between the PCH and CPU to minimize the PCB trace length. Instead, it's positioned to the left of the PCH, away from the CPU- because the board layout explicitly follows the diagram hosted on Wikipedia (CPU <-> PCH <-> Thunderbolt).
Once again, Thunderbolt is PCI-e.
Don't believe me? Fine, here's this, straight from Wikipedia:
Source:
http://en.wikipedia.org/wiki/Thunderbolt_(interface)
The prototype Mac Pro motherboard in question was modified to allow the GPU DisplayPort data to be fed back through the PCI-e slots and into the prototype PCI Express card. The CPUs and PCH were the same chips found in the 2009 Mac Pro (in fact, the CPU daughter card was identical to the one that shipped on the 2009 Mac Pro). Both the GPU and backplane board were modified for the aforementioned reasons.
Had they foregone the ability to drive displays, then that PCI-e card could have been manufactured to work in any Mac Pro simply without the ability to drive an external display...
Here's even more crap to backup what I'm saying:
http://www.anandtech.com/show/4194/intels-codename-lightpeak-launches-as-thunderbolt
So once again, apart from the connection from the TB controller to the CPU's built-in GPU, please provide a citation as to why TB absolutely unequivocally requires CPU-specific features if you drop the DisplayPort compatibility.
-SC
Last edited:
False.
Its a known fact it will never happen.
A multiplexer is not a CPU.
You answered yourself.
That is false information.
Please quote a source, wikipedia is not accurate for any form of information.
Thanks for providing proof there will never be a TB PCIe card.
You already have.
TB will never drop video output.
What? Of course it is. It's just not a very complicated one.
Looks like a CPU to me...
And you still haven't said why it requires a CPU. It requires DMA, not a CPU, which PCI Express is perfectly capable of supplying, just like it does with Firewire and serial.
Intel originally demoed Thunderbolt from a PCI Express card, even.
And video output has nothing to do with the CPU. On the iMacs, that's linked to the GPU, not the CPU. Again, no CPU requirement.
Last edited:
OK, fine.
Then someone like Sonnet *could* make a card that either:
A) has a dedicated input port for DisplayPort data, the card comes with a short (<6") cable that plugs into the back of your ATI/NVidia GPU and the other end goes into the aforementioned plug on the back of the TB card
B) Has an on-board FPGA that simply generates a static image that literally says "DisplayPort not supported on this card"
What is impossible about either of these to you? Such a product would not be cheap, but we're not discussing costs here. We're discussing the physical possibility of someone somewhere manufacturing a card that is compatible with the existing generation of Mac Pros.
There is no physical impossibility. A third party company simply hasn't bothered to do it, and neither Intel nor Apple are interested in doing it themselves since TB is completely pointless on a Mac Pro anyways.
-SC