Lightpeak and Mac Pro

[seek3r]

Yeah, I realized that... Did they add a USB controller to the card?

I *think* the card was just modified to poll the motherboard's controller, but I could be wrong... Now I'm curious, I'll power up my Windtunnel G4 later and see if I can check.

Hopefully, although if you are trying to drive 3 ports that's a lot of bandwidth...

I hadn't thought about that.... they could do what used to be done with dual link DVI, first MDP is TB, other 2 are display only

Yeah, kind of curious to see what NVidia says.

Hopefully something to the order of "hell yeah" (followed by an awesome guitar riff and putting on some sunglasses).

According to Intel the copper connectors aren't going to change, it'll just convert to optical after it enters the cable.

Point taken, though I do wonder if that'll change, otherwise I'd be worried about cables being out of consumer price ranges.

Macrumors: We're pretty much making it up.

making it up *with style*

 

[dyn]

AFAIK, Thunderbolt won't work on current and previous Mac Pro's. It's physically impossible with the current architecture and isn't going to magically happen. TB requires graphics processing too and it's not something you can do with PCIe.

Of course it doesn't require graphics processing! If it does that would mean every TB device needs a gpu... The magic happens in the TB processor, it directs the incoming traffic to the correct bus: DisplayPort for audio/video and PCI-e for everything else. So it is technically possible to use it in a Mac Pro and they have even proven that when they first announced Light Peak. The demo machine that was used to show it working was...a Mac Pro! People tend to forget this.

And then there is the way TB works. It uses PCI-e so it can present the various stuff as a new PCI-e device to the OS. Videocards also use PCI-e so you could in theory use an external device for video stuff. I don't think it will work this way since DisplayPort is explicitly used for the audio/video stuff.

Intel never ever has said that a GPU is mandatory, neither have they said they will never ever make a TB PCI-e expansion card. They also haven't said they will. In other words: we simply do not know.

 

[goMac]

Intel never ever has said that a GPU is mandatory, neither have they said they will never ever make a TB PCI-e expansion card. They also haven't said they will. In other words: we simply do not know.

I think the secondary issue that crops up is if they allow data only TB, you have some ports that will be display only, some ports that will be data only, and some ports that are both! Bleh!

They should have gone with a new port.

 

[dyn]

And I think just that is the discussion Intel is having: "should we or shouldn't we?". Just because something is technically possible doesn't mean you'd want it.

 

[300D]

Of course it doesn't require graphics processing!

It absolutely does on the computer side.

 

[DeeEss]

It absolutely does on the computer side.

Yeah this is what I've come to understand also. Well I certainly do hope it will be possible.

 

[dyn]

It absolutely does not on whatever side. If you want/need audio/video you might need it. Technically TB will create a virtual PCI-e device and present that to the OS. Since you can buy audio and videocards for PCI-e...do the math.

There is only one reason why they use mDP: it already does audio/video so there is no need for something else. Simply use whatever is already there. This also eliminates the problem of having too many audio/video devices (a practical problem, not a technical one!).

If you want a really good working port you want it to be the most convenient port there is. Just plug it in and it will work. That's why you do not want to put it in a Mac Pro. On the other hand that sort of defeats the point of being universal and the ease of use for the TB devices (you can't use them on every computer). That's the dilemma Intel is facing.

 

[300D]

It absolutely does not on whatever side

That is false information per Intel's own statements.

 

[dyn]

That is false information per Intel's own statements.

Your post is false information per Intel's own statements and demo indeed. When will you stop with the false information, you seem to be living off it...

 

[goMac]

That is false information per Intel's own statements.

It seems to be a paper requirement from Intel, but a graphics signal is in no technical way needed to add Lightpeak to a computer.

Intel has demoed it repeatedly without a video signal as well.

 

[hector66]

I remember reading that add-on cards would not be offered and that Thunderbolt would (could?) only be supported at the motherboard level. I cannot remember if that was Apple's official position which wouldn't be a surprise but would not keep third parties from doing it, or if it was Intel's statement.

Since this has such a direct connection to the PCI-e bus, I'm thinking that the limitation may not be Apple's need to sell brand new computers.

 

[300D]

Your post is false information per Intel's own statements and demo indeed.

That is false information. Please provide documentation proof that states otherwise.

Intel has demoed it repeatedly without a video signal as well.

What Intel demoed was Lightpeak and what was released is Thunderbolt, two completely different technologies that only share traits.
Lightpeak is a standard of its own that operated on its own terms.
Thunderbolt is based in PCI Express, uses the Mini-displayport format and is hardware tied to the GPU, stealing some of its bandwidth.

The only way thunderbolt will ever be sold as an expansion card is included on a videocard.
Since its currently a proprietary format with Apple having "exclusive use" of it, there won't even be the possibility of expansion cards until next year.

 

[nanofrog]

That is false information. Please provide documentation proof that states otherwise.

There's a lot of confusion going on over this, but it's technically possible to use TB for DATA only, as it seems the GPU data (DisplayPort) is just a Pass-Through connection (keeps it simple and cheap).

Intel's own public statements haven't really cleared this up, but if you look at both the Tech Brief (.pdf file) and Thunderbolt Tech Page, both show laptops as the intended target of this particular chip.

BTW, they've not yet released a full Datasheet for designers. Combine this with the above information, it lends me to think they're not intending to sell it for systems that aren't either laptops or AIO systems right now, as they want both PCIe and DP data on the connector to avoid the inevitable confusion that would arise over some users with DATA only capabilities (both have GPU's either integrated into the CPU or separate chips that can have their DP output routed to the TB chip).

What Intel demoed was Lightpeak and what was released is Thunderbolt, two completely different technologies that only share traits.

No, TB is just a cut-down version of LP in a sense.

LP's demo (workstation), used optical signals rather than electrical. So it included the optical trancievers needed to make that aspect function (they didn't test out DisplayPort at that time IIRC).

All the did with the shipping version of TB, is take those transcievers out, which are now to be included as part of the optical cables when they show up. That's it.

It's made it simpler and cheaper on the system end.

BTW, the laptop demo was on a MBP, and is the TB version (port is electrical). But they did demonstrate DisplayPort over it, as well as daisy chaining. Video of the MBP demo.

The only way thunderbolt will ever be sold as an expansion card is included on a videocard.

That's one possible method to bring TB to desktop users.

Another alternative would be to get GPU card vendors to agree to a standard to move DP data over say a flexible PCB (like a Crossfire connector) to a PCIe based TB card to get a video signal to it. Cheaper for GPU card vendors to do this than include a TB chip (just some trace routing and notching required to the PCB).

 

[goMac]

What Intel demoed was Lightpeak and what was released is Thunderbolt, two completely different technologies that only share traits.

No, TB is just a cut-down version of LP in a sense.

LP's demo (workstation), used optical signals rather than electrical. So it included the optical trancievers needed to make that aspect function (they didn't test out DisplayPort at that time IIRC).

All the did with the shipping version of TB, is take those transcievers out, which are now to be included as part of the optical cables when they show up. That's it.

This. TB and Light Peak are the same technology. I suppose if you want to be correct, you could say that the data stream that is in shipping TB is absolutely identical in every way to Light Peak. You could still implement it as a data only stream and just leave the video stream dead. The only equipment that would care about that would be displays.

 

[Darien Red Sox]

I remember reading that add-on cards would not be offered and that Thunderbolt would (could?) only be supported at the motherboard level. I cannot remember if that was Apple's official position which wouldn't be a surprise but would not keep third parties from doing it, or if it was Intel's statement.

I read this too, I think on Intel's Thunderbolt page.

 

[nanofrog]

TB and Light Peak are the same technology. I suppose if you want to be correct, you could say that the data stream that is in shipping TB is absolutely identical in every way to Light Peak.

It is (no difference in the protocols).

My point as to what was tested out with LP and what's shipped, is a slight variation in terms of the optical cabling (original workstation demo had the transcievers on the PCIe card; TB <version that's made it to market> is electrical only, with the optical cables to contain the transcievers in the cables themselves - possible since TB does have power).

You could still implement it as a data only stream and just leave the video stream dead. The only equipment that would care about that would be displays.

Exactly.

Now whether or not Intel will sell parts for DATA only usage right now, is a bit unclear, but the indications are that they won't in the immediate future (want a DP signal over TB to be consistent between products; not some systems do support both DATA and DP, and some do DATA only).

I expect PCIe cards will be avialable at some point when a DP signal can be routed from a graphics card to a TB card or TB is included in graphics cards (able to share equipment with laptops, such as monitors and storage systems). But TB isn't all that useful vs. laptops otherwise IMO as non-AIO desktops have PCIe slots for user customization (can allow for better use of bandwidth per specific requirements; TB is a big step forward for laptops, but does have bandwidth limitations for simultaneous useage of high bandwidth devices in a daisy chain configuration).

 

[DoFoT9]

There's a lot of confusion going on over this, but it's technically possible to use TB for DATA only, as it seems the GPU data (DisplayPort) is just a Pass-Through connection (keeps it simple and cheap).

so basically, at the moment it doesnt look possible for this with TB - but maybe with future LP releases/expansions (next year?) that it may be possible?

 

[dyn]

That is false information. Please provide documentation proof that states otherwise.

That's not how it works. You claim it is false information so you're the one that should backup his claims. Anyway, if you looked at Apple's and Intel's pages about Thunderbolt you'd have the documentation you want. Also there is something called MRoogle that you can use to dig up some articles on Macrumors about Light Peak and Thunderbolt that also provide the documentation you want. Then there is the live coverage of news.com of the Intel Thunderbolt announcement and their article summarising this live coverage. If you search on YouTube you also get a lot of stuff about this.

What Intel demoed was Lightpeak and what was released is Thunderbolt, two completely different technologies that only share traits.
Lightpeak is a standard of its own that operated on its own terms.
Thunderbolt is based in PCI Express, uses the Mini-displayport format and is hardware tied to the GPU, stealing some of its bandwidth.

That is incorrect information and a complete misunderstanding about Light Peak and Thunderbolt. Light Peak was the name of the project for a new kind of "universal" protocol which eventually was released as a finished product named Thunderbolt. Thunderbolt/Light Peak is based on both the PCIe and mDP formats. It uses mDP for audio/video and PCIe for everything else. This is clearly explained and showed by Intel (see above where you can find that proof). Thunderbolt sits between the mDP connector and the PCIe & videocard. It will automatically split the signal accordingly. If you take a look at the live coverage they'll get a bit more in-depth and explain that it presents things like a PCIe expansion card to the OS.

The only way thunderbolt will ever be sold as an expansion card is included on a videocard.

By saying that you show that you indeed do not have any clue as to what Thunderbolt is. The way you view it, Thunderbolt would be hardware tied to the PCIe controller AND the GPU. That combination makes any kind of expansion card impossible. However, Thunderbolt is not something that requires both PCIe and GPU which is why in theory you can have something like an expansion card. Like I said earlier it's not about what you can do on a technical level, it's all about wanting to do it or not. Not tying GPU and PCIe together means you get two kinds of Thunderbolt devices and this is what makes it hard for users. If you advertise a device uses Thunderbolt people think it will work with any computer with Thunderbolt. In reality it doesn't. It has the ability to create confusion and you don't want that.

Since its currently a proprietary format with Apple having "exclusive use" of it, there won't even be the possibility of expansion cards until next year.

It is not since Intel is promoting it actively and denying Apple has any exclusive use (they're simply the first as they were with the Xeon Nehalem cpu's). Canon is the fourth manufacturer to use the standard (the others are Apple, Lacie and Promise). Obviously Thunderbolt needs to mature but it is known only for a couple of weeks. New standards need quite some time (like DisplayPort, USB3).

To be fair I actually don't expect Intel/Apple/anybody else will bring out expansion cards. Thunderbolt seems to be aimed at notebooks so people now have a real mobile workstation (especially since the 15" MBP configuration can match the cheapest Mac Pro configuration). Using two kinds of Thunderbolt ports (data only & full) only makes things more confusing for the average user. But it still doesn't mean that this isn't possible.

 

[nanofrog]

so basically, at the moment it doesn't look possible for this with TB - but maybe with future LP releases/expansions (next year?) that it may be possible?

It's possible to use the current TB chip in a PCIe card. The problem is, the resulting confusion that would occur (i.e. DATA only support via a TB card that can't be attached to the GPU vs. laptops and potentially AIO's that have full support = DATA + MDP signals over the TB cable).

Such confusion would likely cause anger and have a negative result on adoption. Intel and their partners are banking on this launch being successful, so they're going to take steps to reduce, if not outright eliminate such confusion.

Now the fastest and cheapest way to get a PCIe based TB card (faster because there's less work this way), is to work with graphics card vendors and create a connection standard to get DisplayPort data to the PCIe card (i.e. via a flexible PCB like that used for Crossfire connectors). Unfortunately, this does leave a "hole" in terms of confusion in some cases, as they'd need both the TB card and a newer graphics card that can be attached to the TB card for DP data for full support over the TB interconnect (confusing already isn't it ).

The one way to eliminate even this, is to get graphics card makers to add the TB chip to their cards. But this has drawbacks for graphics card vendors, as it adds cost and complexity to an already difficult task (i.e. PCB real estate as well as PCI bracket real estate, which has to be balanced between any interconnects and vents for sufficient airflow). But it would solve the DATA only vs. full TB support issue for Intel and their development partners (they undertook all of this for a pay-day afterall...), so they may try to take this route if they can get graphics card vendors to go for it.

I do see a need for some users to have TB in a desktop (and really, they'd be after DATA support more than video), due to cost effectiveness of shared TB peripherals with their laptops.

Think of an independent film maker/video editor. Let's say that person goes out on location regularly, and shoots footage for the project they're working on. If it's stored directly to an external TB storage system, such as a Promise R4 or R6 box, they can connect the TB storage directly to their workstation (desktop). This can allow them to go directly to editing, as there's no need to transfer data from the laptop (or storage system on another interface, such as eSATA) to the workstation first, which improves their workflow. This saves them time, which has the potential of allowing them to attain more work and improve their bottom line.

To be fair I actually don't expect Intel/Apple/anybody else will bring out expansion cards.

This will be up to Intel, as they've control over who gets the parts or not (done legally through some sort of signed agreement) as a means of preventing user confusion (DATA only vs. full support mess that could develop from PCIe based TB cards that don't have a DP signal attached for whatever reason).

Thunderbolt seems to be aimed at notebooks so people now have a real mobile workstation (especially since the 15" MBP configuration can match the cheapest Mac Pro configuration). Using two kinds of Thunderbolt ports (data only & full) only makes things more confusing for the average user. But it still doesn't mean that this isn't possible.

It's most definitely aimed at laptops and AIO's for now. But there are reasons that a PCIe card would be a good thing for some users (see above example).

But with the potential for confusion, it's hard to say where Intel will draw the line (i.e. make GPU card vendors add a TB chip, or if they'd be happy with a simple connector to get the DP signal to the TB card for full support for those that want it).

 

[Joshuarocks]

My prediction

3rd Party companies aren't stupid.. I am sure they will come out with a PCIe drop in for the mac pro to support thunderbolt.. the 6-core is an extremely powerful mac pro.. Sandy bridge really shines better in notebooks as its lower power consumption and overall cooling make the macbook pro not so much of a heater.

Better power consumption is what Sandy Bridge is all about. I don't believe Sandy Bridge is going to make a big deal in the mac pro, given what Westmere still has to offer.. 6-core 3.33 is no slouch.

 

[300D]

That's not how it works.

Incorrect. You have not provided any documentation that states differently from Intel's official information release.

You claim it is false information so you're the one that should backup his claims.

Incorrect. I have not posted any information that differs from Intel's information, has you clearly have.

Anyway, if you looked at Apple's and Intel's pages about Thunderbolt you'd have the documentation you want.

I suggest you actually read it since you don't have the first clue what it says.

That is incorrect information and a complete misunderstanding about Light Peak and Thunderbolt.

That is flase information. I suggest you research how LightPeak and Thunderbolt differ to prevent you from making a fool of yourself any further.

Light Peak was the name of the project for a new kind of "universal" protocol which eventually was released as a finished product named Thunderbolt.

They share nothing except the developer.

Thunderbolt/Light Peak is based on both the PCIe and mDP formats.

Incorrect. Thunderbolt alone is based from PCIe. Light Peak used its own protocol.

My statements show that I indeed do not have any clue as to what Thunderbolt is.

That is correct.

The way you view it, Thunderbolt would be hardware tied to the PCIe controller AND the GPU. That combination makes any kind of expansion card impossible.

incorrect. TB can be included on PCI-E video cards because it will have direct connection to PCI-e data lanes and the graphics processor.

Thunderbolt is not something that requires both PCIe and GPU which is why in theory you can have something like an expansion card

Incorrect, per intel's press release.

It is not since Intel is promoting it actively and denying Apple has any exclusive use

Incorrect. They have a 12 month "lead time".

Canon is the fourth manufacturer to use the standard (the others are Apple, Lacie and Promise).

Apple is the only one with chipset access. Those others are accessory manufacturers, not computer manufacturers.

Thunderbolt seems to be aimed at notebooks

Incorrect. The MBP is simply the first product with it. If it had been released on the MP, you'd probably say "Thunderbolt seems to be aimed at desktops" instead.

 

[nanofrog]

Incorrect. You have not provided any documentation that states differently from Intel's official information release.

The difference it seems, is how that information is being interpreted (not as clear as it could be).

That is false information. I suggest you research how LightPeak and Thunderbolt differ to prevent you from making a fool of yourself any further.

Since you think they're two totally different technologies, why not post the sources (particularly on LightPeak) that you're getting this impression.

Incorrect. Thunderbolt alone is based from PCIe. Light Peak used its own protocol.

Think about this logically for a second. PCIe on it's own cannot handle daisy chaining of devices (must use a switch to have multiple end-points, and each of those end-points has it's own wiring to the switch - not over the same wiring). So they had to make some modifications, which will include the protocol (what the TB chip actually does).

In light of this, the differences aren't as stark as you may think. In fact, the only difference, is that they didn't include the optical transceivers.

Now if I had to guess, the optical cables didn't work out to be a cheap as they initially expected and/or were going to be delayed (which haven't arrived yet). Personally, I expect both issues occurred. So they made the change to get a product shipped (they want their R&D money back + a lot more ). It's also cheaper for system vendors this way, and in the case of laptops, makes it a bit easier, as there's less complexity and more importantly, less PCB real estate consumption as a result.

TB can be included on PCI-E video cards because it will have direct connection to PCI-e data lanes and the graphics processor.

This is one possibility. But not the only way it could be done as mentioned previously.

Now whether when/if Intel will go for this isn't clear, but it's in their best interest IMO that they do settle on a method due to users' desire to share peripherals with more than just laptops (or AIO's) that come with TB ports.

They have a 12 month "lead time".

A lead time isn't the same as exclusivity (they're willing to sell to other vendors, but so far, none of the others have a product ready yet).

Apple accomplished this by working with Intel and the other developers from the beginning on the software end. What's happening now, is the pay-off. It's allowed them to beat the competition to market by getting their hardware (MBP with a TB chip) and software done earlier than anyone else.

Incorrect. The MBP is simply the first product with it. If it had been released on the MP, you'd probably say "Thunderbolt seems to be aimed at desktops" instead.

Maybe. TB doesn't really offer much to the desktop market alone though, and wouldn't generate as much interest for adoption.

Users with desktops equipped with PCIe lanes can do better with existing solutions (separate GPU card + RAID storage systems can out-perform what TB can do; even with a Promise R6).

For laptops OTOH, it offers a significant advantage than what's previously been available for most (faster connection than they've ever had before, and potentially fewer wires to boot).

 

[CaoCao]

If Intel can make a MDP1.2 Thunderbolt update you would probably see it appearing on lots of video cards

 

[seek3r]

They share nothing except the developer.

Wait, what? ::sighs:: lightpeak was a codename for the project, thunderbolt was the resulting consumer product released - rather like Apple's internal "Chefcat" --> marketing name "Firewire"

 

[DoFoT9]

It's possible to use the current TB chip in a PCIe card. The problem is, the resulting confusion that would occur (i.e. DATA only support via a TB card that can't be attached to the GPU vs. laptops and potentially AIO's that have full support = DATA + MDP signals over the TB cable).

Such confusion would likely cause anger and have a negative result on adoption. Intel and their partners are banking on this launch being successful, so they're going to take steps to reduce, if not outright eliminate such confusion.

oh i understand. i think for the mean time that a data only solution is the best method - i honestly cant see much audience/attention for GPU solutions in the consumer market, can you?

edit: actually, reading further on - im not sure im following the actual implementation of TB + GPU. i always assumed TB could be used to attach external GPUs onto a machine...?

(confusing already isn't it ).

yes!

The one way to eliminate even this, is to get graphics card makers to add the TB chip to their cards. But this has drawbacks for graphics card vendors, as it adds cost and complexity to an already difficult task (i.e. PCB real estate as well as PCI bracket real estate, which has to be balanced between any interconnects and vents for sufficient airflow). But it would solve the DATA only vs. full TB support issue for Intel and their development partners (they undertook all of this for a pay-day afterall...), so they may try to take this route if they can get graphics card vendors to go for it.

very confusing, but that would solve it. if HDMI/VGA/DVI connectors were abandoned for a singular TB port, would that potentially make enough room to be attractive for the graphics card vendors?

I do see a need for some users to have TB in a desktop (and really, they'd be after DATA support more than video), due to cost effectiveness of shared TB peripherals with their laptops.

compared to eSATA and USB3.0 i can't see anything more attractive, as the bus isnt the bottleneck. i guess thats just me though, as i am generally fine with FW800 still

Think of an independent film maker/video editor. Let's say that person goes out on location regularly, and shoots footage for the project they're working on. If it's stored directly to an external TB storage system, such as a Promise R4 or R6 box, they can connect the TB storage directly to their workstation (desktop). This can allow them to go directly to editing, as there's no need to transfer data from the laptop (or storage system on another interface, such as eSATA) to the workstation first, which improves their workflow. This saves them time, which has the potential of allowing them to attain more work and improve their bottom line.

ok, yes - for that scenario i totally agree. you wouldnt exactly call that a consumer target base though, not even prosumer. that is more professional market.

so is TB attempting to target the professional, or consumer market more?