Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
So glad you said that. I think quite a few people are going to see this and not realise that the harddrive is going to be a massive bottleneck. Especially with that guy in the video saying "700 Megabytes per second!" when actually he means Megabits per second.


A PCIe x4 slot has a 32Gb/s bandwidth. Well, when boards with v3.0 come out anyway.

Does this mean people with an extra x16 slot could potentially add a TB card? what about having an x16 video card with built in TB?
 
No. The old cable + separate transceiver was more expensive than the combo cable + integrated transceiver , but the combo was not less than the cable all by itself. The point is that optical cable is going to be more expensive than copper ( although copper prices are increasing). The optical cable is going to be more expensive anyway. Having to throw electronics into both ends of the cable is going to drive up costs.
Yes,
optical will always be more expensive. But just like with spdif, at some point difference is too small to count. Already good long copper hdmi can be almost as expensive than just little longer cheap optical extender.

My point was that if you use optical, you need to have transiever in both ends and it is cheaper to have them integrated in the cable.

Who wants slower/most expensive video ? Looks at the path. You need some display port to pci-e converter on the graphics card and then another pci-e to display port converter to pump it back into the display channel on the TB controller. Not only that but now you are sucking up PCI-e bandwidth pumping data for no good reason.
Converters are not economical issue when used in massive scale. If 100M graphic cards per year would have this, it woulc cost something like 1 cent per gpu.
Also if you already have fast PCIe architecture on your mb, it doesn't cost anything to use it.
Another economical solution would be TB-pcie-card without dp path.
Most people will need TB for something else than display. TB was integrated in miniDP port just because it was most convenient place in mbp, when Apple didn't want add new port or take any existing away.


The other problem is that the graphics engine on the card would have to share bandwidth to memory/CPU with the non video data being pumped by TB. Right now graphics cards use x8 or x16 PCI-e lanes themselves without adding bandwidth to the external RAID array to the mix. PCI-e v3.0 will help. Folks have wondered a bit about why a graphics card would need 16 v3.0 lanes when 8 would being equiv to 16 v2.0 lanes. So maybe when PCI-e v3.0 becomes mainstream this will be less an issue.
[although kind of wonder if these TB controllers can deal with PCI-e v3.0. ]
I would wonder why they wouldn't make TB card PCIe v3 compliant.
Although they are still making all kind of stuff with usb2, when usb3 would cost something like 1 buck more and macs are still not having bd...

But at technical point there's no problem for even 8 lane v2 pcie (32 Gbit/s) to handle all possible TB card traffic with all possible dp traffic.

Finally, the last reason is non-techinical. It will be confusing to many users. One of the marketing points of TB is that you can hook a display into it. So if someone did a "PCI-e only" card ( only hook the card to the PCI-e bus like what is done in the storage peripherals ) then when someone plugs their dispaly into the card and no video comes out ..... that is yet another support communication about why it is suppose to but doesn't. Additionaly, the more folks do that the more likely display vendors won't pick up TB. The argument will be that more host TB ports don't have video data coming out anyway, so why should I add this expensive controller to my cost senstive display.
People who buys MPs don't get confused so easily. And so few buys them, that they don't mean anything to display manufacturers.





1 display because there are zero , including Apple, displays with TB sockets. They are display port only. So has to be last device on the daisy change. When there are dual ported TB displays ( which will probably cost more ) then probably and drop more than one onto the chain.
Not sure it works that way. Seems likely there is a mode the controller goes into to recognize whether the device on the other side is speaking DP so just communicates purely in DP or whether all the comm is done in TB mode when TB device on the other side. There is just no mix in the traffic if only a DP decoding device on the other side.
I didn't get this.
TB specifications are limited to 1 display at the end of chain, because there are no TB displays?
And when first one pops out, they change the specs?

Have I undersood correctly: TB has only one lane. Dp can have 4 lanes. TB conncetor can only carry one TB lane, which can include one mixed dp lane. If there's more than one dp lane to carry TB can't carry that?
 
...( TB soaks up enough PCI-e bandwidth will need faster PCI-e lanes so that GPU and TB bandwidth don't step on each other. )
No.
TB can carry 10 Gbit/s.
PCIe v2 x4 can carry 16 Gbit/s.
If dp carried by TB is limited to 1 lane, it can't have more than 5.184 Gbit/s and therefore 16 Gbit/s is enough for pipe from both gpu & cpu.

What puzzles me, is what kind of hubs are we getting, when and how expensive they will be.
I was looking to buy Dell's U2711.
Now, I'm not sure if I should wait for TB version of it.
Maybe it couls have integrated TB hub with usb3 & eSATAp?
Or if a TB hub with dp, usb3 & eSATAp comes soon with small pricetag, I could safely buy the current version of U2711.

I hate these "technology unveilings" that leave more questions than answers...

Another thing wich leaves me wondering, is that is TB a really good substitute for not having integrated usb3 & eSATAp?
For example, if I want fastest possible connection to external hdd, how do I benefit from TB?
Changing usb2 connectors to usb3 would have costed Apple only few bucks.
Usb3 hdd's have same price than usb2 hdd's. Usb3 cables can have same length than TB cables.
eSATAp cable would be limited to 2 meters, but you'd get SMART data and bootability.
My guess is that TB will double the price for this single external hdd.

It is not a good excuse to have one expensive fast connection to leave other connections old & slow or missing.
 
Last edited:
You've got to love Intel. Up to their old tricks again.

Touting something so dramatic yet when it comes to the launch it's only at 10% of the speeds claimed originally.

10 Gigabits/seconds is still stunning and more than welcome.
 
This looks very much like a PCI-Express "Thunderbolt" add-on card.

Remember that Thunderbolt merely acts as an external extension of a 4x PCI-Express slot.

From the big, red, grab-your-attention section of that page (just under the photo):
As the final Light Peak standard has not been released by Intel, these products are still under development and are not yet available.

It looks like Thunderbolt is LightPeak+DisplayPort, so unless your current MB has its data channels set up like that there might *not* be a way to add Thunderbolt to an existing machine. (I actually hope I'm wrong about that, even though it wouldn't impact me one way or the other since I only have laptops these days.)
 
...Intel lists the dual core processors used in the 13 inch model at a TDP of 35 watts. That is the same as all CPU's used in the previous generation MBP's...
That's not entirely true. The Core 2 Duo that was used in the previous 13" MacBook Pro had a TDP of 25W. However, since that design also required a NVIDIA system controller/GPU (integrated) the sum of both chip's TDP was still around 35W. Of course, Sandy Bridge allowed Apple to drop the NVIDIA chip so the net change is near zero.
 
Here's something that I haven't yet seen discussed on MacRumors. The Thunderbolt port on the MacBooks contains TWO channels and each of those can carry up to 10Gps in both directions -- it's two channel and bi-directional. Thus, the total bandwidth is 20Gbps both up and down and that means you've got 40Gbps combined in/out.

However, I found one article which claimed that one of these channels is dedicated to DisplayPort (which kind of makes sense). Thus, you have 10Gbps in/out for the PCIe portion of Thunderbolt and a separate 10Gbps output dedicated to DisplayPort (I don't know what happens to the 10Gbps of input bandwidth which is on the DisplayPort channel -- I guess that will be underutilized for backchannel traffic on DisplayPort).

This is also documented in AnandTech's Thunderbold review and is even briefly mentioned in Apple's Thunderbold overview (from which I quote below):
Thunderbolt I/O technology gives you two channels on the same connector with 10 Gbps of throughput in both directions.
Further, AnandTech says the following:
Thunderbolt is dual-channel, with each channel supporting 10 Gbps of bidirectional bandwidth. That’s a potential 20 Gbps of upstream and 20 Gbps of downstream bandwidth.
 
Last edited:
To everyone upset about no PCIe implementation, I feel your pain. Hopefully Apple and Intel will reconsider this if they anticipate this to become a new standard to compete with USB and eSATA. Otherwise, it will be FireWire all over again. This is exactly why I built a Hackintosh. Don't get me wrong, I love love love apple and the hardware that they make, but its this kind of bulls**t that set me off and forced me to walk away. Right now the machine I'm running is on par (and possibly faster) that $2500-$3000 Mac Pro, and I built it myself with no prior experience for a whopping $900!

Oh, and the guy in the video messed up. Unless the drive bank that he was using was full of SSDs from the future, there is no way that it topped out at 700MB/s. The fastest write speed you'll see on a SSD today is around 250-300MB/s. In this test the actual MB/s speed that was displayed was around 85.
 
Last edited:
You've got to love Intel. Up to their old tricks again.

Touting something so dramatic yet when it comes to the launch it's only at 10% of the speeds claimed originally.

10 Gigabits/seconds is still stunning and more than welcome.
10 gigabits/s = 1.25 gigabytes/s

Yes, the Intel page shows Gb/s (10Gbps). The guy in the video used MB/s (800MB/s). Guess what. Those are in approximately the same camp as far as data rates go. (The file transfer was 800MB/s, which is roughly 6400Mb/s. Add in the display signal and there you go.)
The hard drive was the bottleneck on that demo.
 
Yes, the Intel page shows Gb/s (10Gbps). The guy in the video used MB/s (800MB/s). Guess what. Those are in approximately the same camp as far as data rates go. (The file transfer was 800MB/s, which is roughly 6400Mb/s. Add in the display signal and there you go.)
Except, that the display apparently runs on a separate 10Gbps channel (the Thunderbolt port on the MacBooks has two channels).
Braindead360 said:
Oh, and the guy in the video messed up. Unless the drive bank that he was using was full of SSDs from the future, there is no way that it topped out at 700MB/s. The fastest write speed you'll see on a SSD today is around 250-300MB/s. In this test the actual MB/s speed that was displayed was around 85.
Where did you get 85? Also, I think they were using the then unreleased OCZ Vertex 3 Pro SSD (or similar) which apparently offers speeds well over 400MB/s (max rates are quoted as Read: 550 MB/s, Write: 525 MB/s).
 
Last edited:
No add-on cards, eh? (even at reduced rates) :rolleyes:

It's just one more reason Thunderbolt will not be adopted en-masse any time soon. You cannot expect hard drive makers or anyone else, really to bother to add ports for something that almost nothing on the planet supports. By the time this picks up, USB3 will firmly entrenched (you CAN add it to most PCs, including some Macs; even my 2008-era MBP can add USB3 via the expansion slot that newer 15" Macs don't bother to support). I've already got one USB 3 drive here already (3TB media drive). Thunderbolt won't do it a big of good.

They should have made Thunderbolt accommodate USB3 within its bus. That would/could have saved space in the MBP motherboard and it would have then offered USB3 + the newer Thunderbolt port on the same machine and cost less than having to offer them separate (or in the MBP's case, not at all).
 
Here's something that I haven't yet seen discussed on MacRumors. The Thunderbolt port on the MacBooks contains TWO channels and each of those can carry up to 10Gps in both directions -- it's two channel and bi-directional. Thus, the total bandwidth is 20Gbps both up and down and that means you've got 40Gbps combined in/out.
Next we can start creating stories about why Intel represents these basic technical details so confusingly.
And why the real specs are still secret under some NDA of development kit?
Are they still largerly under construction?
What is the competition they are afraid of?
 
Next we can start creating stories about why Intel represents these basic technical details so confusingly...
I think they are actually being unusually honest. They say 10Gbps of high-speed data but that doesn't include the DisplayPort traffic or overhead. They could have claimed a total bandwidth of 20Gbps each way but that would have been a little misleading.
 
it's probably too early, but does anyone out there know if you can boot into windows via bootcamp via thunderbolt?

i've checked apple support, etc - but it's prolly too early in the game to know.
 
I think they are actually being unusually honest. They say 10Gbps of high-speed data but that doesn't include the DisplayPort traffic or overhead. They could have claimed a total bandwidth of 20Gbps each way but that would have been a little misleading.
So is 10 Gbps Displayport only and 10 Gbps PCIe only or can both of them use 20 Gbps? I'm confused.
 
time machine backup's are mostly limited due to two facts.
One: hardrive have a maximum sequential read/write cap (wouldn't be much more in the than 80 MB/s in the 5400 rpm 750 GB HDD).
Two: backing up is most of the time NOT SEQUENTIAL, so in real life you'll average at around 5-15 MB/s.
The time capsule itself even seems to cap at only 15 MB/s (read/write, not throughput/network connectivity).

TB/LP = 10Gb/s raw throughput ~ 900MB/s usable throughput
USB 3.0 = 5 Gb/s raw throughput ~ 150-200 MB/s usable throughput
GBase ethernet = 1 Gb/s raw throughput ~ 90-110 MB/s usable
Firewire 800 = 800 Mb/s raw throughput ~ 80-100 MB/s usable

Backing up will not go any faster using TB in comparison to quite a lot of more frequently used connections.
The power of TB is the extendability of your computer with PCI periferals, without the common high latencies, the scalability, the usage of external storage as a internal. Except for the multiple display you can now easily connect to your mac/pc using just one port I don't expect to see soon any consumer products that'll use TB as a necessity.

So glad you said that. I think quite a few people are going to see this and not realise that the harddrive is going to be a massive bottleneck. Especially with that guy in the video saying "700 Megabytes per second!" when actually he means Megabits per second.

Sjhonny is wrong about USB 3.0 throughput. It is 400 MB/s (or more depending on overhead) usable throughput not 150-200 MB/s. The fastest SSD's on the market are sub-300MB/s write/read so the hard drives ARE the bottleneck. These are INTERNAL SSD drives which aren't connected by USB. There is almost no market in external SSDs so you will almost never see them connected by USB 3.
 
Last edited:
I am going to have to wait and see. The times have always proven to me that there's never really a limit in technology, just a lack of interest or desire or market need from the parties involved if you know what I mean. Since TB is using the PCIe x4 lane, and we already have plenty of bandwidth options as fast or faster than 10GBps in ethernet, FOE, and Fibre channel, there may not be much of a limitation on creating a TB card . . . . even if it only got us half of the proposed bandwidth.





True, but FCP is pulling from the source files on the Promise to create the composite image, and that image can be played back in full uncompressed 1080p. The engadget guy might not be smart enough to set the app for those settings, but it is stellar speed.

Not to mention, that it's pulling the four streams from the Promise, encoding in realtime on the MBP, spitting it back out on the canvas for FCP and the 27" ACD without any dropped frames.

TB is a nice step up from FW800 and in lots of ways eSATA since the bottleneck in many HD edit rigs was the I/O.



There is such a thing as a G-Technology TB drive? Link please :cool:

On the side, the sucky part about TB for current workstation owners is that they've spent $5000 for a decently equipped Mac Pro and are going to feel that lack of TB hurt down the line. Any company with the budget and the mind power WILL make a TB PCIe card for them, and most likely will charge whatever outrageous price they can.

If I were in that camp, and spend the paltry $5000 for a Mac Pro, I would definitely put another $800 or more in for a 2-6 port TB PCIe card.

Sorry to say but I think you are missing some key elements. If the footage is native ProRes then the quad core processor (with hyperthreading so it looks like 8) is more than enough to spit out the footage (in this case the processor would be the bottleneck. Not stellar - just cool. If the footage is uncompressed (which I read somewhere) then what you are calling "encoding" is beyond minimal - all the work is on the HDD. Indeed still impressive that one raid can spit out 4 streams of uncompressed but please let us not exaggerate that the bus is actually handling 8 streams
 
So is 10 Gbps Displayport only and 10 Gbps PCIe only or can both of them use 20 Gbps? I'm confused.
I think the current implementation is limited to 10Gbps on the PCIe/data channel and the other channel is dedicated for DisplayPort.

Actually, it seems that the port has at least two modes. In one mode the port acts just like the original DisplayPort, that's how the port continues to work with the existing DVI/VGA adapters and the current Apple Cinema Display. This is called the "compatibility" mode (based upon what they said in the Thunderbolt demo video that was posted on Engadget, reference the video at time mark 00:25). Then, if you actually plug in a Thunderbolt device to the MacBook it switches to a single 10Gbps channel for PCIe/data and the second channel is then apparently reserved for carrying DisplayPort.

Thus, in the example video where they show transfers from an external RAID connected via Thunderbolt the RAID gets the full 10Gbps bandwidth and the Thunderbolt port on the RAID that connects to the display is run in the compatibility mode where the output from the second channel is converted to legacy DisplayPort for the Cinema display. I suspect if you attached two Thunderbolt devices in a daisychain they would share the same 10Gbps bandwidth/channel but you could still attach the existing Cinema Display to the end of the chain and the display will get a full 10Gbps over the second channel.

The only question remaining might be what if you don't attach a display? Can the second channel supply data to a PCIe device that is downstream from the first Thunderbolt device? I suspect that is the way it is supposed to work (by design), but I wonder if that is not enabled in this first release.
 
Last edited:
No.
TB can carry 10 Gbit/s.
PCIe v2 x4 can carry 16 Gbit/s.
If dp carried by TB is limited to 1 lane, it can't have more than 5.184 Gbit/s and therefore 16 Gbit/s is enough for pipe from both gpu & cpu.

I thought that but isn't TB 10Gbit/s both ways? So the overall bandwidth would be 20Gbit/s? Realistically though most people wouldn't notice the 4Gbit/s loss since they wouldn't saturate connection anyway.
 
They should have made Thunderbolt accommodate USB3 within its bus. That would/could have saved space in the MBP motherboard and it would have then offered USB3 + the newer Thunderbolt port on the same machine and cost less than having to offer them separate (or in the MBP's case, not at all).

There's two theings that could make thunderbolt a winner over USB 3 in the long run.One key question over this is whether Apple intend to license the Apple Mini DisplayPort/TB port to other monitor manufacturers. The other thing is how much it's going to cost hardware makers to license thunderbolt for the peripherals they make.

It was impossible to incorporate USB3 since the USB-IF rejected modifications of the standard to allow for Thunderbolt.
 
There's two theings that could make thunderbolt a winner over USB 3 in the long run.One key question over this is whether Apple intend to license the Apple Mini DisplayPort/TB port to other monitor manufacturers. The other thing is how much it's going to cost hardware makers to license thunderbolt for the peripherals they make...
There is no cost or license for Mini DisplayPort, it's already part of the DisplayPort standard (i.e. Apple developed the form factor, but "gave" it to the DisplayPort standards body). I'm not 100% sure about Thunderbolt, but I think there's no license or royalty required for that either.
 
Thunderbird is not Apple's to license to other vendors. Lightpeak, a.k.a Thunderbird, is owned by Intel.
 
Except, that the display apparently runs on a separate 10Gbps channel (the Thunderbolt port on the MacBooks has two channels).

Where did you get 85? Also, I think they were using the then unreleased OCZ Vertex 3 Pro SSD (or similar) which apparently offers speeds well over 400MB/s (max rates are quoted as Read: 550 MB/s, Write: 525 MB/s).
He was using Pegasus R6 in a RAID setup which allows data to be written to multiple drives at once. That is how he is able to achieve 800 MB/s. This setup is overkill and probably crazy expensive for most MBP users.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.