Why Thunderbolt is dead in the water, aka how Apple killed Light Peak(extremetech.com)
extremetech.com
Why Thunderbolt is dead in the water, aka how Apple killed Light Peak
http://www.extremetech.com/article2/0,2845,2385615,00.asp
9 comments
> I'd be really surprised if you could drive, say, a very high resolution monitor off USB 3.0.
You definitely can't. But then again, you don't do that either with Thunderbolt. You do that with the DisplayPort signal that goes into the same cable.
Technically, you could implement a pretty crummy external video card driven by thunderbolt (may be useful for laptops, as additional oomph), but I don't think you'd usually do that.
But if you want to go into technicalities, USB3's SuperSpeed raw throughput is 4Gbps (5Gbps theoretical max), 3.2Gbps expected after protocol overhead. That's 400MB/s (full-duplex) (note that this 64% effective throughput is an impressive progress over USB2, which rarely reaches 50% of its theoretical 480Mbps throughput).
Thunderbolt is specced at 2x10Gbps (two channels per cable), full-duplex. And because it has a much lighter protocol (close to PCI-E) it should be able to reach a much higher efficiency than USB3.
It also has a significantly lower latency, even at the end of a daisy chain, than USB.
You definitely can't. But then again, you don't do that either with Thunderbolt. You do that with the DisplayPort signal that goes into the same cable.
Technically, you could implement a pretty crummy external video card driven by thunderbolt (may be useful for laptops, as additional oomph), but I don't think you'd usually do that.
But if you want to go into technicalities, USB3's SuperSpeed raw throughput is 4Gbps (5Gbps theoretical max), 3.2Gbps expected after protocol overhead. That's 400MB/s (full-duplex) (note that this 64% effective throughput is an impressive progress over USB2, which rarely reaches 50% of its theoretical 480Mbps throughput).
Thunderbolt is specced at 2x10Gbps (two channels per cable), full-duplex. And because it has a much lighter protocol (close to PCI-E) it should be able to reach a much higher efficiency than USB3.
It also has a significantly lower latency, even at the end of a daisy chain, than USB.
...you don't do that either with Thunderbolt. You do that with the DisplayPort signal that goes into the same cable.
No and yes, respectively... sending DisplayPort over the wire is a feature of Thunderbolt, not something you 'do with the same cable'. You can also theoretically drive a monitor with PCI Express, the other half of the Thunderbolt bus.
No and yes, respectively... sending DisplayPort over the wire is a feature of Thunderbolt, not something you 'do with the same cable'. You can also theoretically drive a monitor with PCI Express, the other half of the Thunderbolt bus.
I am sure that there are many technologies that are inherently better than USB -- but it's not always the better technology that wins. (VHS vs. Betmax...)
Don't generalize, by the way - if details are false or irrelevant, please mention them so that they can be rectified.
Don't generalize, by the way - if details are false or irrelevant, please mention them so that they can be rectified.
Not to mention that the tech has been out for about 5 minutes. Proclaiming the death of an interconnect that is this new is in and of itself pretty ridiculous.
I run a 24" monitor running at 1920x1200 off of a USB 2 displaylink adapter. Of course video or games slow my computer if they're running on my secondary monitor. But it's fine for a web browser or a text editor.
I posted this on the site but I'll post it here, too.
The last point is wrong. The USB Implementers Forum complained last year about lightpeak using the usb interface. The other points have some merit but, considering sony is supposedly putting thunderbolt into their new laptops, it's still too early to call the interface a disaster. Or to blame Apple about it bombing.
http://www.engadget.com/2011/05/17/sonys-thunderbolt-impleme...
The last point is wrong. The USB Implementers Forum complained last year about lightpeak using the usb interface. The other points have some merit but, considering sony is supposedly putting thunderbolt into their new laptops, it's still too early to call the interface a disaster. Or to blame Apple about it bombing.
http://www.engadget.com/2011/05/17/sonys-thunderbolt-impleme...
Don't forget, Sony was the other large proponent of FireWire. They have a big vested interest with their video gear, I guess.
Could the USB-IF actually block a fiber-and-wire USB-socket-compatible implementation?
Could the USB-IF actually block a fiber-and-wire USB-socket-compatible implementation?
Even in its non-optical, crippled copper state, Thunderbolt is prohibitively expensive. USB 3.0 controllers cost just a few dollars, while Thunderbolt hardware, we've been told, cost no less than $90.
This is the key claim in the article, and "we've been told" is not exactly authoritative.
This is the key claim in the article, and "we've been told" is not exactly authoritative.
That figure is from an OEM. I was just shying away from the (magnificently over-used) phrase 'according to unnamed sources familiar with the situation...'
Also, look at the Matrox pricing! (Or the prices of any other Thunderbolt-enabled devices that are coming to market -- if you can find them; there are not many yet...)
Also, look at the Matrox pricing! (Or the prices of any other Thunderbolt-enabled devices that are coming to market -- if you can find them; there are not many yet...)
Wait, you mean early products implementing a barely released smart protocol are going to be more expensive than products implementing an incremental update to an existing dumb protocol? Surely that can't be.
What's your point? Nobody's expressing mystification at the price difference. That's not the topic at hand.
Fair enough, still seems really high considering you can get entire motherboards for around that price.
Also, look at the Matrox pricing!
Sure they'll soak the early adopters, but that's not necessarily representative of the marginal cost.
Also, look at the Matrox pricing!
Sure they'll soak the early adopters, but that's not necessarily representative of the marginal cost.
No, true enough.
It would be cool if Intel spoke out about the actual cost of Light Peak -- I mean, _can_ it be as cheap as USB 3.0, or is it inherently an expensive technology?
FireWire controllers, for example, are still many times more expensive than the USB equivalent.
It would be cool if Intel spoke out about the actual cost of Light Peak -- I mean, _can_ it be as cheap as USB 3.0, or is it inherently an expensive technology?
FireWire controllers, for example, are still many times more expensive than the USB equivalent.
> It would be cool if Intel spoke out about the actual cost of Light Peak -- I mean, _can_ it be as cheap as USB 3.0, or is it inherently an expensive technology?
It's going to depend what you do with it. Thunderbolt is basically an external PCI-E, with some additional brains to handle chaining and the DisplayPort signal. I'd say it's always going to be more expensive than USB3.
Not necessarily by much, but with USB the vast majority of the work is really done in software and on the machine's hardware, which pretty directly controls the USB device. That's, in fact, why FireWire was so much superior to USB early on and could reach much higher data rates without straining the machines: the dedicated hardware meant almost no resources (outside of providing or consuming data) were required of the machine (I remember machines in the late 90s and early 00s having trouble playing MP3s in Winamp and driving a USB communication at the same time)
It's going to depend what you do with it. Thunderbolt is basically an external PCI-E, with some additional brains to handle chaining and the DisplayPort signal. I'd say it's always going to be more expensive than USB3.
Not necessarily by much, but with USB the vast majority of the work is really done in software and on the machine's hardware, which pretty directly controls the USB device. That's, in fact, why FireWire was so much superior to USB early on and could reach much higher data rates without straining the machines: the dedicated hardware meant almost no resources (outside of providing or consuming data) were required of the machine (I remember machines in the late 90s and early 00s having trouble playing MP3s in Winamp and driving a USB communication at the same time)
> USB 3.0 is more than fast enough for current and near-future applications.
USB3.0 Bandwidth: 3.2 GBit/s
HDMI Bandwidth: 10 GBit/s
I am still waiting for a truly universal bus.
USB3.0 Bandwidth: 3.2 GBit/s
HDMI Bandwidth: 10 GBit/s
I am still waiting for a truly universal bus.
To quote the article "Just like FireWire, though, Thunderbolt is off to a slow start." So is firewire dead, now, after a long wait for a second, third, tenth, twentieth company to use it? No. So wait for Thunderbolt to be used or not, whichever it may be within the next ten years.
Article writer is article submitter.
Article writer is article submitter.
If Intel isn't bundling it into every motherboard they ship I don't see how anyone could blame Apple for killing it.
I suspect Intel is a bit torn between USB 3.0 and Light Peak integration. Light Peak would be very expensive to integrate into every motherboard they ship -- but yes, Light Peak definitely has a stronger chance of taking off because of the Intel link.
Don't forget, Intel is due to integrate USB 3.0 at some point, too...
Don't forget, Intel is due to integrate USB 3.0 at some point, too...
What was the pricepoint of initial USB hardware interfaces?
The original USB interfaces which apple made mainstream in the iMac? I don't know. But I do know that MacBook Pro prices with Thunderbolt are similar to without, probably because the thunderbolt controller also does DisplayPort (I wouldn't be surprised if Apple built the connector themselves, actually. Maybe ifixit knows.)
The market is different than when they put USB into the iMac now mainly because there are large amounts of cheap USB peripherals which USB 3 is backwards compatible with. I think (but haven't done any research) that the number of companies making USB peripherals now is much (one or two orders of magnitude) greater than the number of companies making PS/2 or parallel SCSI when the iMac was released. The personal computer market has grown a lot, and USB now has much more momentum than anything else had when the iMac was released. People (such as my mother) are going to buy a new computer and are going to want it to work with their existing external hard disk -- they're not going to transplant the disk to a new case. So USB, especially is USB 3 peripherals can run at slower speeds over USB 2, is going to stay for a little while. Thunderbolt is going up against an entrenched competitor, and if you compare to the HD format wars, both players worked with DVDs. If Blu-Ray didn't and people had to have two boxes instead of one, HD-DVD probably would have won. The best Thunderbolt is probably going to do for the near future is be the new FireWire.
The market is different than when they put USB into the iMac now mainly because there are large amounts of cheap USB peripherals which USB 3 is backwards compatible with. I think (but haven't done any research) that the number of companies making USB peripherals now is much (one or two orders of magnitude) greater than the number of companies making PS/2 or parallel SCSI when the iMac was released. The personal computer market has grown a lot, and USB now has much more momentum than anything else had when the iMac was released. People (such as my mother) are going to buy a new computer and are going to want it to work with their existing external hard disk -- they're not going to transplant the disk to a new case. So USB, especially is USB 3 peripherals can run at slower speeds over USB 2, is going to stay for a little while. Thunderbolt is going up against an entrenched competitor, and if you compare to the HD format wars, both players worked with DVDs. If Blu-Ray didn't and people had to have two boxes instead of one, HD-DVD probably would have won. The best Thunderbolt is probably going to do for the near future is be the new FireWire.
re: your first point -- I looked into the MBP tear-down on iFixit when I wrote this post. They identify a chip that is 'probably the Thunderbolt controller' -- but yes, I don't think they identified if there was a separate DisplayPort controller, or whether it's built into the Thunderbolt controller.
Intel and Apple should market Thunderbolt not as a peripheral interconnect designed to compete with USB but an external PCI format designed to let you connect PCIe style devices (video cards/displays, RAID cards, realtime AV, docking stations, specialty or future ports -- USB3? 6Gbit eSATA? Fibre Channel, 10Gig Ethernet) to laptops, tablets, all-in-ones, etc.
These probably aren't completely mainstream needs (is USB3 even a mainstream need though?) but they represent things that can only be done with traditional form factor PCs now. Why buy a Mac Pro when the only thing holding you back from a compact low power mini or laptop is some port you can only get a on PCI card?
These probably aren't completely mainstream needs (is USB3 even a mainstream need though?) but they represent things that can only be done with traditional form factor PCs now. Why buy a Mac Pro when the only thing holding you back from a compact low power mini or laptop is some port you can only get a on PCI card?
What a POS article-- all speculation, no facts. And who in their right mind thinks Thunderbolt costs anymore than a single digit?
Thunderbolt has a major technical advantage with a protocol lighter in weight and superior handling of asynch communication. USB "beat" FireWire because it was cheap, and it was cheap because a lot of the logic was offloaded and host-driven. You can't get many gigabytes per second with that strategy. I'd be really surprised if you could drive, say, a very high resolution monitor off USB 3.0.