Slashdot Mirror
Thunderbolt On Windows: Hardware and Performance Explored
MojoKid writes "Intel's Light Peak technology eventually matured into what now is known in the market as Thunderbolt, which debuted initially as an Apple I/O exclusive last year. Light Peak was being developed by Intel in collaboration with Apple. It wasn't a huge surprise that Apple got an early exclusivity agreement, but there were actually a number of other partners on board as well, including Aja, Apogee, Avid, Blackmagic, LaCie, Promise and Western Digital. On the Windows front, Thunderbolt is still in its infancy and though there are still a few bugs to work out of systems and solutions, Thunderbolt capable motherboards and devices for Windows are starting to come to market. Performance-wise in Windows, the Promise RAID DAS system tested here offers near 1GB/s of peak read throughput and 500MB/s for writes, which certainly does leave even USB 3.0 SuperSpeed throughput in the dust."
If that is a Promise RAID box, I Promise the numbers are totally imaginary. Maybe they got that performance for about a second, on a full moon, in the dark, with no one watching.
I don't doubt thunderbolt can do it, but I doubt anything Promise says.
TLDR: Promise sucks.
What I expect from Thunderbolt is not to use it as a link to a storage device, but to a graphic card. This way you could have a CPU and memory heavy laptop to carry around, but then you could dock it at home and connect it to the external graphic card and play some video games.
Apparently this interface can do 10 Gbps, and that sounds like a good start.
OEMs are more pragmatic than that. If it's a salable tech that is already developed and offered by a major player like Intel, they'll use it, whether it smells of Apple seconds or not.
I don't think that's the target. Look what Apple has done with Thunderbolt: it's their primary docking adapter for their laptops and they've made their new monitors the equivalent of docking stations. Basically, it has just enough bandwidth to carry a DisplayPort signal plus USB.
I have a 2009 MacBook Pro which commutes with me to and from the office. It's a tad annoying to have to plug in six different cables every time I get to my desk and unplug them when I leave (which is a few times a day due to meetings). I've wished for a decent docking station; Apple seems to be averse to including a connector for this purpose, and the third-party solutions I've tried are as kludgy as one might expect. The addition of Thunderbolt doesn't have me rushing out to replace my laptop (obviously), but I'll be happy to have it when the time comes to retire this machine.
(As for why I have a MacBook vs. a Windows laptop... well, it's rather well built (and has survived a few drops to date), is Unix-y enough to allow me to develop on it and still deploy the results to our Linux servers, and has built-in grep and zsh.)
With an IOMMU in between, which the OS can use to protect sensitive memory.
No it's been a year of geeks like you claiming that Thunderbolt was somehow an Apple technology when it was not. It is an always has been Intel technology; Apple helped Intel develop it. Apple did not get an exclusionary deal for their efforts; they simply got a year head start on all the other computer manufacturers. In that year others have implemented it. OEMs have been slower no doubt because some have wondered if I was worth implementing.
Well, there's spam egg sausage and spam, that's not got much spam in it.
That's all true of the PCI-E slots you already have on your motherboard. Do you hot glue those too?
If only they could have just multiplexed a USB over the displayport, or firewire, but no, they had to provide a root access connector that is now standardized across many devices.
You don't have even the slightest idea what you're talking about. You grabbed hold of a few concepts that you apparently don't fully comprehend and then used them to rant about surveillance. A sibling of mine posted the IOMMU thing already, but that wasn't the only howler in your post. Firewire also allows DMA so your purported solution wouldn't work for exactly the wrong reasons you were complaining about Thunderbolt. And even if they were legitimate objections, you're screwed if an attacker has physical access anyway.
I have developed a truly marvelous proof of this comment, which this signature is too narrow to contain.
It is not much different than firewire with DMA access and hotplug? IOMMU's plugged that hole years ago.
No sir I dont like it.
There are already multiple alternatives to thunderbolt.
That's why some of us are less than excited about a closely held standard that requires the purchase of a Mac or a $400 Intel motherboard and also requires $50 cables.
The enclosure in the article is pretty expensive too.
> There is a market, but you're not it.
You are even less in it than I am.
A Pirate and a Puritan look the same on a balance sheet.
First, that complaint is true of essentially all external busses, including SCSI, SAS, eSATA and virtually everything else except USB. They're setup that way for a reason -- DMA is much, much faster.
Second, memory access on modern busses is routed through an IOMMU. This provides both memory abstraction (which is vital on modern architectures) and allows the OS to control which devices, if any, can access a particular memory location.
(I'll assume you're speaking from a windows point of view on FW?)
As does the vast, vast majority of computer users out there. I don't think anyone would argue that fact, right?
That's why USB 3.0 is going to ultimately be the standard...it's backward compatible and everyone is still using mostly USB peripherals. Until that changes (which it probably won't, regardless of capability, look at how long VGA has been hanging on, and that standard is 30 years old), USB x.0 will likely be the dominant standard for peripherals based on that fact alone.
Geeks like going out and buying new peripherals to take advantage of the new capabilities of new standards. Most people, though, just want something that's going to work with the shit they've already got.
I don't think that's the target. Look what Apple has done with Thunderbolt: it's their primary docking adapter for their laptops and they've made their new monitors the equivalent of docking stations. Basically, it has just enough bandwidth to carry a DisplayPort signal plus USB.
I realize you're not trashing it and it was probably a verbal slipup, but I have to say, you seem to have an odd definition of "just enough". ;)
One Thunderbolt connector carries two full-duplex 10 Gb/s links, or 20 Gb/s total (bidirectional). 60Hz refresh of a 2560x1440 27" display with 8 bits per channel needs 2560*1440*3*8*60 = 5.3 Gb/s. One lane of PCIe 2.0 is equivalent to 4 Gb/s (5 nominal, but 8b10b line coding means it's 4 actual, while Thunderbolt has a much-closer-to-100% efficient line coding). So Thunderbolt can refresh Apple's Thunderbolt Display with enough bandwidth left over for >3 PCIe 2.0 lanes.
The Thunderbolt Display doesn't just have USB, by the way. It also has a gigabit ethernet port and FW800. Those, and the USB, are all local PCI Express host controllers which communicate to the computer by tunneling PCIe through Thunderbolt. That's how Thunderbolt works: it tunnels PCIe and DisplayPort packets. All other protocols require a PCI Express host controller at the far end.
Should I even get you started on FireWire? :P
Non impediti ratione cogitationus.
I remember USB being very rare in the Bondi iMac days... Intel might have put the controllers on their boards but a lot of manufacturers don't use Intel motherboards, and even if they used the chipsets. Nobody (for large values of n) who owned a PC in 1998 was ready for USB when MS "got on board," they got USB when they bought their first Windows 2000 or XP system in the subsequent three years.
The question is, will Microsoft getting on board even a factor this time? Thunderbolt doesn't require drivers, it's just serialized PCI Express -- manufacturers can put these ports on their motherboards and they work out of the box.
Don't blame me, I voted for Baltar.
Get out the hot glue gun... Any device with thunderbolt has the full PCI bus exposed. Plug in the right gadget, which cops and crooks WILL have, and you completely and utterly own the system down to the lowest level, memory and drive contents.
Sorry, no. I'm a professional linux kernel developer. Unless you have something cooperating within the OS to set up a mapping any DMA request from the thunderbolt device is going to get dumped on the floor by the IOMMU. (See the IOMMU wikipedia article if you're unsure how this works.)
Yeah, but back then it stood for Useless Serial Bus.
Pull my finger for my public key.
I remember having a ten pin USB header on my motherboard (AT style) for many months before I knew what an actual USB port even looks like.
I never ended up even using USB on that motherboard; that machine just got upgraded to a newer one that did have external ports (my first ATX motherboard!) and even then, I didn't actually USE those ports until I bought a USB optical mouse months after *that*.
USB definitely got a VERY slow start on the Wintel side. Which is ironic, because the opposite just happened recently with USB 3.0; generic whitebox PCs have had 3.0 for over a year now, whereas Macs are only just getting them right now.
If memory serves me correctly, it took years before USB became the defacto peripheral port. In that case Apple also was an early adopter. And you are complaining about Intel's newest technology taking a year. Please let's not let history and logic get in the way of your Apple hatred.
Thunderbolt is not meant to replace USB which is another Intel technology. So if you think USB3 is going to compete with Thunderbolt, I think you need to read more about what Thunderbolt is. Thunderbolt is meant more to replace eSATA and more as eSATA is really only for the one purpose of connecting drives. You can't run USB and ethernet over eSATA.
Well, there's spam egg sausage and spam, that's not got much spam in it.