Slashdot Mirror
New Thunderbolt Revision Features 20 Gbps Throughput, 4K Video Support
hooligun writes "The next-gen Thunderbolt tech (code-named Falcon Ridge) enables 4K video file transfer and display simultaneously in addition to running at 20 Gbps. It will be backward-compatible with previous-gen Thunderbolt cables and connectors, and production is set to ramp up in 2014. An on-stage demo with fresh-off-the-press silicon showed the new Thunderbolt running 1,200 Mbps, which is certainly a step up from what's currently on the market."
Asus has a motherboard.
--Joakim Ziegler
That's simply false. There's a large amount of Thunderbolt accessories, including video gear, PCIe expansion chassis (very useful for laptops), and docks. Sonnet just announced this Thunderbolt dock, which seems to be a pretty great deal for laptops.
--Joakim Ziegler
But in the end, it all comes down to cost. Current Thunderbolt displays are rather expensive. Heck, I picked up a dual-link DVI monitor of the same resolution for $275 on ebay! why pay three to four times as much for something with only a small few bells and whistles added on?
Thunderbolt, overall, is great in terms of performance, but it just seems to be well beyond what most folks are willing to pay. It's like that guy who brags about how "My car has a Turbo Kit option from the dealer" but he NEVER SPENDS THE MONEY TO GET IT.
The external drives, the only situation that I'd actually be interested in, are also stupid expensive. In the long run, just better off either using E-SATA, USB3, or internalizing the drives. Same goes for daisy chaining monitors. Want to run tons of monitors? Install more video cards! woo.
no more coffee for me after 5pm, k? ._.
I don't know what you've been smoking but Thunderbolt is an Intel invention. They worked with Apple on implementation with Apple's most obvious contribution being the VESA compliant mini-Displau port connector. For their efforts, Apple got a good six month lead on their competition as they had products the day Intel released the specs. Incidentally, Apple got the Thunderbolt trademark and then transferred it to Intel.
Well, there's spam egg sausage and spam, that's not got much spam in it.
What could I connect this to?
Several RAID arrays, gigabit ethernet, multiple monitors, misc external storage (like single disks or a DROBO).
All with one connector...
Yes Thunderbolt stuff was slow to come out, but the rate of arrival has picked up.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Thunderbolt (codenamed Light Peak)[1] is a hardware interface that allows for the connection of external peripherals to a computer. It uses the same connector as Mini DisplayPort (MDP). It was released in its finished state on February 24, 2011.[2]
On the same day, Apple released new iMacs with Thunderbolt.
Now what are your facts?
Well, there's spam egg sausage and spam, that's not got much spam in it.
Do you have any insight why they even bother with TB when 10Gb Ethernet already exists and has been deployed for ages? I.e. why not just use 10GbE instead?
It seems like reinventing the wheel for no real gain.
When all you have is a hammer...
The main reason for using Thunderbolt over 10Gb Ethernet is that one has a fairly significant protocol overhead (Ethernet) while the other is primarily a bus protocol, and operates at a much lower level than Ethernet does. Each has their strengths and weaknesses, each has their application.
...si hoc legere nimium eruditionis habes...
Ummm. No.
That enclosure doesn't do RAID, it's a JBOD enclosure. The peak transfer rate for the mini-SAS interface is 3Gbs (3 Gigabits, not bytes, per second) this is an absolute maximum of 375 MB/sec. The real-world performance of the unit will then depend on the RAID card you're using and will typically be somewhere lower than the peak theoretical performance of the interface. I don't know what drives you're putting in there that can each do 500MB/sec (SSD?) and I don't know what RAID card you propose to use that'll let all eight SSDs run at their peak rate.
The unit I was talking about (http://www.areca.com.tw/products/thunderbolt.htm) on the other hand, with 8 drives in it has a measured real-world performance of 650MB/sec read or write via a single Thunderbolt cable, using RAID 5 that's done in hardware in the enclosure itself.
This 650MB/sec is the actual performance that the BlackMagic Disk Speed Test gave me on a MacBook Pro 13" laptop connected to the RAID with 8x 1TB Western Digital hard drives in it.
Thunderbolt is faster than SAS, SATA and SATA II. Thunderbolt is faster than 2, 4 and 8 Gb/sec Fibre Channel - Thunderbolt is a 10Gbs full-duplex interface, so can transfer 20Gb/sec at it's peak. That's 2.5 Gigabytes per second (1.25 in each direction).
Specialist Mac support for creative pros, Melbourne
Because 10GBe doesn't expose PCI to your peripherals
How about ExpEther technology virtualizes PCI Express over Ethernet.
Why are manufacturers coming out almost-but-not-good-enough connector standards one after another?
Both tablets and TVs are leaving PC displays in the dust, and new PC connector standards that aren't even available yet already don't have the required bandwidth to support displays that are coming to market now, let alone in the future!
For example, support for full 4K video over 20 Gbps is bullshit, because some aspect of the full spec has to be abandoned:
Resolution: 3840 x 2160
Bits per pixel: 30 or 36 (10 or 12 bits per color channel)
3D or High Framerate: 120 fps
This adds up to: 3840 * 2160 * 30 * 120 = 29.8 Gbps.
Sure, you can drop the framerates, but then expect to have a headache viewing 3D. The bit-depth can be lowered, but then expect visible banding when using gamuts that are wider than sRGB. The resolution can't be lowered, because calling 3840 pixels "4K" is already a stretch.
Originally Light Peak was supposed to basically just be an external PCIe bus (and it could be internal too). The idea was a connector for things that need lower overhead than USB, and also hopefully eventually a single connector for all kinds of things. With the original goal of 100gbps, that would have been realistic (optical was the original interface design).
However things got changed pretty quick in part for cost reasons, but also because Apple got involved (meaning gave Intel money). Apple is obsessed with less cables because cables = evil in their mindset. So it got changed to be display + PCIe on one cable.
That had negative implications for the bandwidth, but also for the cost and ability to implement it. If it was just PCIe, well then a PCIe-thunderbolt card would be real feasible, and you could add a thunderbolt port by hanging it off the PCI bus. However with display integrated, it needs to work with the integrated display adapter and all that jazz.
Ultimately more cost, and thus less interest. While some Apple types might salivate over the prospect of one cable that goes from a laptop to a monitor, and then a bunch of non-monitors ports on that monitor, most people don't care.
Why, most of the exploits these days are on OS X anyways.
I haven't thought of anything clever to put here, but then again most of you haven't either.
Speaking for myself, 48-bit image processing in a layer-based, non-destructive paradigm. Software defined radio -- extremely demanding, that. High speed data, maximized low latency requirements, no particular limit to the amount of processing one might like to apply to the signals / spectrum segment. I use Logic for musical performance, and that's absolutely got to stay local, again latency must be managed to then nth degree and the more processing that can be done within that bound, the better. None of this can be handed off down the network; it just wouldn't work well. Or at all.
I sure would like to see core-per-file parallel compilation, too, but instead, all of the dev environments I use keep the source and object on HD and do them one at a time, serially. Big projects take much longer than the hardware at hand could manage. XCode, QT, gcc/gmake... all serial sluggards.
Well, sort of. My understanding (which could be wrong) is that those 12 cores are all hyperthreaded, so as long as you're not requiring FPU or blowing cache a lot, you're more-or-less running 24 cores. Not sure as I have an older generation dual four-core w/o hyperthreads (but with two sets of four FPUs.) It's a fair bit of computing power; my Macbook pro, a dual core with lots of resources, can't even come close to keeping up with my older Mac pro.
That varies. We use an integer-math approach that maximizes CPU power and doesn't rely so much on the highly variable GPU capabilities of the various Mac models (not to mention having to drink more of the Apple kool-aid than really needed... we really like portable code.) There are a number of advantages to this, first among them the availability of a great deal more RAM (it's useful to stack a hundred or so 48-bit astro images, for instance) and so more layers-per-image, but also a more consistent performance for the end user. Some Macs -- the current minis, for example -- use bottom feeder Intel shared RAM GPUs. They aren't anything to write home about. Even so, the machine can be carrying a 2.6GHz i7 w/16gb. We often outperform Aperture, Apple's poster child for GPU use in image processing; can't really say why, but there it is.
For some value of bigger jobs, sure. For the things that I do on my desk, no.
In any case, a Macbook isn't going to provide even close to the same level of hardware as a Mac pro, which was really my only point above.
I've fallen off your lawn, and I can't get up.