Slashdot Mirror
Can Thunderbolt Survive USB SuperSpeed+?
Lucas123 writes: "The USB SuperSpeed+ spec (a.k.a. v3.1) offers up to 10Gbps throughput. Combine that with USB's new C-Type Connector, the specification for which is expected out in July, and users will have a symmetrical cable and plug just like Thunderbolt but that will enable up to 100 watts of power depending on the cable version. So where does that leave Thunderbolt, Intel's other hardware interconnect? According to some analysts, Thunderbolt withers or remains a niche technology supported almost exclusively by Apple. Even as Thunderbolt 2 offers twice the throughput (on paper) as USB 3.1, or up to 20Gbps, USB SuperSpeed+ is expected to scale past 40Gbps in coming years. 'USB's installed base is in the billions. Thunderbolt's biggest problem is a relatively small installed base, in the tens of millions. Adding a higher data throughput, and a more expensive option, is unlikely to change that,' said Brian O'Rourke, a principal analyst covering wired interfaces at IHS."
I figured that all along. It took off on Apple hardware, with almost no pickup on normal PCs. That has finally started to happen a little - some upper end motherboards have 1 or 2 Thunderbolt ports now, and Asus has an add-on board for a few others - but it is really a niche thanks to its odd hardware requirements and lack of early adoption outside of Apple. USB is easier to use, and at least up to 3.0 has been backward compatible with older devices. With an even faster option, as long as they don't screw something up, I don't see how USB could not continue to be the leading connectivity standard.
William George
Thunderbolt isn't going to replace USB in all cases, but Thunderbolt isn't about the speed. It's about the protocol. Thunderbolt is basically PCI-E over a wire. Can you connect a GTX 780 Ti (http://techreport.com/news/26426/thunderbolt-box-mates-macbook-pro-with-geforce-gtx-780-ti) with USB 3.1? No? Not really a replacement then. Same goes for any other device that has traditionally been a PCI-E card. Or, you know, you can get an adaptor (http://www.sonnettech.com/product/echoexpressiii.html) and directly connect a PCI-E card.
Speed wise Thunderbolt is evolving too. At this rate there isn't much of a chance of USB 3.1 catching Thunderbolt. As the OP mentioned, Thunderbolt is still ahead of USB 3.1 and 40 Gbps Thunderbolt is coming soon (http://www.extremetech.com/computing/181099-next-gen-thunderbolt-details-40gbps-pcie-3-0-hdmi-2-0-and-100w-power-delivery-for-single-cable-pcs). But again, even is USB catches Thunderbolt, or both become fast enough, the protocols and designs of the connections makes them entirely unsuitable for each other's uses (you wouldn't connect a mouse and keyboard to your PCI-E bus directly via Thunderbolt.)
I have a really hard time caring about "up to 100 watts of power depending on the cable version", mostly because of the "depending on the cable version" part of the statement.
How is this different from DVI, which much or might not have multichannel audio, might or might not be analog, might or might not support 5 channel digital sound, etc., etc.?
One thing Thunderbolt has going for it is that a cable is a cable, and you don't have to worry about it. If you want negotiated power supplied over USB, fine, but don't make me search my cardboard box for the "most sincere USB 3.1 cable". Thanks.
Will the new spec be able to solve the problem of not knowing which way up to insert the USB plug?
*ducks*
READY.
PRINT ""+-0
A niche technology, used mainly by Apple fans. Part of it is just lack of need, and increased cost. Most devices work just fine on USB and Thunderbolt, being a PCIe bus more or less, has more hardware requirements on the device side than USB.
However it is also because of Apple's meddling. Apple got involved with it back when it was an Intel project called Lightpeak and paid Intel to influence the development. They wanted an exclusive on it, since Apple loves being "first", for a year and convinced Intel to integrate it with DisplayPort video. The problem with the DP integration is that it means you could no longer just drop in a PCIe card that would add it to a system, it has to be integrated in to a device to work with the GPU. So there's been little interest in it overall.
That'll probably continue for the foreseeable future. It isn't totally worthless, but there are few cases where it would matter much instead of USB, so its adoption is likely to be lackluster not so much because USB keeps getting better (though that helps) but because most of the things people want to do with an external connector, USB3 does "good enough" and everything has USB of some sort or another.
Can Thunderbolt Survive USB SuperSpeed+?
No.
Especially with
Look where all this talking got us, baby.
Maybe this is a "Let them eat cake" moment, but It's not like I'm a teenager in 1994. I don't really drag peripherals along with other upgrades. ... I can use my 5 year old USB/Thunderbolt/Whatever Display/Drive Array with my new computer, but I'll probably just buy a new one using the connector that works best. So, Is Thunderbolt going to be niche? Probably. Is it because I can't use my old thunderbolt display with my new computer? No.
Thunderbolt 2 will continue to exist until Apple chooses to abandon it. There is little reason for them to do so now.
Even when SuperSpeed+ products running 40Gb/s ship out, their biggest competitor will be USB 2.0, not Thunderbolt. USB 3.0 devices have been on the market for years, but popular webcams, mice, flash drives, and external hard drives are still made with USB 2.0 heads. Though I can't claim to know *the* reason, two possible reasons include manufacturing cost (which makes them unappealing to hardware companies) and a wide disparity of performance among 3.0 devices (which makes it difficult for customers to trust their purchases).
This is exactly what I came here to post. It's a shame, because FW400 was far superior to USB2.0. The problem lay with the peripheral manufacturers who didn't want to put in more expensive controllers and dual-ports on their enclosures. Heck, wasn't the iSight the only webcam for Firewire? No demand=no supply=high prices. FW800 was pretty much the same. Better tech, limited market, high prices, bang, whimper. I love that my old Mac Mini can transfer data between 3 daisy-chained FW400 drives much faster than it can transfer to a single USB2.0 drive, but the fact that enclosures are expensive and basically non-interchangeable with any of my other devices makes it a pretty niche market. Thunderbolt will probably follow the exact same progression, right down to the "new" faster Thunderbolt. Sure, its PCI-E, but 95% of consumers don't know, care, or need that capability. They buy on price and availability, plain and simple.
Thunderbolt 2 allows me to connect a 4k DisplayPort screen (or daisy chain two lower resolution DisplayPort monitors). Its connector is the same as mini-DisplayPort. It's small and convenient. Apple fit two TB 2 buses next to each other on my 13" MacBook Pro. Nice. Very high bandwidth, PCIe.
I don't want to plug a keyboard into this bus, because its overkill. Thunderbolt will probably never have any cost effective way to do a hub/star type topography. For general use lower bandwidth (haha, 1 gigabit is low bandwidth now!) peripherals I need USB. And my MacBook has that too. I wouldn't want it any other way.
That said, USB 3.0 seems like a ball of hurt compared to the difference between USB 1.0/1.1/2.0
Just look at the ads for USB 3 hubs. Most of them state which chipset revision they use, so you can look up whether or not your motherboard / OS will have difficulty with them. I built a FreeBSD 9.1 file server using usb 3 / usb 3 docks, but I failed them all back down to using their 2.0 interface due to persistent flakiness/dropping off the bus type issues. Rock solid on USB 2.0. YMMV, but I hope that USB 3 gets over its growing pains soon.
Fireside wasn't just more expensive hardware, it was more than 10x the license cost of USB.
I figured that all along. It took off on Apple hardware, with almost no pickup on normal PCs.
This, just like Firewire.
Thunderbolt is a solution looking for a problem.
Sure the fanboys can tell me I can hook up a small supercollider via thunderbolt but really, who needs to do that. What advantage does it provide me over USB when all my current devices are connected via USB and do everything I need them to. I use USB for connecting peripherals and storage devices. USB does this brilliantly. In order to switch to thunderbolt I need to buy a new PC (from a brand I cant stand), adapters for my current devices and add strenuous requirements when looking for new devices... So why not just stick with USB?
Calling someone a "hater" only means you can not rationally rebut their argument.
Heck, wasn't the iSight the only webcam for Firewire?
No, I had firewire webcams from 2 different vendors long before Apple released the iSight.
The original demonstrations I saw at conferences with the early firewire boards only had a Sony firewire webcam as firewire was originally only designed for moving video.
Using firewire for external hard drives and other tech came long after firewire/i.link was added to video cameras.
If your interface allows DMA, does that not mean that a malicious device can own your computer as soon as it gets plugged in?
Also, I thought I'd read that USB had DMA and hence this security problem.
Could someone who actually knows what they're talking about comment on this please?
Quattuor res in hoc mundo sanctae sunt: libri, liberi, libertas et liberalitas.
You might want such a thing with a Mac Mini or Mac Pro since they are more or less totally un-upgradable. So you might want an external card, even a graphics card.
For a standard desktop? No, you'd put it inside. Much cleaner, easier, and more sensible.
CPUs have gotten really, really, fast and for many things are seriously undertasked. Like I said, not knocking Thunderbolt for certain uses, but they are limited. USB3 on a modern system is capable of being "good enough" for most things. Audio? No problem, even USB2 has that licked. Video? Yep, USB3 can handle that. Data transfer? Well it is fast enough that even fast sticks are slower than it so no big deal. Network? It'll do 1gbps no issue.
Thunderbolt is faster, and lower latency, no question, but for most uses it isn't relevant. Same deal as it was with Firewire. That was not common at all.
Actually Thunderbolt has the additional issue of USB now being much better. Back when Firewire was first introduced, USB was not able to do many of the things it could do, at all. So it was use it or do without. However for quite a few things all Thunderbolt can claim over USB3 is that it is lower latency, or lower CPU load. Ok, fine, maybe that matters, but USB can still do it and so people will just use it.
I work in IT and I've seen next to zero uptake on Thunderbolt. Most of the places I've seen it is A/V type places, and mostly because they use Macs. They'll buy a thunderbolt LaCie drive not because they need it, but because that's what the new Macs use. That's only for the lower end stuff too. The higher end still seems to be all PCIe directly. For example the Avid Nitris DX won't work with the new Macs via Thudnerbolt. That's coming, but will be a separate adapter, in place of their native PCIe card.
How big is the installed base of USB 3.1 with the new C-type connector? ZERO! Apples and oranges, ladies and gentleman...
Sig?
DisplayPort lets you connect a 4k DisplayPort screen, or multiple streams (specifically the 1.2 MST). Thunderbolt is not required. It's fine that it is a Thunderbolt connector as well bunt don't get confused here. A DP connector coming off a regular videocard in a desktop will drive the monitor just the same. It is the DP 1.2 signaling that matters, not the PCIe lane of Thunderbolt.
If all you are doing with your Thunderbolt connector is hooking up displays, that's an argument AGAINST Thunderbolt since you aren't using it, you are just using DisplayPort.
Thunderbolt FTW
USB is the "mainstream, use for anything" connector. USB SS+ with type-C and 100 W power delivery makes it even moreso.
Thunderbolt is external PCI Express. Over long distances with optical cabling. Yes, there are few places in which TB is better than USB SS+, but in those places, USB SS+ can't compete - at all.
Need a 20 Gb/s connection to your storage array in the next room over? USB SS+ can't do that. Need an effectively-zero-latency connection to an external sound/video editing rig? Yeah, PCIe is your format, over Thunderbolt.
And don't expect Thunderbolt to sit still, either. While USB has plans to increase speed, so does TB. TB has PCIe3 coming up, and other improvements.
No, I never expect Thunderbolt to become even as mainstream as FireWire was, but it most certainly won't just go away, either.
Another non-functioning site was "uncertainty.microsoft.com."
The purpose of that site was not known.
Amen to the "Ball of hurt".
I design USB3 H/W....what. a. piece. of. shit. I have truly given up hope of engineering anything that will ever work universally, even Intels interfaces which you would like to believe would be a model reference design look like crap when you plug them into a gizillion dollar Agilent USB3 analyzer. Should I be be surprised? Probably not, USB has never exactly been the premium interface has it? Firewire didn't go away because USB was technically superior thats for sure. Thunderbolt just friggin' works, day in, day out, incredible and reliable performance. Sure cables are expensive, they have all sorts of clever active electronics...because...thats what it takes to make 10G in a consumer application work...not a $1.99 piece of injection molded crap from god knows what Asian hell chemical works. In fact Thunderbolts worst problem is ....Intel.....who seem to have a bizarre attitude towards people who want to buy components from them to make peripherals...I honestly don't get it.
I figured that all along. It took off on Apple hardware, with almost no pickup on normal PCs. That has finally started to happen a little - some upper end motherboards have 1 or 2 Thunderbolt ports now, and Asus has an add-on board for a few others - but it is really a niche thanks to its odd hardware requirements and lack of early adoption outside of Apple. USB is easier to use, and at least up to 3.0 has been backward compatible with older devices. With an even faster option, as long as they don't screw something up, I don't see how USB could not continue to be the leading connectivity standard.
Yes, we want technology that is 1/2 as good as what Apple offers and we are good with that. What dill weeds
This. Firewire was designed for video. Almost all video equipment of the era had firewire ports. Pro and prosumer video cams still do.
==================
Hippie Logger Jock
==================
I have a bunch of thunderbolt ports and handful of USB 3 ports on my Mac. I would love to trade my thunderbolt ports for some more USB 3s and maybe 2 HDMIs.
I remember for years my fellow mac people blah blahed about how firewire was so much better than USB and how it was the future. Basically it was the future for about 3 minutes.
I am smelling the same thing with Thunderbolt.
Basically my experience with thunderbolt is summed up by my monthly search for a reasonably priced thunderbolt to USB 3 adapter.
I thought USB had to pass thru the CPU/driver. Firewire had device DMA access and PCI is well, as low as you can go. I read about Thunderbolt 3 which is also being worked on; since both come out from intel you can expect Thunderbolt to be ahead of USB in terms of speed and flexibility. Don't see much need for high power output when Thunderbolt devices like displays will probably need their own power supply anyway.
Democracy Now! - uncensored, anti-establishment news
Failing to provide a normal micro USB connector is a pratfall everybody but Tim Cook could see coming a mile away.
When all you have is a hammer, every problem starts to look like a thumb.
good to see APK is as lucid as ever
---Saying gnome 3 is better than windows 8 not so much a compliment as it is damning with light praise.
They should just call Thunderbolt "USB 4". It's as related to what people call USB as USB 3.1 is. Different connectors, standards, etc.. The only thing it shares is the "USB" name itself. So might as well just call Thunderbolt "USB 4".
This is exactly what I came here to post. It's a shame, because FW400 was far superior to USB2.0. The problem lay with the peripheral manufacturers who didn't want to put in more expensive controllers and dual-ports on their enclosures. Heck, wasn't the iSight the only webcam for Firewire? No demand=no supply=high prices. FW800 was pretty much the same. Better tech, limited market, high prices, bang, whimper. I love that my old Mac Mini can transfer data between 3 daisy-chained FW400 drives much faster than it can transfer to a single USB2.0 drive, but the fact that enclosures are expensive and basically non-interchangeable with any of my other devices makes it a pretty niche market.
Thunderbolt will probably follow the exact same progression, right down to the "new" faster Thunderbolt. Sure, its PCI-E, but 95% of consumers don't know, care, or need that capability. They buy on price and availability, plain and simple.
One of the security failures of firewire was that it provided direct access to memory. In other words a malicious external device could gain complete control of the computer. Having your peripheral interface be PCIe is just as bad. USB for all its overhead is still more secure (assuming you finally fix some of the stupid windows autoexecute bugs)
I would be the first to agree... Thunderbolt is technically better than USB3.
However, that is not the point.
What is the advantage of USB? Simple. The connector.
I can plug a USB1 device into any USB3 port and it works. The reverse it also true, albeit at a pretty slow transfer rate.
The point it, the USB plug is ubiquitous while thunderbolt is already planning to change the connector again. That means buying adapters.
That also means that the next motherboard I buy most likely wont have thunderbolt on it. Which means I would not be buying any thunderbolt devices.
My impression is that thunderbolt is marketed specifically toward "Apple" people. Since I am not an apple person, I cannot say if this next part is true or not, but the impression that non-apple folks have is that "Apple" people will replace their gadgets every year or two with what ever new thing Apple has come out with.
If that is true, the connector thing is no big deal since you would be starting over every time.
That we keep talking about the two in language that exactly describes the two, but we completely ignore the language?
EVERY spec for USB refers to the "up to" speed and quotes the maximum theoretical burst transfer rate that is sustainable for only fractions of a second in host to single peer communication.
Thunderbolt's speed is the speed. period. 1 peer or 16 peers doesn't matter. You get 20Gb/s every second after every second. USB has never and is likely to never achieve that.
This was true of Firewire vs USB as well; USB claimed "up to 480Mb/s" but could never sustain that for any human sense-able time. Firewire 800 was flatly 400Mb/s. Firewire didn't advertise a theoretical maximum speed that you could get once in a while; it was a real-world measurable throughput when you were copying files.
So as long as people are ignorant enough to fall for marketing hype instead of actual useful data then USB will continue to dominate (and people will continue to purchase cars based solely on HP ratings)
Article X: The powers not delegated... by the Constitution...are reserved...to the people
Aside from the technical advantages that people keep on bringing up, one of the main non-technical advantages that Thunderbolt has is its certification process. Any USB chipset that is faster than USB2's theoretical speed is certified as USB3, whereas in order to get certified as Thunderbolt 1 or 2 you must actually reach the advertised speed.
When you buy a USB device (unless it's from a reputable manufacturer such as Intel), its actual speed is usually an order of magnitude worse than the advertised speed. That is a huge difference!
It wil have the same position as SCSI, Firewire etc...meaning CPU independent throughput, and higher-end systems, and more reliable.....
as for market share compared to USB...it was Apple who made USBpopular, with their first iMac.....USB was also original by Intel, and did hardly anything for 5 years...then came the first iMac, and suddenly almost overnight USB was hot...it took PC's about 2 years to catch up, from PS/2
would not underestimate the influence of Apple ...
Famous last words:"but...."
same niche market that Apple made popular ;)......it was Apple who made USBpopular, with their first iMac.....USB was also original by Intel, and did hardly anything for 5 years...then came the first iMac, and suddenly almost overnight USB was hot...it took PC's about 2 years to catch up, from PS/2
would not underestimate the influence of Apple ...
Famous last words:"but...."
On amazon instead of the stupid thunderbolt connectors. Only things you lose out on are: No/expensive express card versions, and displayport over PCIe.
Other than that there are already PCIe x1 to PCIe x16 adapter cards that act basically like a copper thunderbolt cable for externally connecting video cards for bitcoin mining and what have you. They're on Amazon, every dropship site that carries computer equipment, etc for 10-30 bucks (basically entirely passives so the 30 dollar price is a scam).
Given that and USB 3.1: Who *WOULD* buy thunderbolt hardware given the markup?
Isn't this exactly how everybody competes with Apple?
In the future, we will ship something WAY better than what Apple is shipping now, so obviously what Apple is shipping now is worthless.
Sleep your way to a whiter smile...date a dentist!
There is one thing Thunderbolt does have going for it, however.
Since it's essentially a PCI bus extension --- this means, you can add an external PCI chassis attached via Thunderbolt without needing special drivers, and in theory.... you can do things like add additional GPUs and arbitrary PCI devices to your desktop way beyond the expandability of your physical motherboard's or primary chassis' form factor.
There's really no way to accomplish something like that using USB, at least.... not with complicated specialized drivers being developed
I think you're talking about specific firewire host controller implementations rather than firewire in general. I imagine the same issue could happen with pcmcia or cardbus.
Except that USB is the badly designed but good enough for average users standard. Like choosing IDE over SCSI, not a decision made on the technical merits but by low margin motherboard manufacturers. The problem USB has won't be fixed with higher speeds, it'll still be a master-slave polling architecture and have latency issues.
Correct, it is possible for the host controller to disable DMA access. At boot, the OS can set the host controller into this mode so that peripherals cannot access physical memory directly. However, doing do has a significant impact on Firewire performance.
Just had a look at my local online camera shop. None of the Canon home, prosumer or pro-broadcast cameras have FW anymore and only the Sony pro-broadcast cameras have it, not their prosumer or home devices. Sad.
Trying to become famous by taking photos. Visit my homepage please.
Thunderbolt is the new Firewire 800
Minor nitpick: the Thunderbolt connector is not symmetrical. The writer must be confusing it with Apple's Lightning connector for iDevices, from which the new USB connector probably copied this feature.
(Actually, I believe the Thunderbolt connector is more or less symmetrical with respect to the x-axis, but this is undoubtedly not what the writer meant.)
R.Mo
this means, you can add an external PCI chassis attached via Thunderbolt
That still sounds like it would be far more useful for Apple products that tend to come in tiny cases with no other way of expanding them. Take a look at a mac mini and see how everyone has to plug in external devices for everything from additional monitors to backup hard drives.
In the future, we will ship something WAY better than what Apple is shipping now, so obviously what Apple is shipping now is worthless.
You're confusing usefulness with relevance. Thunderbolt is, and will reman, irrelevant to PCs, largely because PCs have plenty of internal expansion capability and sufficient USB ports, Display Ports, HDMI ports, etc.
I figured that all along. It took off on Apple hardware, with almost no pickup on normal PCs. That has finally started to happen a little - some upper end motherboards have 1 or 2 Thunderbolt ports now, and Asus has an add-on board for a few others - but it is really a niche thanks to its odd hardware requirements and lack of early adoption outside of Apple.
The real problem with Thunderbolt is that it is niche as designed, while USB is a general-purpose interface.
Thunderbolt is a slower-speed (because copper) variant of Intel's "lightning" interface... which has lots of potential because it's optical. I think doing it first over copper was a rational stepping-stone to fiber, but the problem is that it doesn't seem to be general-purpose. Instead it was made as a carrier for faster version of existing standards: PCI and Display Port. This is a big limitation.
USB, on the other hand, beat out the (in some ways) technically superior Firewire for (I think) the very reason that it was less specific, more generic.
To be really honest, I am mystified by the lack of a universal optical interlink by now. I mean, USB over fiber, or something like it. Hell, TOSlink has been around for a long time now. It works. It isn't terribly fragile. I mean, you don't want to try to wad up your cable in your pocket, but otherwise no big deal.
Intel's Lightning might have been it, but again I think it was too specific. Something a lot closer to just "moving bits over cable" is what is needed, IMO.
Is what will get me back on a laptop. I have a 'gaming rig'. There used to be all manner of reasons why laptops were worse - screens, speed etc. Now there's just GPU.
I'd love to be able to have a laptop I could lug around, get home and dock with a proper mouse, a proper keyboard and a GPU plugged into a great big monitor.
Using firewire for external hard drives and other tech came long after firewire/i.link was added to video cameras.
This was the problem with Firewire. It has lots of technical things going for it, but unlike USB it was not actually designed to be general-purpose. Neither was Thunderbolt. Thunderbolt was designed to support the specific protocols that Intel was building into its chips and boards.
Oh man, that is sad. Doesn't seem that long ago I was looking for helmet cams to use mountain biking, and they were all devices that connected to a recorder through firewire. Just needed a camcorder or laptop in your bag to do the recording. I must be quite a bit older than I thought.
==================
Hippie Logger Jock
==================
If they are changing the connector type, there is absolutely no reason to consider the installed base of USB. USB-with-C-type-connectors has an installed base of zero, not billions.
Bogtha Bogtha Bogtha
Adding external GPUs is a pretty niche application though, and of course you have to supply them with 300W+ for a high end one so that means yet another power brick. For external storage the performance gains are unlikely to be enough for most people to care and spend the extra money, especially once USB gets its shit together and provides enough power to run HDDs without an extra power supply.
Security is also a problem with Thunderbolt, Like Firewire any Thunderbolt device has full and unrestricted access to your computer's memory space. You really don't want to be plugging in random Thunderbolt devices you ordered off Amazon. If you encrypt your data you had better disable those ports, otherwise anyone with off-the-shelf forensic hardware can just read the key out of memory without even unlocking your machine.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
If only they invented a device with a C connector on one end and a A, B or MicroB on the other, they could call it something like "adapter cable"...
Same speed for an external hard drive. 30% CPU consumption for USB, 2% for Firewire.
I also know my usb 3 dock very well - it sometimes decides to connect at usb 2.0 speeds for no particular reason.
And don't get me started on the microusb connector that breaks if you sneeze in its general direction.
So I hope that TB takes its sweet time to die out so I can get some use out of it.
I apologize for the lack of a signature.
I had to buy three USB3 sticks until I found one that was bootable, I was lucky though because the same stick with a later revision lost this capability.
I figured that all along. It took off on Apple hardware, with almost no pickup on normal PCs. That has finally started to happen a little - some upper end motherboards have 1 or 2 Thunderbolt ports now, and Asus has an add-on board for a few others - but it is really a niche thanks to its odd hardware requirements and lack of early adoption outside of Apple. USB is easier to use, and at least up to 3.0 has been backward compatible with older devices. With an even faster option, as long as they don't screw something up, I don't see how USB could not continue to be the leading connectivity standard.
Try hooking an external SSD up to you machine via USB3 and then via Thunderbolt and you'll see why Thunderbolt is desirable if you are transferring large amounts of data: http://gizmodo.com/5980157/thu.... Take a look at the "Time to write 16.9 Gb of data" row in the table at the bottom and imagine you are transferring 3,4 or 500 Gb. There is about 250 Gb of data on the SSD in my MacBook Pro, large amounts of that data can change frequently meaning long backup times and cutting the time it takes to write that stuff up to disk in half is a major bonus. The problem Thunderbolt has had is not just backwards compatibility, i..e. that here are so many USB 3 devices out there that it is going to take a looooong while to put a dent in the USB monoculture (as you correctly pointed out). Thunderbolt devices have also had a tendency to be more expensive which didn't help either nor did the fact that up until now you have only started to benefit from Thunderbolt for real when using SSDs and they are also expensive which just aggravates the cost problem. When the USB 3 alternative is 2-3 times less expensive than Thunderbolt the choice for the consumer is obvious. If there is going to be a USB standard that is comparable in speed to Thunderbolt, backwards compatible with all the old USB2 and USB3 devices and that has a better connector, Thunderbolt is doomed. Intel should have pushed Thunderbolt way more aggressively i.e. handed out Thunderbolt product licenses liberally, provide motherboard and peripheral manufacturers with incentives or even sell Thunderbolt chips at cost.
Only to idiots, are orders laws.
-- Henning von Tresckow
That sounds amazing! You're saying that USB 3.1 means I won't be able to suck out the entire contents of your RAM like I can with a Firewire or Thunderbolt dongle?
Less really is More! Sign me up for USB 3.
I hear in the next version, USB 3.14, will come with a 360 orientation-less connector. It will be shaped like a standard headphone jack.
*LightPeak
especially once USB gets its shit together and provides enough power to run HDDs without an extra power supply.
It already does... for 2.5" drives.
systemd is Roko's Basilisk.
Correlation is not causation. What really made USB popular was when 2.0 came out, which was developed mostly by Intel and HP. At that point you could connect hard drives and get reasonable speeds, and hardware costs for slow devices like keyboards dropped too as HP found a way to do the timing required for USB 1.1 cheaply.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
"Heck, wasn't the iSight the only webcam for Firewire? "
Nope, not even close. Not only were there dedicated FireWire based webcams, but almost every digital video camera had FireWire (and could be used as a webcam) until they went from tape to flash/HD.
People who see FireWire as some kind of failure must have been completely absent in the digital video industry for almost a decade.
Where are the latency issues?
If you need low latency, you just need to use USB the right way.
It took off on Apple hardware, with almost no pickup on normal PCs.
It's not for PCs, it's for high-end video equipment. Thunderbolt will supplant SDI.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
I take it you don't do video?
Faster! Faster! Faster would be better!
To my mind that's where Apple went wrong with the new Mac Pro. Looks pretty, no internal expansion. So you have this nice object on your desk - with a bunch of trailing wires to your storage, optical drive, card readers etc...
If it had an internal optical drive, multiformat card reader (SD and CF at least) and room for at least one or two extra drives (hot swappable would be nice), it would be a great machine. I'd imagine more people would find that useful than the dual graphics cards. I know I'd prefer it all in one box, with just one external drive, or array (USB would do) for backups. And of course I could build a PC to the spec that meets my needs for a lot less.
I do actually use a Mac, and prefer OSX to Windows, but the hardware choices are sooo limited.
Sigs are so 1990s. No way would I be seen dead with one.
It's got titanium armour & protruding landing gear, it can survive anything.
VRRRRRRRRRP!
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
Adding external GPUs is a pretty niche application though
Not for laptops it isn't. I'm sure you can see how valuable that is for a thin/light laptop, and there have been quite a few successful DIY projects doing this. Though it hasn't taken off commercially yet, that may just be due to the fact that thunderbolt is really only ubiquitous on Apple laptops.
Wow, that's a lot of current - 20A at 5V. So, unless they change the voltage spec for this version, they'll have to use 14AWG wire under 6ft, and 12AWG up to 10 feet. That's going to be a big cable.
It's useful for anything that wants lots of power or space, which includes things like high-end GPUs, disk arrays, anything with lots of physical connectors (like analog capture devices), as well as various special-purpose ASICs and the like. It's all stuff you *can* do in a big case, or with a custom cabling system, but it's nice to have external be an option and use standard cable.
One of the obvious applications is desktop use of laptops. I'd like a super-thin, low-weight, long-battery-life laptop for mobile use. I'd also like a high-end GPU, full-sized Ethernet ports, and a slew of USB/etc. connections when I'm at my desk. Currently the solution to that problem is a device-specific dock that supports precisely what the laptop manufacturer decides it will (if the even offer one). External PCIe means nVidia can sell me a GPU that plugs into any laptop (and can be used with more than one laptop), and I can add whatever other bits I like individually or as packaged by arbitrary third-party manufacturers.
I take it you don't do video?
Your statement only prove Fireware is a niche interface, too. And, why, dou you still use tapes? ;)
Not all boards are designed to allow arbitrary DMA; modern memory control systems can provide segment protection for hardware just like they have for decades for software. And on systems that do not provide such protections the OS can disable DMA access -- Windows 8.1 disables DMA at boot and whenever the system sleeps/hibernates/etc., and linux has provisions to forbid the use of DMA while otherwise allowing normal access to devices.
USB has many more types of devices currently available, but that doesn't speak to the general-purposeness of the interface. Arguably USB is quite a bit more "specific" in application -- it defines specific roles for all devices on the chain, only allows a single host which drives the chain via polling, has lots of requirements for the way devices interact, etc. Thunderbolt, on the other hand, just provides a master-less low-level transport mechanism and lets devices do more or less whatever they want with it -- which is why you can use it for PCI traffic without any significant modification, whereas the design of USB prevents it from being used in a similar manner, even if you have a fast enough version of USB.
I figured that all along. It took off on Apple hardware, with almost no pickup on normal PCs.
This, just like Firewire.
Thunderbolt is a solution looking for a problem.
Sure the fanboys can tell me I can hook up a small supercollider via thunderbolt but really, who needs to do that. What advantage does it provide me over USB when all my current devices are connected via USB and do everything I need them to. I use USB for connecting peripherals and storage devices. USB does this brilliantly. In order to switch to thunderbolt I need to buy a new PC (from a brand I cant stand), adapters for my current devices and add strenuous requirements when looking for new devices... So why not just stick with USB?
OK, I can completely see how this is almost exactly like firewire, but next time you want to make an argument, please don't act so stupid about it. Thunderbolt (just like firewire when it came out) is considerably faster than the USB alternatives. This is why it is being used, and why people do choose it over USB today.
Just because "good enough" works for you doesn't mean it works for everyone.
And the #1 reason to not stick with USB? Murphy will all but guarantee you'll try and plug it in wrong the first time, every time. Fucking connector..
same niche market that Apple made popular ;)......it was Apple who made USBpopular, with their first iMac.....USB was also original by Intel, and did hardly anything for 5 years...then came the first iMac, and suddenly almost overnight USB was hot...it took PC's about 2 years to catch up, from PS/2...
Sadly enough, we're still getting new corporate models in...with PS/2 connectors.
Apparently we have not exhausted the world's supply of PS/2 connectors. And someone voted there will be no landfills full of them like AOL CDs or old Atari games.
You're confusing usefulness with relevance. Thunderbolt is, and will reman, irrelevant to PCs, largely because PCs have plenty of internal expansion capability and sufficient USB ports, Display Ports, HDMI ports, etc.
Not often I wish for mod points and don't have them, but this pretty much nails it.
Thunderbolt is solid technology - basically PCIe on a cable - but its relevance to machines that don't need PCIe on a cable (or provide an equivalent - ie: a docking station) is close to nil.
The use case for Thunderbolt on Macs, due to their typical design focusing on form factor over other factors, is reasonable.
The use case outside of Macs, is niche (to say the least).
Does USB 3.0 also allow for peer-to-peer interaction without the CPU? That was another cool thing about Firewire. (anon to keep mods)
For many use cases, USB3 seems fast enough but it seems inadequate in some ways.
Will we ever have "universal" connectivity system? Ie, will USB 4 have ports on the motherboard for drives and the drives themselves having USB 4 ports? Enough bandwidth to drive 4k displays (with maybe the display itself having some of the GPU logic)? Supported by "enterprise" software for things like OS-defined RAID or usable as a storage bus for VMware (or even just as a normal bootable drive for Windows)?
It seems like no matter how fast USB gets, it still ends up short on support for a range of useful tasks which end up falling to other interfaces -- SATA, SAS, DisplayPort/HDMI, etc.
Potential do-it-all replacement interfaces like Thunderbolt seem superior, but then lack ubiquity or seem to suffer from overreach, offering too much that complicates their adoption or makes them too expensive for ordinary tasks.
Notice how they are using a 4 drive RAID0 array for their benchmark. Also notice how the read speeds the 4x HDDs on Thunderbolt are almost identical to the SSDs (939 vs 936MB/sec). 234MB/sec average read for each HDD is insane, well beyond what they are capable of. It is obvious that on Thunderbolt there is some massive caching going on somewhere and they are effectively measuring a read from RAM.
IOW the test of completely flawed and obvious bullshit. In reality USB 3.0 speeds are absolutely fine for most people and you are unlikely to see much real world difference using only a single drive.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
I see a lot of concern here about backwards compatibility with any new interface. Why are we really concerned about this?
Your brand new server with quad-port Gig-E interfaces still auto-negotiates down to 10Mb speeds. Why? Because you might hook up your new $10,000 server to a $20 network hub you bought off eBay? Uh...no.
Apple had literally billions of devices in the market with the old sync connector. Then, they came out with an all-new connector, alienating entire lines worth of products. Did they go bankrupt? Was there some massive revolt in the industry? No, not quite.
My point is we should learn to move on. Stop worrying about backwards compatibility to ensure that we address scenarios that rarely happen, if ever. What exactly was Thunderbolt compatible with when it came out? Or Firewire? Didn't stop them from innovating.
Besides, there's a damn good chance that every single piece of computing hardware in your hands today will be replaced within 3-5 years, so I fail to see why we care even from a logistical standpoint. You won't even have the hardware in your hands to worry about backwards compatibility, and vendors will always see replacement as THE solution, so don't expect many long-term favors from them either.
so let's have a discussion of the technical merits. scsi:
true multi-drop bus, and i can't think of any others*. ide:
simplier and less expensive hardware, simplier command
set (one command 0xec to get all drive info), etc. at the
time, a multi-drop bus was not useful on pcs. nobody had
more than 1 fast (maybe 500 KB/s) drive. the real option
in 1986 was the ST-506, and it was worse in all respects.
* i've searched around for a few minutes, and can't find any
others (even that i disagree with) that apply at all to when
ide got traction in the late 80s/early 90s.
I recently copied some hundred GB of data from my workstation to an iSCSI datastore with maxing out the gigabit connection - while several other datastreams also read and wrote to it. If I'd had 10gig ethernet (and a good offloading driver on the workstation) I probably would have come near to the thunderbolt data, which is point-to-point and not (for now) able to support several computers at once on the same storage.
Did I mention the storage was in another building, while 30 meters (i.e. next room) for TB cable cost 649$ (discounted) plus transport fees? See http://eshop.macsales.com/item/OWC/CBLOPTTB30M/
I simply don't get the excitement over "but it's for data storage". *real* fast datastorage has for years been Fiber Channel or maybe iSCSI over 10GigE (perhaps 40gigE nowadays), and for "it has to sit on my desktop" USB3 is fast enough.
The problem lay with the peripheral manufacturers who didn't want to put in more expensive controllers and dual-ports on their enclosures. ... They buy on price and availability, plain and simple.
And they were right. Overall what is cheaper? Six devices and a computer with expensive controllers in them? Or one computer with one less expensive controller in it and six devices with really inexpensive controllers in them?
Daisy chaining - I'm sure it was nice, and I can even think of one or two cases where I might have used it if it were available, but in the end the fast majority of my devices were all within reach of me, which meant they needed to be ~the same distance from the computer, which meant they could both use the same length cables (give or take). Daisy chaining doesn't change the # of cables I need. At best is lets me use one shorter cable and slightly declutters the back of my computer. But that is what a USB hub is for, and combined it was cheaper than FW controllers, enclosures, and cables.
It was a very expensive solution to a problem very few people had: the need to move massive amounts of data in/out of a peripheral at a time when the user is unwilling to wait. That is to say, not even external-backups really needed it because those tend to be fire and forget.
At least that was my experience and I actually did buy an external FW HDD for backups because I really was just that impatient :)
If you can't be good, be good at it!
That sounds amazing! You're saying that USB 3.1 means I won't be able to suck out the entire contents of your RAM like I can with a Firewire or Thunderbolt dongle?
What would really be amazing would be if instead of snark, your comment contained some sort of awareness of the existence of an IOMMU, which prevents this sort of attack on modern hardware.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
It is a feature of the design, and the reason firewire is fast. It can access memory directly, that makes it fast, and that makes it a giant security hole.
Funny thing. Thunderbolt has the same issue.
USB won because it was less expensive, which permitted it to be more generic. We could have had firewire everything down to keyboards, but that would have cost us a lot. So we got USB. Today USB is faster than 1394, so we lost nothing. 1394 continues in fact to be a massive PITA. Daisy chaining often doesn't work, the controllers support half as many devices as promised, and the state of driver support is poor even on Windows, let alone Linux. Using 1394 on Linux is a neverending nightmare of abandoned interfaces.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
DMA access is no longer a security issue because modern CPUs use IOMMU visualization to effectively disable bootable DMA access for external devices.
Any modern system requires that DMA devices negotiate DMA access with the BIOS or OS. Secure Boot is one example how to easily block drive-by DMA booting. And no, you can't just walk up to a computer with a TB port and plug in to dump memory. If the OS doesn't map the DMA ranges, your device will get nothing. Behold, the power of virtualization DMA mapping techniques.
http://en.wikipedia.org/wiki/USB
in my timeline windows 98se was the first wide spread use of usb technology. a full 2 years before apple decided to add usb, based on your '5 year' timetable. personally i thought usb 1.1 too slow, and while i did eventually buy a 4 port powered hub, the main thing i used usb for was joysticks and mice, at the time you got 2 usb ports on a computer, and so when printers went usb i had to get one to have 3 usb ports. also digital cameras and scanners both went usb, and then i needed 5 ports... and then i quit using printers and scanners, but then external hdds were cheapest backup options, then flash memory came out and those eat usb ports for breakfast... connecting a whole lot of them is easy and sometimes it is nice to have several pluged in for various uses... luckily motherboards come with 6-12 ports for decent desktop computers, and some laptops have 4 usb ports which then means using a powered hub, because now bluray burners, cell phones, wifi/ethernet dongles, flash memory, mouse, hdds, etc etc etc. unplugging things is possible but not ideal.
https://www.gnu.org/philosophy/free-sw.html
I don't disagree. I was heavily involved with HD video for a while, so I'm aware of the prevalence of FW connections in that realm. However, that's hardly mainstream. I daisy-chained FW400 and FW800 stacks, and appreciated the speed, but 99% of the consumers out there wouldn't know that it was even possible. They were willing to accept slower transfers in return for cheaper devices. The tragedy of the commons strikes again.
Are still going to be USB3.
Thunderbolt is too expensive and offer no benefit in this use case since 5 Gbps is enough for any hard drive. RAID arrays are mostly SATA, internal. External Thunderbolt RAID arrays will remain a niche product.
"Why buy a desktop when you can simply plug the PCI-E cards straight into your laptop?"
Because it's cheaper and offer more expension possibility. Have you seen the price of these external Thunderbolt cases? It's even cheaper to get both a laptop and a desktop than to get a high performance laptop with Thunderbolt and such a case.
Also you can't do anything serious on an external Thunderbolt graphic card because it will be limited to x4 speed instead of x16.
More importantly, can the A10 Thunderbolt 2 survive congress, and Obama wanting more JSF's
>It took off on Apple hardware, with almost no pickup on normal PCs.
Are you talking about USB or Thunderbolt? Apple was the first to use USB as well. Look how popular it is now.
MABASPLOOM!
You forgot Infiniband and SRP/iSER ;)
20Gb infiniband gear is suprisingly cheap compared to 10GbE.
Isochronous USB is "low latency" relative to human perception, around 250us. TB is "low latency" as in tens of nanoseconds. It's a difference of about 2-3 magnitudes. Isochronous is also low bandwidth, so you need to use higher latency bulk transfer modes for high bandwidth, which increase latency into the millisecond ranges. They're talking about 40gb USB. At 1 ms of latency, you will need a 5MB buffer, but with TB, your buffer would only need to be around 64KB.
For some work loads, latency will become very important at high bandwidths, like 10gb and 40gb. Most people just doing bulk copies of data in one direction probably won't notice.
I take it you don't do video?
The monitor I'm using to type this is hooked up via USB 3.90 via a docking station. Thunderbolt is far more optimized for video but for any situation where there is a choice, USB is almost certain to win largely due to the size of the installed base. Furthermore Thunderbolt is not the only or even most common way to hook up a monitor. It's not that Thunderbolt is a bad standard but the problems it "solves" really are just not a big deal. USB covers 90%+ of the potential use cases for Thunderbolt and other video and network standards cover the remaining ones adequately.
This is why it is being used, and why people do choose it over USB today.
Except, not so much.
Outside the Apple foodchain, I have yet to see a single Thunderbolt device (actually used... yes, I technically have seen a monitor that supports it - connected by HDMI), from the home to the recording studio to the server room. Nada.
Thunderbolt may well have a world of advantage over USB, just like Firewire did. And just like Firewire, history will relegate it to a niche of obsolescence caused by a simple lack of adoption in the face of a number of much more popular standards.
This is just nuts. Your USB device can pretend to be a keyboard whenever it wants to and start opening shells.
Since it's essentially a PCI bus extension --- this means, you can add an external PCI chassis attached via Thunderbolt without needing special drivers
Problem is that there is barely any use case for doing that. I can think of a few specialized applications where an external PCI chassis makes sense but not nearly enough to justify Thunderbolt. USB covers 90%+ of the use cases of thunderbolt and other existing technologies cover the rest. Thunderbolt may be a bit more elegant than some but Firewire was more elegant than USB and we all know how that turned out. A low end "good enough" technology will beat a high end expensive technology in the long run almost every time.
I would disagree. The actual production cost difference was small, but manufacturers wanted to shave pennies, rather than raise the price and worry about the competitor getting placed in stores due to a $1 per thousand unit discount resulting from USB only implementations.
Consumers definitely noticed the slow speed of USB. Think back to when scanners became a desktop reality, before MFPs. USB was 12Mbps, and firewire was 400Mbps. Scanning was atrociously slow over USB. Also, when external hard drives were becoming more common, customers were frustrated at how long it took to transfer even over USB2. And the more you transferred, the slower it got. God forbid you try to use the computer during that time.
No, Firewire was killed by retailers and other penny pinchers. Consumers did not care about $2 on a $1000 computer, or $2 on a $100 scanner.
The brains of a chicken, coupled with the claws of two eagles, may well hatch the eggs of our destruction.
Like Firewire any Thunderbolt device has full and unrestricted access to your computer's memory space
Is that always the case with DMA?
Imho USB drives are slow & demanding as fuck. I've wanted DMA support.
for corporate desktops, there is absolutely nothing wrong with a ps/2 connector, so why not use them?
"They were pure niggers." – Noam Chomsky
Not so.
USB was developed back in the early 90s and was ready by 95 or 96. I remember seeing motherboards that supported it.. Long before Microsoft announced USB support, which officially only started with 98. (On 95 it's complicated. There's a patch for USB and AGP support, and it was integrated in later OSR releases. 95 releases were damn messy. As far as Microsoft is concerned, though, AGP and USB were officially a 98-only ordeal)
Even then, USB ports were more a curiosity, and reletively unused.
Know what made them take off? The iMac. The original "Bondi Blue" space age looking bubble iMac with it's 233 mghz G3. And no ADB ports (ADB is the Mac equivalent of PS/2. Legacy keyboard and mouse connectors, but a true chain-able bus) or serial ports. And no floppy drive.
The iMac was widly popular and the only peripheral expansion was USB. Suddenly everyone needed USB keyboards and mice and Floppy drives and printers. And you know what? Those devices also worked on PCs too.
Know who's brainchild the iMac was? Steve Jobs. So really, you have Jobs to thank for USB.
It only has pci-e 2.0 X4 - overhead so that is limited for a GPU even more so with any other thing on the same bus.
The macpro really needed 2 cpus for more pci-e lanes they planed to have 2 pci-e flash slots but that was cut due to lack of pci-e IO / DMI limits.
seriously? future USB spec that is years away may be faster than thunderbolt 1? No fucking shit.
Think thunderbolt is standing still? No.
Besides, they are not competitor technologies. Thunderbolt can do things that USB can not. And it actually works, pretty much 100% of the time.
Not saying there is no place for USB. It is a cost effective, fairly high speed bus. But comparing it to thunderbolt is just not legitimate. Most people don't need thunderbolt, and that's fine. But for those who do, USB 3 (and no doubt USB 4) are not and will not be competitive.
I run: Windows, OS X, Linux, FreeBSD. Just because you have a hammer, doesn't mean everything is a nail.
USB has always sucked badly, especially sustained transfers when all that speed is badly needed not to mention when older usb protocols are used when it get really, really sloooow.
To me the real question is: Can usb3.1 survive Thunderbolt 3,4 or 5???
USB has ALWAYS lagged behind everything else. The pattern seems to be they are slightly better for a very short time then something else comes along and... WHAMMO! HEAD SHOT, BABY! YOU SUCK; YOU GO HOME NOW.
You are not a target user. Good for you, pay less for USB. Move along, nothing to see here. If, however you would like to say, hook up 10 gigabit networking to your laptop (or mac pro for that matter), a fiber channel SAN, a GPU, or all manner of other possible options, then it is relevant.
I run: Windows, OS X, Linux, FreeBSD. Just because you have a hammer, doesn't mean everything is a nail.
Quote from the article:
"Even as Thunderbolt 2 offers twice the throughput (on paper) as USB 3.1, or up to 20Gbps, USB SuperSpeed+ is expected to scale past 40Gbps in coming years. 'USB's installed base is in the billions. Thunderbolt's biggest problem is a relatively small installed base, in the tens of millions."
USB install base in the millions? Yes - with current connectors and technology. Making a giant assumptive leap that a new interface, with new cable will suddenly have an install base in the "billions" is - a giant assumptive leap. It might well do so - but not because there is already an install base in billions - for an adapter/chipset/cable that is incompatible with the newer adapter/chipset/cable. It will do so for other reasons - mainly what it costs to implement $$$, not because of an existing install base.
I think what killed USB initially was the Windows demo where Bill BSODed a machine by plugging in a scanner.
I run: Windows, OS X, Linux, FreeBSD. Just because you have a hammer, doesn't mean everything is a nail.
No in theory about it.
I run: Windows, OS X, Linux, FreeBSD. Just because you have a hammer, doesn't mean everything is a nail.
Both Windows 98 and the original iMac (the first 'USB-only' Mac) were released in 1998.
For pro video and pro audio use cases
Pretty much the very definition of niche use. When your livelihood depends on performance of a technology, it's not hard to justify the extra cost of a specialized technology. That does not describe the vast majority of use cases.
And, for what it's worth, firewire is still very popular in the pro video and pro audio world.
And RS-232 is still very popular in industrial machinery. Neither one is going to displace USB at this point as the go-to general purpose device interface. There are lots of interfaces out there which are popular for specific uses but looking forward will remain niche. USB is good enough for enough stuff that it is unlikely to be displaced any time soon. I don't think USB is great but it's not going away either.
Actually, I kind of miss PS2 keyboard/mice systems. The thing is they work. If you plug a USB keyboard/mouse into a Windows system it will work until Windows takes over from the BIOS and then won't work until the driver is set up in Windows.If you plug in a PS2 device it works the whole way through the process. This may have changed now with UEFI since I haven't done a lot of work on these systems yet. It may also have changed with Windows 8 but again haven't used enough systems with Windows 8 to know for sure.
Now I realize that I have just opened myself up to a series of trolls about how "Microsoft sucks and that is your real problems" so bring it on. ;-)
For external storage the performance gains are unlikely to be enough for most people to care and spend the extra money, especially once USB gets its shit together and provides enough power to run HDDs without an extra power supply.
And especially as, at some point you just get a NAS / SAN for roughly the same that a TB array would have cost, and now you have shared network storage.
IIRC USB3 has some improvements designed specifically to fix the whole polling issue when using a device as storage.
http://en.wikipedia.org/wiki/U...
No it doesn't. At least not spinning rust - those things easily require 2+A to start up. On USB alone, that would prevent them from even starting, nevermind running.
The only reason we "get away" with it is due to cost reasons - each USB port is limited to 500mA, but often they're all ganged together to one big overcurrent switch. E.g., if you have 8 USB ports, you have a 4A overcurrent switch controlling power to all of them.
Since the vast majority of users do NOT use all 4A simultaneously, a USB hard drive can easily spike its current draw without doing anything bad.
But problems arise in more mobile situations - devices with less USB ports suddenly it can matter if they have all sorts of things plugged in - mouse, a thumbdrive, etc. and suddenly plugging in their hard drive kicks everything off the bus as the overcurrent switch cuts power.
Or you may also see it with heavily loaded chains where certain actions suddenly lose USB completely because every drive decides it needs to start simultaneously, fail as they trip the switch, etc.
Right now a lot of stuff only works because the USB implementers were lazy and cheap. But if you have devices with limited USB ports, you run into issues quick.
Heck, even some thumbdrives can't start up on 500mA. Don't ask me why.
A good use case is a thunderbolt docking station. You can get USB ones, but they're chintzy and barely work. But thunderbolt ones that provide a pile of ports that are real, legit ports (e.g., a real serial port using native 16550A drivers, real parallel ports suitable for bit-banging), external graphics ports that are suitable for running huge displays (USB display sucks), real Ethernet networking (USB3.0 required for GigE), etc. Plus piles of USB3.0 ports for all those USB peripherals that remain at your desk.
While you're out and mobile, your PC has limited connectivity (WiFi/BT/WWAN), uses onboard graphics, etc., which are great for a light and mobile laptop. But you then go to your desk, plug in a single cable (Thunderbolt will soon support device charging, both ways) and your thin and light becomes an ultra powerful workstation with external high end GPU (multiple monitors, too, since the internal GPU can drive displays through Thunderbolt), networking, etc.
Anyone who's tried a USB docking station junks it within the first week as they're quire useless and your displays lag to hell and back. (If you're lucky, you can do YouTube in QVGA mode)
Setting USB legacy mode in the BIOS addresses the issue of not being able to F8 into windows to tell it you really don't want to run a chkdisk right now.
GENERATION 25: The first time you see this, copy it into your sig on any forum and add 1 to the generation. Social exper
I'm tired of hearing about any of these standards.
I want something where every port can be host or client or maybe everything is just peers and it doesn't matter. I want it to work in a star topology AND also be capable of being daisy chained. That way I can hook up as many peripherals as I want to a limited number of ports but I can also chose to buy a hub for better performance.
Come on, we had these kinds of features in networking technologies 30 years ago already!
I have zero faith in the USB spec since version 3 essentially killed compatibility with almost all audio interfaces. Many high-end audio interfaces will only use USB2, FireWiire or Thunderbolt. If PC manufacturers ditch Thunderbolt for USB3.1 it will shrink the A/V market for PCs and push people back to Apple. Also, backwards compatibility is important when you have an expensive piece of hardware that may well last 10-15 years as you move from PC to PC.
Personally, I would have liked a raid array controller that I could hook up my 12-drive raid-6 array and being able to attach it to a laptop.
Or attaching my OCZ Revo for a superfast scratch disk.
Or a 10Gbe connection (or faster) so I can access my local NAS box.
But I understand these aren't mainstream use cases, so what I could use to for isn't popular enough to keep such an option viable by itself.
How am I going to charge my Tesla with only 100 watts of power capacity? This sucks.
This security hole can be easily plugged via the use of the IOMMU: http://en.wikipedia.org/wiki/DMA_attack#Mitigations
The low latency of memory mapped interfaces such as PCI, Firewire and now Thunderbolt make them excellent buses for hard real time systems which need timing down in the micro or nano second range. I work with a motion control system that uses fire wire to connect smart servo drives to a Windows host running a real time kernel (Ardence RTX). It can maintain >10us latency between the drives on an 8000MHz Pentium 3 and Windows still has plenty of CPU time left. You can't do that with USB.
Another disadvantage of USB is that devices cant send interrupts and have no DMA capabilities (a security advantage). The host has to poll each device which is costly in both time and CPU cycles. Firewire devices copy data to memory without the CPU using DMA thus avoiding interrupts. Thunderbolt and PCIe can both take advantage of DMA and message signaled interrupts.
Now, change out the word "desktop" with "laptop" and you start to see some of the value. Yes, you can have a fiber channel controller on your laptop, and mount a couple exabytes of storage as a local volume.
Slashdot still doesnâ(TM)t support Unicode after it was added to the HTML standard in 1997.
I hear that. Imagine, oh that is a Fibson guitar, it needs Fibson cord and amp. If USB breaks the standard, obsoletes my gear, I will look around. I have dodged Apple because of this merry go round, from the horrendous iTunes, to expensive firewire mistakes, that I did not make. It is the single reason I do not buy Apple gear, or anything with despised proprietary connectors. Never.
If USB dumps this advantage, and I am forced to change? Stupid move.
OT Rant: Woman last week, new evidently 1st gen VW Hybrid. Car would not start because the trunk was open. No joke. I did not ask her why she didn't just buy the 5th gen Prius. Some folks willingly walk into hell. Just open the door, they line up
Firewire smoked USB all the way through. 6 devices connected to a USB hub will drop throughput to abysmal levels. I currently run on a mix of USB 3, USB 2, FW 400/800, and eSata. My preferences today are eSata, FW 800, FW 400, USB 3, and finally USB 2. The reasons are that USB 2 gets about 20MB/s throughput with more than 2 devices on the controller, USB 3 slows down significantly as well although currently I can only get 2 devices on my controller, so it's not as obvious. FW 400/800 just keeps going without a significant drop, and eSata, even though this is limited to just a single drive, I can peak out FW800 with up to 130MB/s transfers which eSata can handle.
just a question, is there hardware out there that is not protected, but can be attacked if someone put in some connectors in their PC? or is this something that was taken into consideration from the start?
have you seen my sig? there are many others like it but none that are the same
No, Firewire was killed by retailers and other penny pinchers. Consumers did not care about $2 on a $1000 computer, or $2 on a $100 scanner.
This is misleading, there were never anywhere close to just $2 difference on a FW vs USB scanner, from anyone - regardless of sales channel (direct, online, wholesale, B2B). The part and production cost difference added up to more than this, and you will have your add-on on top of the extra cost too, and on top of this Apple required $1 per port in license fees. At a minimum you would have 10+% difference on a $100 class scanner. And if you don't think consumers care about that, you haven't worked much sales and marketing of this type. The retailers care because they know the customers do care (even when they in research say they are willing to pay more, their actual buying actions often prove otherwise).
Thunderbolt only exists to be proprietary. So sad apple no longer innovates.
And the #1 reason to not stick with USB? Murphy will all but guarantee you'll try and plug it in wrong the first time, every time. Fucking connector..
As the summary states,"The USB SuperSpeed+ spec (a.k.a. v3.1) offers up to 10Gbps throughput. Combine that with USB's new C-Type Connector, the specification for which is expected out in July, and users will have a symmetrical cable and plug just like Thunderbolt but that will enable up to 100 watts of power depending on the cable version."
first to use? no, first to mandate its use however would be correct (the imac only had USB this is correct, but USB was out before hand)
have you seen my sig? there are many others like it but none that are the same
until i got an external tb ssd that is so damn fast it is indistinguishable from internal ssd.
*LightPeak
Thank you. My mistake, and I am properly corrected.
just a question, is there hardware out there that is not protected, but can be attacked if someone put in some connectors in their PC? or is this something that was taken into consideration from the start?
Basically, only recent machines even have an IOMMU, and only the very latest machines necessarily have a supported one. And even then, it's typically only in the fancy processors, not the budget ones.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
yes, it's the new firewire - a lot of professional audio and video equipment already uses thunderbolt (and has skipped usb 3.0). it's the de facto standard that followed firewire. firewire provided a more reliable & stable transfer rate than usb and it looks like thunderbolt does the same. also, daisy chaining is nice if you got a lot of harddrives.
USB won because it was less expensive, which permitted it to be more generic. We could have had firewire everything down to keyboards, but that would have cost us a lot.
But my point was that USB was designed to be generic, from the start, while 1394 wasn't. Thunderbolt also was not designed to be generic, but rather to carry the specific standards that Intel was already putting in its chips and boards.
It's usually easier to take something generic and adapt it to specific uses than to take something designed for specific uses and make it generic.
And the #1 reason to not stick with USB? Murphy will all but guarantee you'll try and plug it in wrong the first time, every time. Fucking connector..
Not just the first time, the second time, too. It's only on the third try it will fit.
But my point was that USB was designed to be generic, from the start, while 1394 wasn't.
IEEE1394 has generic transport, and you've always been able to use it. You're just mistaken. Because of the design of IEEE1394, it requires more hardware to implement it. USB devices can simply be dumber than 1394 devices, they don't even need a microprocessor. 1394 devices do, consequently they have a higher minimum cost. You could always have had a QWERTY keyboard on 1394, but it would have cost fifty bucks for the cheapest one instead of five bucks, and an Apple keyboard would have cost two hundreds instead of one.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
You are ignoring his actual point - he's talking about hooking an SSD up externally in high-end applications (but this would also apply to any high-throughput device, including RAIDs). It is very clearly faster than USB 3. The HDD test is rather irrelevant (but if you look at the 16GB file transfer test, which doesn't hit the HDD caches in any meaningful way, you'll see it's still 2x faster than USB). But Gizmodo sucks as they always have, yes.
The number one reason to be wary of USB is that half of the hubs and adapters out there barely work right. I can still go to Fry's and find cheap crap that will go wonky even interfacing with just USB 2.0 devices. USB 3.0 is the Wild, Wild West in terms of quality, and they are already jumping the gun to 10 Gbps? I am very afraid. Put it this way: To achieve USB certification, all you have to do is pass a ten minute functional test. Intel's stuff may cost a lot more and may never get great market penetration, but their stuff is almost always rock solid. You get what you pay for.
I write multi-platform software that utilizes some fairly dense data streams. Some of the devices are USB; some are Ethernet.
For the curious, it's software defined radio; it's not unusual to see 16-bit data at 4 MHz coming in across the interface, which is a 64 Mb/s load, doable, but heavy, for a 100 Mb/s connection. There are devices that do 18 MHz at 16 bits. That can take up a considerable portion of a 1 Gb connection - 288 Mb/s. Day to day on my machine, I work with 800 kHz 16 bit data, or about 12 Mb/s for a 400 kHz receive bandwidth. All of these add some overhead on top of the main data stream too, as does TCP itself, so really, the loads are heavier.
Ethernet support basically has to be written once. How networking operates can be abstracted to be the same across OSX, linux, and Windows without much effort at all (Qt does this), and the TCP network stack is generally pretty fast and well behaved. And it is stable. (I could insert a rant about Apple's half-assed UDP support here, but...)
USB support varies considerably from one platform to another. Abstracting it is a huge PITA, requires lots of conditional compilation, and doesn't turn out that well, either. And that's without dealing with new USB drivers coming from the OS manufacturer that overwrite (and break) previously installed drivers. Nothing as fun as a user telling me "everything was working, now nothing is."
It's to the point where I won't directly support new USB SDRs. Every one I've done so far has added to my support load; whereas adding a new network-based SDR is painless, works the day I release it, and is still working a year later. In fact, if they adhere to the established SDR networking protocols, I don't even have to do *anything*, it'll just work. When someone asks me to support a new USB SDR, I tell them to write a network server for it using the established networking protocols. That way, when the USB breaks, as it seems it inevitably will, it's the responsibility, and support load, of the network server author, rather than mine. :)
It's also worth noting that as networking speeds have gone up, previous networking devices still work, and the new ones can crank pretty fast.
Then we get to remote issues; how far can you place a USB device from the computer that's talking to it? How about firewire or Thunderbolt? Now consider Ethernet. I can run an SDR that is quite a bit more distant from my computer, and I don't really have to do much to make that happen. I can have a bunch of them on the same subnet or even remote them out in the world. And then there's sharing. Same device, no reconfiguration, I can use it from OSX or Windows or linux without it even knowing that's happening. I can even go wireless transparently. I routinely sit on the couch and surf the HF and VHF bands on my laptop using WIFI as the data transport from the SDRs.
For any device that can manage it and still obtain acceptable performance, Ethernet looks like an excellent solution to me. If you really must have bus-like speeds, as in graphics engines and SSDs, ok, fine, Thunderpants, Displaypunt, Firepliers, etc. But please, if you're designing a peripheral with any kind of lesser bandwidth requirements, consider Ethernet as your interface. Everything is so much easier to deal with.
IMHO.
I've fallen off your lawn, and I can't get up.
Actually, there are a few things wrong.
I haven't checked lately, but PS/2 wasn't designed as a universal port. Keyboard and mouse had dedicated ports, although the connector was the same and easily confused. Most servers corrected this issue electronically in the mid 2000s, but I still ran into it on desktops where the user couldn't figure out why their mouse didn't work.
A keystroke logger can be attached to a PS/2 cable and be entirely transparent to the OS.
PS/2 connectors have pins that can bend, and are more difficult to orient correctly.
Gamingmuseum.com: Give your 3D accelerator a rest.
I assume you mean Firewire. And you may be correct about the license costing 10x USB, but that makes the difference look far larger than it practically was.
IIRC, when I bought my first and only Firewire peripheral, a CD-RW drive (both were relatively new technologies at the time), that drive cost me $397. I believe the cheapest USB CD-RW was on the order of $300 and painfully slow by comparison. The license cost for Firewire was $1/port. So if the manufacturer didn't want to lose money on the license fee, it could have cost $399. Firewire lost for a variety of reasons, not least of which being that only Apple and Sony really ever embraced it, but licensing cost wasn't one of them. Sad really, since it was a superior standard in all measurable ways.
Correlation is not causation. What really made USB popular was when 2.0 came out, which was developed mostly by Intel and HP. At that point you could connect hard drives and get reasonable speeds, and hardware costs for slow devices like keyboards dropped too as HP found a way to do the timing required for USB 1.1 cheaply.
Correlation does not imply causation
This is just nuts. Your USB device can pretend to be a keyboard whenever it wants to and start opening shells.
Nope. Not if you have your system configured not to automatically install new keyboards.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
If it had an internal optical drive, multiformat card reader (SD and CF at least) and room for at least one or two extra drives (hot swappable would be nice), it would be a great machine.
They really should be offering a machine like that, and nobody is sure why they aren't. There's also always the option to offer the machine with some kind of external expansion bus and offer some kind of sidecar, but historically that's a fool's game.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Is there a rule against plugging it in with the lights on or something? Or do you just blindly throw cables at your PC until they stick?
Unity? Screw that: XFCE. Slashdot Beta? Screw that: SoylentNews. Australis? Screw that: Pale Moon. UX developers DIAF
I had it with Windows 95 OSR/2. As usual Apple fanatics reinvent history by claiming Apple is responsible for innovation they clearly aren't responsible for. They did not develop the spec, did not develop the prototypes, did not develop most of the market, ad nauseum.
Interesting, which device are you using to use the FireWire interface? I would love a programmable arduino like device with faster io for low latency signal acquisition.
IEEE1394 has generic transport, and you've always been able to use it. You're just mistaken.
I am not "mistaken". Firewire was intended to be a replacement for SCSI. While it includes the ability to transfer generic data, it was not -- I repeat: not -- designed from the beginning to be primarily a general-purpose data transfer. It was deliberately optimized to be a SCSI replacement and to carry video.
Yes, it is capable of doing those other things, but the primary aim of the design was specific, not generic. The "generic" capabilities might be adequate or even well done; but that was not the main design goal.
It is the "specific" part that caused (causes) it to be more expensive than USB.
yeah, but that's why i said corporate desktops. they'll mostly be used in a static configuration and very rarely would be keyboard or mouse be removed and reinserted. when they are, it would be by IT.
the logging issue is interesting, i guess, but using usb for that reason would be "security by obscurity" and bruce schneier would come down and rape your children to death.
also, why is it impossible to make a completely transparent usb 'logger'? at worst, can't you just sample the voltages and then do protocol analysis (or whatever you would call it) on the stream? i really have no idea but it seems plausible, just more difficult, though someone only has to build the database.
"They were pure niggers." – Noam Chomsky
USB's ubiquity/success can be attributed to Apple. Had they not made the push, parallel port printers and PS2 keyboard/mice would have stuck around for a lot longer than they did.
MABASPLOOM!
If they are changing the connector type, there is absolutely no reason to consider the installed base of USB. USB-with-C-type-connectors has an installed base of zero, not billions.
The connector is backwards compatible - USB2 devices can be plugged into USB3 hosts. Same applies for USB3/USB3.1.
So John Doe can still plug all his old stuff onto his fancy new PC with USB3.
Thunderbolt, on the other hand, is completely different. Manufacturers need to add an extra controller AND an extra connector, and people will be hesitant to buy something "different".
USB has a huge overhead, is slower, etc, etc. Thunderbolt is techonologically superior.
USB will win because it's more popular. It's happened over and over, and will keep on happening, because it's not the tech-savvy people that make these decisions.
The iSight is an IIDC camera, while practically all consumer digital video cameras are DV devices. while DV streams a compressed video stream, IIDC can be used to load (parts) of uncompressed individual frames. IIDC is still common in industrial image acquisition, but I think it's slowly being replaced with GigE.
No joke...my low end Compaq laptop purchased in '05 had firewire. In fact, I did digital video capture/editing on an Apple-free setup for almost a decade.
> When the USB 3 alternative is 2-3 times less expensive
What does '2-3 times less expensive' mean?
No. (I have one of these kicking around somewhere...worked much better than most of the USB webcams that were on the market back in the day.)
20 January 2017: the End of an Error.
That's not true, I work in the (relevant) industry to this topic. Intel will phase out PCIE.
http://www.accountkiller.com/en/delete-slashdot-account Stop visiting Slashdot.
No it doesn't. At least not spinning rust - those things easily require 2+A to start up. On USB alone, that would prevent them from even starting, nevermind running.
I've been using external USB-powered 2.5" HDDs for years, connecting to laptops, desktops and USB hubs. Works just fine in all situations, which is why companies such as WD are still selling these things like hotcakes.
It is the "specific" part that caused (causes) it to be more expensive than USB.
Which specific part? You haven't actually named one.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Your entire target user lineup sounds like stuff that
1) I would not be using on a laptop, I'd be setting up a server / workstation (if for nothing else something powerful enough to take advantage of it, I have a pair of SATA 3 SSDs paired with an i7 and sure I can saturate the gigabit port on my laptop but I sure don't do that often, I have the SSDs so I can work on the files locally).
2) Are all available via PCI express cards already (I'd be interested in how your OS would respond to your GPU being unplugged accidentially). The Mac Pro is the only device where I can see it being useful (mainly because the extra options desired cannot be added via more traditional offerings).
Seriously, I wouldn't rule Thunderbolt out yet.
It's taken how many iterations for someone to design a symmetrical cable?
Sure, Apple got there first, but it still took decades!
We clearly are not working with the best or the brightest on either side!