USB 3.0 Is Ten Times Faster; Get It In 2010

thefickler writes "Seagate and Symwave are jointly demonstrating the first consumer applications of USB 3.0 at CES, showing a Seagate FreeAgent drive running through a Symwave USB 3.0-compatible storage controller device. According to Symwave, this will result in 'speeds previously unattainable with legacy USB technology.' Which means, if you understand PR-write, it will be much faster."

Comments

What about the rest of us?

[Tubal-Cain]

Which means, if you understand PR-write, it will be much faster.

What does it mean if you don't know PR-write?

Re:What about the rest of us?

[lorenzino]

What does it mean if you don't know PR-write?

Clearly much slower

Re:What about the rest of us?

[jd]

Hell, in PR-speak it could also mean it's much slower. So much slower that legacy devices would time out. Or it could mean it's so expensive that the marketing people can double the amount of speed they buy.

Re:What about the rest of us?

[zippthorne]

You're going to have a lot of trouble writing drivers for it?

Re:What about the rest of us?

[iNaya]

Haha, that's something I always mean to say, and forget to when someone says "if you know ... then ... is ..."

Re:What about the rest of us?

[Anonymous Coward]

If you don't know PR-write, you'll look up that USB2 usually delivers 25MB/sec max, then calculate that ten times faster is 250MB/sec, and then realize that a Seagate FreeAgent drive tops out around 100odd MB/sec being a 1TB SATA drive internally.

thus, you'd guess that it is somewhere between 2 and 4 times as fast.

Re:What about the rest of us?

[Hoi Polloi]

It means you need $50 Monster cables to get maximum performance with it!

Re:What about the rest of us?

[omeomi]

It means Firewire and External SATA will still be faster, but everything will come with USB connections anyway for no apparent reason...

Re:What about the rest of us?

[hairyfeet]

Uhh....there is a very good reason why even though Firewire and ESATA are faster that everything will be USB- Because it is cheap and it is the one connector that you KNOW that your customer will have. I tried carrying a Firewire external drive a few years back. I ended up selling it because I almost never got a chance to use it. Every PC I came across, unless it was truly ancient, had at least a USB 1.1 port. If Apple would have given away the Firewire patents for free as Intel did with USB we would all have Firewire ports. But IIRC they wanted $1.00 a port for Firewire compared to $0.0 for USB, and in the cutthroat world of computers it simply wasn't worth it for the manufacturers.

So while it would have been nice if ESATA or Firewire 800 was everywhere, I just don't see that happening. From what I understand USB 3.0 will be backwards compatible with 2.0 which is backwards compatible with 1.1. So that is over a decade that you can be assured that your device will be able to connect with their machine. And that kind of mass penetration is simply worth more to the customers and manufacturers than a speed boost.

Re:What about the rest of us?

[Yvan256]

Or 5$ MonoPrice cables.

Re:What about the rest of us?

[necro81]

One other problem Firewire had in becoming widespread was that it required a more beefy, dedicated chip. As far as I understand it, Firewire is implemented mostly in the chip, cutting out the CPU, and creates a more-or-less guaranteed bandwidth. This is why it was popular with camcorders - you could always be sure that you could transfer video in realtime. Same for high-end sound equipment.

USB, on the other hand, while it has its own controller chip, is moderated largely by the CPU and memory bus. If the computer is under heavy load, the USB throughput suffers. Peripheral devices are at the mercy of the host to control things. This is fine for things like mice and keyboards, which transfer relatively little data semi-asynchronously. You don't need such a robust high speed bus for such lightweight peripherals. But for hard drives and other devices, USB has some catching up to do.

Royalties aside, the Firewire chipset that could implement a high-speed bus robustly cost more than the more lightweight USB controller. In the particular case of the iPod, a portable device, having the Firewire controller eventually took more board space than Apple was willing to provide. While every computer had a mouse and keyboard, not every computer was made with the ability for connecting external drives. So there, too, the economics played in USB's favor.

Re:What about the rest of us?

[gad_zuki!]

Why isnt esata everywhere? The mb manufacturer has already paid for a sata license, put the chip in, put in the interface for the internal connectors, etc. How hard is it to put in a damn physical connector for the esata? esata should be replacing usb for disks. Another physical connector and external connector costs pennies. Instead we have millions of computer users pegged at well under 400mbps for disk operations. In the real world USB2 is something like 30 or less MB/s because of all the overhead.

I can understand cutting off firewire, but the sata is already there. It would be nice if one of the big OEMs started promoting it. Save your usb ports for flash drives and mice.

Re:What about the rest of us?

[supernova_hq]

Whooosh

Re:What about the rest of us?

[Anonymous Coward]

But IIRC they wanted $1.00 a port for Firewire compared to $0.0 for USB, and in the cutthroat world of computers it simply wasn't worth it for the manufacturers.

Are computer manufacturers, or consumers for that matter, really concerned about an additional $1.00? So a laptop computer costs $500 instead of $499. The additional cost incurred from a patent hardly seems to be an issue.

Re:What about the rest of us?

[im_thatoneguy]

Since another reply covered the problems with Firewire. I'll also add another problem with eSATA besides large swaths of incompatibility with a diverse range of computers. The connector for eSATA is garbage. It is incredibly weak and unreliable. Using an eSATA plug in a mobile environment is just asking for a disconnection. It can't take any abuse. A USB plug on the other hand will stay connected through a hurricane without any disconnects.

Re:What about the rest of us?

[im_thatoneguy]

Power? I have a 2.5" drive I power off of USB2.0. With eSata I have to plug a drive into a wall (or power it off of a USB port. :D)

Re:What about the rest of us?

[timeOday]

Why isnt esata everywhere?

Everywhere, as in mice, keyboards, mp3 players, gps, and every other gadget released in the last 10 years? I doubt you meant that, but I sure don't want a different interface for every peripheral if I can get by with a "universal serial bus." USB lives up to its name and I like that. With a 10x speed boost it'll be very viable for a while longer.

Re:What about the rest of us?

[Dogtanian]

Are computer manufacturers, or consumers for that matter, really concerned about an additional $1.00? So a laptop computer costs $500 instead of $499. The additional cost incurred from a patent hardly seems to be an issue.

Doesn't that apply to any port, whether on a computer or on a peripheral?

Bear in mind that $1 is the *cost* price to the manufacturer; to sell a device with the same (slim) profit margin they'd have to charge a bit more to the consumer.

And while it doesn't sound a lot, such decisions are probably one of many cases where you could say "oh, it's only a tiny difference in cost for a big improvement", but when you go for that in every case it mounts up to a $10, $20 or $30 difference. Ever shopped for food and realised that the deceptively small differences in price that seem so harmless when choosing individual items are the ones that mount up?

And the great unwashed frequently *do* buy on price, even when it seems like a false economy to more informed people. If a manufacturer is building to a price-point and selling at razor-thin margins, the consumer probably isn't that bothered, but that $1 is a significant difference to the manufacturer's profit.

Also, the fact that USB is universally accepted probably brings in more economies of scale, both in terms of buying the controller chips *and* in terms of selling X units of that peripheral. USB's small advantage at the start probably spiralled, as these things tend to do, so that this is the position we're in now.

Re:What about the rest of us?

[Yvan256]

Given that USB 3.0 cables will have fiber optic cable(s) inside them, I don't think Hoi's comment is much of a joke. Be prepared for very expensive super-speed USB 3.0 cables.

Re:What about the rest of us?

[rfuilrez]

Exactly. Take out that $1 part on 1 million laptops for example. Bam. That's $1 million that they didn't have to front, and eat the loss if no one purchased them.

To bring in a car analogy. I believe it was chrysler a few years back, when doing a new year minivan left out a single piece of trim from the previous year that cost them $1. Now, I don't know how many they sold, but I guarantee they probably sold quite a few at the same price as the previous year. $500,000 profit for not putting a $1 part on? Sounds good to me.

Re:What about the rest of us?

[jc42]

According to Symwave, this will result in 'speeds previously unattainable with legacy USB technology.' Which means, if you understand PR-write, it will be much faster."

What does it mean if you don't know PR-write?

Clearly much slower

Heh. While you might be right, a careful analysis of the statement leads to the same result whether you know PR-write or not. What it really means is that, under ideal test conditions, it will be at least a tiny bit faster. The statement is satisfied if a single measurement shows USB 3.0 to be 0.0001% faster than any previous USB.

Of course, for the majority of us who won't be using it under ideal, carefully controlled conditions, it well might be slower. That wouldn't make the PR statement false. The only way to falsify it is to show that under all conditions, there exists some earlier USB hardware+software that will do one particular task at the same speed or faster.

When reading such things, it's useful to ask yourself "How bad could it be and still make the PR claim true?" And you should understand that words like "faster" don't necessarily mean "by a large amount"; they just mean "by any amount greater than zero".

Re:What about the rest of us?

[gad_zuki!]

Your comment is like saying "WHoa, I dont need USB. I have two serial ports and a parallel port on my computer. Not to mention two ps/2 ports. Everything made in the history of personal computers uses serial and parallel." Yet here we are.

Re:What about the rest of us?

[gad_zuki!]

Powered esata is here. It came later in the spec, but its here. This 32gb drive supports it:

http://www.kanguru.com/eflash.html

29mb/s in USB2 and 80mb/s in esata.

Re:What about the rest of us?

[dgatwood]

A single USB hard drive on a USB 2.0 bus at full throttle sucks down something like 40% of a CPU core by itself. With a 10x boost....

:-)

Re:What about the rest of us?

[Golddess]

Aren't PS/2 ports just for keyboards and mice? And could serial and parallel ports power devices attached to them? Not to mention that you're comparing three different ports to a single, speedier port.

Now I'm not saying that a better spec couldn't be designed, but as long as the USB spec keeps getting speed increases, why would the average user require any other kind of plug?

Re:What about the rest of us?

[timeOday]

The fundamental problem with serial, parallel, and ps/2 ports is that they were just ports - not busses. My livingroom PC is out of serial ports (already using one for controlling the DTV converter box, and one for the remote control IR). I could add a couple more serial ports at the cost of a PCI slot (if I had one free, which I don't). Since USB is a bus, I can connect my mouse, keyboard, external hdd, gps receiver, digital camera, and 3 family members' mp3 players - all at the same time if I wanted. Granted it would be better if bluetooth were displacing USB on all sorts of devices and were fast enough for external HDDs, but that doesn't seem to be happening.

Re:What about the rest of us?

[badkarmadayaccount]

And now you know why Intel is pushing USB. ;)

Re:What about the rest of us?

[supernova_hq]

I guess that will make USB lights a little easier to make :D
Assuming they don't use IR or something...

That speed comes at a cost

[stokessd]

Given that USB is PIO and not DMA, the faster the bus runs the more processor intervention is needed. Given how cheap and fast our processors are, that's not a huge deal, but it's not like a DMA based transfer just got faster, it means that the processor is going to be more busy too.

PR-write or not, it will be a PITA just like USB2.0 until it's built in and common.

Sheldon

Re:That speed comes at a cost

[Anonymous Coward]

So basically, change is hard and oft met with resistance, there's no such thing as a free lunch since the world is made of tradeoffs, and technology that's still in development is also not yet perfect. Surprise!

Re:That speed comes at a cost

[beakerMeep]

I'm not very knowledgeable about USB flavors but I think I'm like most people in hoping they just settle on a standard. The Marketriod Speak of the USB people is exactly what's been wrong with all of this. First there was Hi-Speed USB, then USB 2.0, then Hi-Speed USB 2.0, Then Ultra-Titanium Jet-Powered-Turbo USB, then something where they claimed you'd get 76 virgins for using USB.

Honestly, I think even some of the geekiest computer users start to not care when bombarded with all this nonsense. Ultimately it's a cable -- we want it to be universal, and fast. Nothing complicated there. Sadly the USB standard seemed to gain the most traction in the market despite it's fractured flavors/versions.

So if it ends up being eSATA, Firewire, or USB 3.0 or something else, I hope they just make it simple and fast. It's a cable - it shouldn't have compatibility problems or be used to confuse users with marketroid speak.

Re:That speed comes at a cost

[Vihai]

USB is PIO and not DMA? You understand that PIO/DMA transfer modes only meaningful for Parallel ATA devices?

That's because the ATA interface was originally THE 16-bit system bus (AT bus) and the disk controller was onboard on the disks (thus the name IDE - Integrated Drive Electronics).

So, the CPU accessed the disk controller through the AT bus which was originally programmed I/O and then started using DMA.

Of course, the system bus quicky become different and faster (FSB + PCI) thus the ATA interface became a disk attachment interface instead of the system bus and an additional controller was put between the system and the disk.

USB is a completely different beast. The "bus" actually transfer packets (URBs) and all USB controllers use DMA to transfer URBs to the main memory... So, no PIO/DMA stuff is involved....

Re:That speed comes at a cost

[Surt]

USB has basically zero compatibility problems that i've ever seen. There are 3 port varieties (for small, medium, large devices), and you have to match that properly (or be really strong). There are 3 protocols (1.0, 2.0 low speed, 2.0 high speed), and you get the fastest that both sides of the cable support.

The disastrous mistake the USB forum made was to allow 2.0 low speed to exist. They should have just bitten the bullet and said that 2.0 labeling is for high-speed only. But they had too much 2.0 low speed product in the channel that didn't want to be labeled 1.0, waaah waaah i want to put a 2.0 label on my product!

But still, in spite of the poor labeling, there has been zero incompatibility that I or anyone I know has experienced. Some 2.0 devices work slower than you would expect (because they are low-speed), but they WORK.

Hopefully with 3.0 they won't make the same mistake, and will only allow labeling with 3.0 for devices that use the full speed link.

Re:That speed comes at a cost

[Hal_Porter]

Given that USB is PIO and not DMA,

Bzzt! Wrong.

Here's how it works. A USB device driver wants to read from a device. It calls the host controller driver with the type of the frame (IN) it wants to send, the device address, endpoint and a address of a buffer for the received data. The host controller driver links a transfer descriptor into a structure in memory containing all this information. Some time later the host controller reads the structure using bus master DMA. It reads the frame info out of the structure, and sends the frame over the bus. The device responds and the host controller DMAs the result on the fly into the structure and signals and interrupt. Then the host controller driver unlinks the transfer descriptor and returns back to the USB device driver. No PIO is involved, and no buffers are copied.

It's all designed to avoid loading the CPU down with interrupts too.

Re:That speed comes at a cost

[sjames]

Not quite. USB devices are polled. That is, they have no way of initiating a transaction with the host, they can only respond to a transaction.

The USB controller in the host does DMA and interrupts. There is nothing in the USB spec that precludes a controller automatically generating periodic device polls and firing an interrupt when the device returns the wanted status. Naturally, that would cost more.

For USB 3.0, it may become worthwhile to develop such a controller.

Re:That speed comes at a cost

[commodore64_love]

>>>then something where they claimed you'd get 76 virgins for using USB.

I have 76 *movies* of virgins on my USB hard drive. Does that count?

Re:That speed comes at a cost

[sjames]

Fortunately, thus far, all USB devices are backward compatible. Naturally, you get best performance if any hubs you use are at least as fast as your PCs USB ports, but even if not, the devices will work, they'll just go slower.

Re:That speed comes at a cost

[Anonymous Coward]

are YOU my virgins? I hope not! If this is paradise, I've been screwwwwwed!

Re:That speed comes at a cost

[xwizbt]

Do you know, I just read all the replies to your comment and I noticed no-one dared to say 'I love you'? Because, really that's all I have to say. I agree totally. Your post was perfect in every way, and I want to have your children. Thanks, and before you ask 'No, there's no sarcasm.' Seriously. This man gets it. One USB that works for everyone. That's all we want. I still plug my 2.0 camera into my keyboard. I have to wait a little longer for the photos, but they get there...

Re:That speed comes at a cost

[Waccoon]

I don't know the real cause of the problem, but I do recall that many high-speed USB devices, such as scanners and digital cameras, will only work from a root port, and not a hub.

Any technical input on this? It drives me nuts when people try to plug things into any brand of hub and the device often won't even be recognized, but plug it directly into the computer, and it will work just fine.

Re:That speed comes at a cost

[frehe]

Ultimately it's a cable -- we want it to be universal, and fast.

It's not just a cable, but a obvious male created penis metaphor for the patriarchical oppression of women worldwide during all ages, created just to mentally abuse us women even more by reminding us of "our place" every time we use a computer. And of course you want it to be "universal" and "fast", just like all men want women to be generic models of the unrealistic Barbie stereotype, instantly ready for some quick sex with the only goal being selfish gratification for the male. All men are pigs! Snivel... sob... where's my chocolate...

Re:That speed comes at a cost

[barzok]

No, you have 76 movies of "virgins" on your USB hard drive.

Re:That speed comes at a cost

[rvw]

Ultimately it's a cable -- we want it to be universal, and fast. Nothing complicated there.

So if it ends up being eSATA, Firewire, or USB 3.0 or something else, I hope they just make it simple and fast.

I don't know about eSATA, but Firewire is more than a cable. It has its own cpu which handles the traffic. This makes Firewire 400 faster than USB 2.0 (480), because USB requires CPU time. And thus Firewire needs more space, and it's more expensive (which is probably not a big problem). The space is problematic in USB sticks, and even Apple dropped FW on the iPod for this reason.

Re:That speed comes at a cost

[toddestan]

Are you sure it's not a power thing? Unpowered hubs aren't capable of providing the full 500mA of power to attached devices. If a device requests more power than the port it is connect to says it can provide, the device won't come on. You may want to try some powered hubs with those devices.

Re:That speed comes at a cost

[Anonymous Coward]

I have 76 movies of "virgins"on my USB hard drive. Does that count?

Fixed that for you.

Re:That speed comes at a cost

[Xaoswolf]

missed the point...

All we see as an end user is a cable. We plug one end into the peripheral, and the other into the PC.

Doesn't matter what happens inside the case, we just want a cable

Re:That speed comes at a cost

[228e2]

If I had mod points, I would rate this insightful.

Re:That speed comes at a cost

[timmarhy]

are you sure about that? i remmeber writing applications for usb scanners that worked by recieving a notification from the scanner. unless of course the api i was using was a smoke screen that really just polled the deivce a lot?

Re:That speed comes at a cost

[sjames]

Either the userspace library or a driver was doing the polling for you and making it look like notifications. I'm sure because I wrote a USB driver for a firmware bootloader. It talked directly to the UHCI.

Re:That speed comes at a cost

[Rockoon]

That 5Gbps speed isnt even very attractive now. With SSD-based drives approaching the very maximum throughput that USB 3.0 is expected to offer, this pretty much cripples high speed SSD's potential as an external plugin device (yeah... eSata, gimme a break)

They should have worked with manufacturers to come up with an effective strategy to grow into a large bandwidth overshoot, such as 32Gbps, which is essentialy the bandwidth of PCIe 1.0 with 16 lanes, the kind of data rates nearly every motherboard manufactured in the last 5 years can already handle. Lower cost devices wouldn't have to spit out such bandwidth, so essentialy the cost would be on the motherboards supporting such an aggressive standard (such thinking worked for PCIe.)

Re:That speed comes at a cost

[Khyber]

I've seen PLENTY of compatibility problems.

480megabits for USB 2.0, yea?

Why does a LifeCam VX-3000 and Freetalk Wireless Headset somehow eat up ALL of that bandwidth? If the cam's on first, if I try to use the headset in skype I get "There is not enough bandwidth" in the lower right corner.

What the hell is up with that? EVERY port on my machine is USB 2.0

Only things hooked up to USB *RIGHT NOW* are:
Logitech MX wireless laser mouse
Compaq USB wireless keyboard
Lifecam VX-3000
Freetalk Wireless Headset.

Both wireless mouse and keyboard are on the front ports. Cam and headset are on the back.

Re:That speed comes at a cost

[Waccoon]

Yes, I'm sure these are self-powered hubs. I have this problem myself with my digital camera, and I've tried using the powered USB ports on my power switch box. I can charge my PS3 controller with my computer off with no problem, but my scanner will only scan less than half an image before locking up.

I used to work in a camera store, so I had the chance to mess around with a lot of equipment, and helped numerous people with this problem. Every time someone called the store and told me the equipment they bought wasn't working, I asked them to plug the device directly into the back of the computer. Almost all the time, it worked. Maybe it's just an issue with cheap, badly designed USB controllers, and I assume there's plenty of those around.

Re:That speed comes at a cost

[markhahn]

USB is not PIO.

Re:That speed comes at a cost

[jnork]

Hopefully with 3.0 they won't make the same mistake, and will only allow labeling with 3.0 for devices that use the full speed link.

Then they'll just label it "3.0 compatible" or some such.

Wouldn't be the first time.

Re:That speed comes at a cost

[StuartHankins]

I've had this issue with certain HP Laserjet printers (1320 & 2300 for instance). Won't work from a powered hub, will work directly through the computer's USB port.

Re:That speed comes at a cost

[jnork]

All we see as an end user is a cable. We plug one end into the peripheral, and the other into the PC.

Doesn't matter what happens inside the case, we just want a cable

Speak for yourself.

I'd like to see more take-up of Firewire, because I prefer having more performance. And I don't like this trend of saving money by loading down the CPU. What's the point of having dedicated peripheral processors if the first thing they do is dump on the PC?

Re:That speed comes at a cost

[Anonymous Coward]

The speed terminology of USB is indeed pretty confusing, so let's be clear:

- "Low-speed" (1.5 Mbit/s) is the lower of the original USB 1.0 and 1.1 speeds. This is used for devices that transfer very little data, such as mice and keyboards. Also a USB 2.0 peripheral can choose to support only this speed.

- "Full-speed" (12 Mbit/s) is the higher of the original USB 1.0 and 1.1 speeds. Also a USB 2.0 peripheral can choose to support only this speed.

- "High-speed" (480 Mbit/s) is the new speed added in USB 2.0. Any USB 2.0 host is required to support this speed, but for peripheral it is optional (peripheral can choose Full-speed instead). That optionality is what allows the USB 2.0 peripherals that are still as slow as in USB 1.x days. The "low speed" 2.0 devices that you refer to are actually Full-speed devices in the USB terminology.

- "SuperSpeed" (5 Gbit/s) is the new speed added in USB 3.0.

Re:That speed comes at a cost

[Enleth]

There actually is a difference. 2.0, high-speed or not, is much easier on the CPU than 1.0 and 1.1 because it allows for a device to wake up the host, instead of the host having to poll the bus every x miliseconds, preventing the CPU from switching into a deeper C-state and saving power.

Re:That speed comes at a cost

[Shamenaught]

I once heard that modern webcams are very wasteful compared to older ones. Back when USB bandwidths weren't sufficient to transfer a stream of bitmaps, they used to compress the images on the camera and transfer the compressed stream.

Nowadays, because transfer speeds are higher, it means they can cut hardware costs by transferring 30 uncompressed bitmaps per second and using the PC's CPU to do the encoding. This means more of the USB bandwidth is taken, and your computer has to do a lot more number-crunching.

Doing some math, assuming a 30fps 640x480 stream in 24bit colour: 30x640x480x24 = 221,184,000. That's almost half of the 480Mb already. I don't know for certain if webcams do the same as other digital cameras, and record at more than 24bpp to provide some range for light adaptation, but that'd take-up even more bandwidth.

Add overhead on top of that, more cutting hardware costs by shifting control to the computer over the high-speed link, more wasteful programming by all parties involved, and I wouldn't be completely surprised if all of that bandwidth could be used.

Re:That speed comes at a cost

[Detritus]

PIO and DMA have been around for over 50 years. That's a bit longer than the ATA interface.

Re:That speed comes at a cost

[drinkypoo]

That's not a compatibility problem. That's a "USB is shit" problem. Personally, I have a laptop which has USB 1 and USB 2 controllers (Intel 82801G chip...) and the system apparently connects the device to the appropriate controller - plug in a USB1 device, it gets hooked up to the USB1 controller, unplug it and plug a USB2 device into the same port, it gets hooked up to the USB2 controller. This is probably the best way to mitigate the problem given the fact that USB sucks. The simple truth is that it has more overhead because it's dumber than IEEE1394. Another truth is that a firewire controller is CHEAP (even for PC Card) and USB2/Firewire enclosures are not much more expensive than USB2-only types. It costs NOTHING more to get IEEE1394 alongside USB2 and eSata on a WD MyBook. I am using it right now, to make an old IBM Thinkpad A21p into a fileserver for a 1TB WD MyBook. I have a Margi-branded TI 1394 PC Card adapter I got at Halted for $5 or so :)

Re:That speed comes at a cost

[drinkypoo]

You do realize that people were doing Programmed Input/Output and Direct Memory Access before the Industry Standard Architecture bus was invented, right?

Re:That speed comes at a cost

[drinkypoo]

All men are pigs! Snivel... sob... where's my chocolate...

In my pants.

Re:That speed comes at a cost

[cthulhu11]

The concepts of PIO and DMA transfers are much older than PATA disks, eg. they were issues back in the Unibus PDP-11 days. IIRC the DZ11 did PIO and the DH11 did DMA.

Re:That speed comes at a cost

[CarpetShark]

said that 2.0 labeling is for high-speed only. But they had too much 2.0 low speed product in the channel that didn't want to be labeled 1.0

Yep, the problem is that some devices don't, and will never be able to utilise a high speed link. Standard keyboards, for instance. What they really need is some kind of label that means "Uses the a USB version that allows full throughput". Call it USB Unlimited, or something. That way, a keyboard can be USB Unlimited, and the best HD cams can be USB unlimited, but the cheap-ass model you get free with your netbook clearly won't be.

Re:That speed comes at a cost

[code4fun]

Not sure what you mean. I wrote a USB Ethernet driver a few years back and I don't recall doing programmed I/O. The chip handles DMA. You set up request blocks and the chip handles the data transfers. You're not using processor cycles to get data in/out of the chip in that respect. You only need to feed buffers to the chip and keep it busy.

The only thing with USB is that there is a significant overhead associated with USB protocol. If I recall, it was something like 30% of the bandwidth. That was USB 1.1 with 11 Mbps. I haven't looked at USB since.

Re:That speed comes at a cost

[Anonymous Coward]

amen! I hate the fake integrated audio which they sells on today motherboard, which inists on offloading all the dsp on the cpu.

Re:That speed comes at a cost

[badkarmadayaccount]

"Vista Compatible" v 2.0

And how much cpu power is needed at that speed?

[Joe The Dragon]

And how much cpu power is needed at that speed? firewire 1600 and 3200 seems better same cables and ports as firewire 800 unlike usb 3.0 that needs new cables and ports for usb 3.0 speed.

Re:And how much cpu power is needed at that speed?

[Daniel Dvorkin]

I hate to say this, but Firewire's dead. Apple invented it and they've been the main ones pushing it; now that they're pretty clearly planning to get rid of it, there are no major industry players with an interest in its survival. I agree that it's a far superior standard for pushing any meaningful amount of data around, so I'm not at all happy about this state of affairs, but so it goes.

Re:And how much cpu power is needed at that speed?

[CarpetShark]

And how much cpu power is needed at that speed?

CPU usage is fixed in all USB standards. It's 140%.

Re:And how much cpu power is needed at that speed?

[TechForensics]

I hate to say this, but Firewire's dead.

Oh boy are you wrong about that. Firewire, unlike USB, does not suck up scads of CPU cycles, the more the faster the transfer. In our shop doing server backups (FreeBSD) to external drives there was significantly less impact on server responsiveness when using firewire drives.

And yes, there was a weird reason why we needed to do backups of live filesystems.

Re:And how much cpu power is needed at that speed?

[lysergic.acid]

USB is cheaper, and not everyone needs FireWire's level of performance, so USB is more prevalent and dominates the casual computing market; that's only natural. but most motherboards sold these days still come with at least 1 FireWire port.

and while FireWire can technically replace USB, USB will never be able to replace FireWire. even though the average user doesn't need sustained 100MB/sec transfer speeds to transfer their MP3s onto their iPod or text documents onto their thumbdrive, there are a lot of professions where USB just won't cut it. if you're in multimedia production or otherwise need to transfer large amounts of data regularly, then the extra cost of FireWire is more than worth it.

high-end external hard drives, cameras, professional audio equipment, etc. will all continue to use FireWire for this reason. heck, IEEE 1394b is even used by NASA for monitoring launch debris and by the U.S. military in jets like the F-22 and the F-35. so FireWire is far from dead. it's just found its niche. at the very worse, users will have to buy expansion cards to add FireWire S1600/S3200 controllers to their computers. but FireWire will continue to be available for a long, long time.

Apple's decision to remove FireWire from their low-end systems is just a sign that they're no longer catering exclusively to the prosumer/media-production crowd. they're still selling systems with FireWire, but they're also recognize that the Mac-using demographic has changed over the years, and there are a lot of Mac users that will never need FireWire.

Re:And how much cpu power is needed at that speed?

[peragrin]

um if not everyone needs firewqire level performance why does USB try getting faster. why not simply use firewire for drives, and USB for simple things like keyboards, mice and monitors?

Re:And how much cpu power is needed at that speed?

[lysergic.acid]

that's like asking why cars keep getting faster when not everyone needs a Ferrari.

faster transfer speeds are always better, and as technology moves forward faster and faster speeds will be available for lower and lower costs. but there will always be a disparity between each generation of FireWire and USB because USB is in a lower cost bracket.

so far all USB versions have continued to rely on the host processor to manage most low-level USB operations. the emphasis continues to be on lowering costs. that's why Intel's UHCI controller interface is primarily software-driven to lower costs.

FireWire OTOH uses more expensive interface hardware to offload tasks from the CPU. this comes at a higher cost, but it will always provide more throughput and use less CPU resources than than the USB approach. so it doesn't matter how faster USB gets in future generations, this difference in design will continue to create a disparity in the real-world performance of the two bus families.

Re:And how much cpu power is needed at that speed?

[commodore64_love]

>>>firewire 1600 and 3200 seems better same cables and ports as firewire 800 unlike usb 3.0 that needs new cables and ports for usb 3.0 speed.

This argument is non-persuasive. Everything I own is USB, therefore to use Firewire 1600 or 3200 I'd have to not only get new cables, but new devices. Pass. I'd rather stick with USB thank you very much.

Re:And how much cpu power is needed at that speed?

[Anonymous Coward]

even though the average user doesn't need sustained 100MB/sec transfer speeds

USB 3 will do at least 200mb/s sustained. And home users will love it when external harddrives gets faster. Professionals may need firewire for other stuff but the measly 100mb/s will not be an argument for firewire.

IEEE 1394b is even used by NASA for monitoring launch debris

And every computer at NASA uses USB for the mouse, it doesn't mean shit. 1394b is an interconnect system. It can't track launch debris.

Face it, firewire is dead. It's only been used for DV by consumers. Now DV is gone and firewire will soon follow as far as consumers are concerned.

Re:And how much cpu power is needed at that speed?

[Anonymous Coward]

The security (or lack of it) on firewire with its peer-peer design and any device being able to read and write memory on other devices should be the death of firewire though.

Re:And how much cpu power is needed at that speed?

[commodore64_love]

Digital8 and DV camcorders are going to the same place as VHS - obsolete. So your "my camcorder uses firewire" is not really helping support it. All newer camcorders like the Hard Drive Everios are using USB.

Re:And how much cpu power is needed at that speed?

[Waxwing Slain]

I hate to say this, but Firewire's dead.

Don't tell me that. I just dropped 3 Gs on a Digidesign 003 deck for my DAW. Actually, maybe Firewire is dead, but I'm betting IEEE 1394 will have a long productive life, at least in my studio. I do enjoy people who pronounce the death of a particular technology, usually with their index finger pointing skyward and in a deep basso profundo. Shitfire, I still know people who are getting work done on Nubus Macs.

Re:And how much cpu power is needed at that speed?

[WillyDavidK]

clearly you don't understand his original point. Just because a protocol claims a higher maximum transfer rate does NOT mean better performance. Anyone who has actually used usb and firewire drives or high-bandwith peripherals side by side can attest to the much better performance of firewire devices. And firewire is still very essential in the pro audio industry.

Re:And how much cpu power is needed at that speed?

[0100010001010011]

Planning to get rid of it... by putting it on the higher end laptops just released.

Firewire may be dead to the home consumer, but just like BetaMax, it's likely to live on in the pro-sumer and professional world.

Re:And how much cpu power is needed at that speed?

[lysergic.acid]

USB 3 will do at least 200mb/s sustained. And home users will love it when external harddrives gets faster. Professionals may need firewire for other stuff but the measly 100mb/s will not be an argument for firewire.

when USB 3.0 can do 200MB/s (not sustained), FireWire S3200 will have sustained speeds of 400MB/s. and it's not faster external hard drives that will push consumers to upgrade from USB 2.0 to 3.0--current UDMA133 hard drives are already far outstripping the 33MB/s transfer speed USB 2.0 is capable of providing. it's the ever increasing disk sizes, especially in portable media players, along with the proliferation of HD video, hi-res cameras, lossless audio, and other applications exposing consumers to ever-larger file transfers, that will increase demand for faster bus interfaces.

And every computer at NASA uses USB for the mouse, it doesn't mean shit. 1394b is an interconnect system. It can't track launch debris.

i know you're just trolling, but if every computer at NASA uses USB--whether it's for mouse/keyboard/printer/scanner/whatever--then that clearly means USB isn't going anywhere. i mean, FireWire is a high speed serial bus. of course it's being used as an interconnect--in this case to connect debris-monitoring equipment. what did you expect them to use it as? a CPU? NASA and the military use IEEE 1394b for high-speed interconnects because it's the best solution. IEEE 1394b has been standardized by SAE AS5643 as a data bus network for use in future military & aerospace projects (such as the Orion crew exploration vehicle) that require a high throughput data bus. USB doesn't even come close to the same performance.

Face it, firewire is dead. It's only been used for DV by consumers. Now DV is gone and firewire will soon follow as far as consumers are concerned.

yea, you can repeat that as many times as you want, but that won't make it true. FireWire's current applications extend far beyond the DV format. as long as most consumer laptops and computers still support FireWire, it's not considered dead, especially as there are no viable alternatives to FireWire for sustained high speed data transfers.

clueless armchair analysts have been predicting the death of FireWire since USB 2.0 came out. but anyone who's actually worked with both interfaces or is media production (or understands the difference between PIO and DMA) knows that USB's real-world performance doesn't even compare to that of FireWire.

Re:And how much cpu power is needed at that speed?

[Anonymous Coward]

Not surprising given your handle.

Re:And how much cpu power is needed at that speed?

[ion.simon.c]

*nods* LSD has this on lockdown.

Unless USB 3.0 offloads all of the protocol work from the CPU to the USB controller and attached devices, FireWire will still have its place as a relatively high-performance interconnect.

Re:And how much cpu power is needed at that speed?

[Frankenshteen]

I agree this is most disturbing - loosing firewire. What can the community do to save it, propagate firewire, compel manu's to include the hard port on devices?

Re:And how much cpu power is needed at that speed?

[WuphonsReach]

Planning to get rid of it... by putting it on the higher end laptops just released.

I think a lot of the griping about Apple abandoning Firewire comes from the MacBook Air which didn't have a Firewire port (but had USB ports).

Re:And how much cpu power is needed at that speed?

[necro81]

Usefulness or necessity aside, the main reason that USB is angling for higher speeds is that there's money to be made there, and they want to be the ones to make it.

Re:And how much cpu power is needed at that speed?

[noewun]

USB 3 will do at least 200mb/s sustained.

And Vista is the fastest, easiest to use operating system ever, and 2009 is the year of Linux on the Desktop.

USB 2.0 was supposed to do 480 Mbps sustained, but you're lucky if you get 20 MB/s sustained. Given Intel's track record, I expect USB 3.0 to, maybe, be as fast as Firewire 400 for sustained transfers. There's a reason my TB backup drive is Firewire 800.

Re:And how much cpu power is needed at that speed?

[Anonymous Coward]

Using a dead transport on a dead OS? You might as well call your server room "The Graveyard".

Re:And how much cpu power is needed at that speed?

[TrancePhreak]

Only USB in the Air and the MacBook. Only the Pro has firewire.

Re:And how much cpu power is needed at that speed?

[martinX]

You may like to wait, but FW is deader than a dead thing.

Two things FW gave you that USB didn't when it came to DV. Control of the camera's mechanism (so you could seek for the spot on the tape you wanted to capture from) and isochronous capture so you could be assured of capturing the audio in sync with the video.

Video cameras now are going to card-based storage and video files are just that: files. This means you can access the storage via USB just like it's another external drive and you don't need to be able to control the camera. The ability to grab at file at random means you just copy all the data over, not stream it.

A lot of pro equipment still relies on firewire, but that, too, is changing. It won't change for a while because professionals tend not to change their equipment just because they read about something really cool on Engadget. For the pros, there's FW cards. Audio guys may still need the low latency that FW provides, but that may change with USB3.

Apple was a bit late to the (USB camera interface) party but the release of iMovie 8 (crappy though it was) was a sign - the future is AVCHD files grabbed over USB.

FWIW, I have a FW camera at home, two at work (just bought one 6 months ago), a FW deck at work (the 3 cams and the deck are Sony) and an analog -> digital converter with a FW interface.

I'll keep using FW at work for the foreseeable future because it's solid. And I have all that equipment. My next camera at home, though, will be AVCHD stored on SD cards with a USB interface.

I will lament the passing of FW like I lamented the passing of ADB, SCSI and parallel ports: part "meh" and part "damn, I have to buy new stuff to go with my new stuff".

Re:And how much cpu power is needed at that speed?

[Anonymous Coward]

faster transfer speeds are always better

Yeah right... tell that to my wife...

Re:And how much cpu power is needed at that speed?

[Anonymous Coward]

Apple's decision to remove FireWire from their low-end systems is just a sign that they're no longer catering exclusively to the prosumer/media-production crowd. they're still selling systems with FireWire, but they're also recognize that the Mac-using demographic has changed over the years, and there are a lot of Mac users that will never need FireWire.

Apple is trying hard to differentiate their "pro" and "consumer" product lines - they've been doing that essentially since the original iMac. For the previous MacBooks, it was the glossy screen and the plastic case, Since the new aluminium MacBooks just look too much like the Pro, they have to be crippled in some other respect. There just has to be something to justify the steep markup for their "Pro" offerings.

Re:And how much cpu power is needed at that speed?

[Anonymous Coward]

I hate to say this, but Firewire's dead.

Sure whatever.

Oh wait all High end video camera and production gear comes with firewire on it. Must be that Video gear is dead!

Just like the claim that SCSI is dead for the past 5 years..... Oh wait all real servers use SCSI.. well lookie that! how silly!

Re:And how much cpu power is needed at that speed?

[commodore64_love]

>>>Not surprising given your handle.

Ha! Well just think of all the thousands of dollars I saved by Not buying the next best thing. My IBM-compatible PC is a 4000 megahertz Pentium 4 that's going on 6 years age. Other people might have thrown it in the trash and upgraded 4 times by now, but I did not. And that's saved me around $7000 of cash.

I don't believe in throwing away perfectly good stuff. Which is why I still have USB drives and Super VHS-C camcorders, rather than "upgrade" to Firewire 1600/3200 gadgets. It saves money. Which means I can work fewer hours.

And most importantly: Take time to enjoy life.

Re:And how much cpu power is needed at that speed?

[Jorophose]

FireWire is used for cable boxes.

Shit ain't dead yet, and you better hope it stays this way.

Re:And how much cpu power is needed at that speed?

[peragrin]

two points.

Cars aren't getting faster. More efficient but the every day car doesn't go a lot faster now than 20 years ago. It does so on better mileage, but that can be attributed to weight loss.

As another poster was pointing out USB can spike your CPU when doing large file transfers, at high data rates. why not use the slightly more expensive firewire for drives, as you can transfer files faster with less CPU So you can do more than one thing at a time? It is my biggest concern with USB 3.0 how will a notebook CPU handle the processor requirements if it is anything like USB 2.0 only faster.

Re:And how much cpu power is needed at that speed?

[dgatwood]

That argument is also non-persuasive. Your computer will not cease to support USB devices merely because you chose to buy a device that uses a different standard. Given the fact that most keyboard these days are USB, it is safe to say that USB will not die out any time soon. That doesn't make it a particularly good way to connect a hard drive, though, inasmuch as gasoline's ready availability does not make it a good home heating solution. You can do it, but if your needs are too great, it can really blow up in your face.

Re:And how much cpu power is needed at that speed?

[commodore64_love]

I still don't see why I would want to buy a FireWire device. My USB 2.0 drive streams movies to VLC Player just fine. There's no need to change.

Re:And how much cpu power is needed at that speed?

[lysergic.acid]

maybe not by a lot, but certainly by a measurable amount. even an '09 Honda Accord is just a bit faster (+2 HP) than an '08 of the same class; and the '08 is faster (+24 HP) than the '07; the '07 is faster (+9 HP) than the '06; so on and so forth. these incremental improvements exist between generations of almost every car, if only for the reason that the same technology/performance gets cheaper to produce with each passing year. heck, this trend can be seen in nearly any product, regardless of whether it's a car, a video game console, or a web service.

and what you suggest is already being done by a lot of people. that's why most computers come with both IEEE 1394 and USB. many people use FireWire for hooking up their external hard drives, digital (cinema) cameras, audio production equipment, etc. and use USB for keyboard, mouse, printer, and everything else. USB 3.0 isn't likely to change this.

Re:And how much cpu power is needed at that speed?

[Anonymous Coward]

Anyone expecting to see a FireWire S1600/S3200 controller in the future is living in fantasyland. Apple has effectively withdrawn support for future FireWire speeds. I know this because my company invested 2 years and millions of dollars developing a FireWire 1600-to-PCIe bridge controller and it was rebuffed by Apple. Without Apple commitment to design-in S1600 or S3200 FireWire ports, the technology will die a slow death until USB3 is mainstream. The fastest FireWire ports/peripherals you will ever see will be 800Mbps (S800 mode), I guarantee! On the other hand, prototype USB3 devices are already showing 3600 Mbps (450 MB/sec) which is more than 4x faster than FireWire 800.

Re:And how much cpu power is needed at that speed?

[Anonymous Coward]

even an '09 Honda Accord is just a bit faster (+2 HP) than an '08 of the same class; and the '08 is faster (+24 HP) than the '07; the '07 is faster (+9 HP) than the '06; so on and so forth.

Dude, HP != speed.

- T

Re:And how much cpu power is needed at that speed?

[lysergic.acid]

so because Apple didn't want to buy some anonymous company's untested FireWire controller, that means they've withdrawn support for future FireWire speeds? right...

Apple isn't in charge of developing IEEE 1394b. IEEE has already approved the specs for S1600 and S3200, and Symwave has already developed an S1600 Physical Layer prototype, the FirePHY-1600, which is fully backwards compatible with S800 and S400 and is expected to cost the same as current S800 PHY chips. so there is plenty of incentive for manufacturers to switch to S1600 as soon as the new S1600 chips start hitting the market.

plus, it doesn't matter if Apple doesn't include S1600 controllers in their computers. using an S1600 cable + an S1600 device, you can still achieve the 1.6 Gbps transfer speeds on FireWire 800 ports. heck, even if computer/motherboard manufacturers suddenly stop including FireWire as a standard interface (which is not likely to happen) the people who use FireWire on a regular basis, and depend on it for their day-to-day work, would in all likelihood be more than willing to shell out $30-40 to pick up a PCIe FireWire adapter.

USB was designed primarily with the aim of providing a low cost peripheral interface. FireWire was designed for performance and thus fills a separate niche. it would be nice if i could believe that USB 3.0 will deliver the transfer rates it promises, but unless USB has switched to full DMA it will come nowhere near its max theoretical speeds. and Apple isn't going to drop FireWire so long as FireWire remains the only viable option for professional users needing high speed data transfers.

Re:And how much cpu power is needed at that speed?

[Anonymous Coward]

I don't understand why you call FreeBSD a dead OS.

Re:And how much cpu power is needed at that speed?

[martinX]

Not where I come from :-)

Anyway, there'll always be areas where one protocol or another is kept on long after it fades from general view. The whole "USB 3 is king" thing is just about connecting peripherals like HDDs to desktop PCs.

Speed

[Anonymous Coward]

Great, so now we have low speed, full speed, high speed and maximum speed, honest this time?

Re:Speed

[eln]

USB 3.0 is known as "ludicrous speed".

Re:Speed

[Anonymous Coward]

No no, ludicrous speed!

Re:Speed

[jlechem]

They better be careful, they don't want to go "plaid"

Re:Speed

[Anonymous Coward]

Here's what it might look like

Re:Speed

[Anonymous Coward]

Actually, the official term for the new data rate is SuperSpeed.

Crap

[Idiomatick]

According to the summary USB 2.0 is already legacy tech. Excuse me while i go to the corner and cry for being horribly outdated.

Re:Crap

[iNaya]

Oh crap, I'm still using USB 1!!

(Actually my computer died, and I had to pull an old one out of the closet).

Re:Crap

[neokushan]

I didn't know computers could have sexual preferences. So THAT'S why I only seem to get lesbian porn popups...

Re:Crap

[iNaya]

Yes, well most computers don't like any kind of physical interaction with humans, from my post, it is quite apparent that I found one that rather enjoys it ^^

Re:Crap

[neokushan]

I'd just like to point out (at the cost of Karma, probably, but fuck it) that I love the mods on slashdot. Only the /. community would be crazy enough to mod the above comment as "informative". Thank you, mods, for giving me a giggle.

eSATA is here already

[camperslo]

With many chipsets/motherboards already supporting SATA and the drives being widespread as well, many of US could enjoy much (about 6x ?) better than USB 2 port speeds for external drives by simply having the external ("e" in eSATA) connectors available.

Re:eSATA is here already

[martinw89]

Yeah, I don't need a faster USB for my keyboard and mouse. Even my flash drive probably won't benefit without faster memory. eSATA seems like a better solution for the widespread use of external HDs. Plus, it doesn't have to go from USB-SATA in the external HDs controller.

Re:eSATA is here already

[Strange Ranger]

eSATA is great for external drives that stay connected and turned on. But for removable (i.e. flash) drives they can be a pain. Every time you pop a card in or out and then reboot the BIOS makes you redefine Boot Order, eSATA drives are just like regular SATA drives, not a "removable device".

As a photographer who unloads about 20-30GB of raw files every week from CF cards in multiple readers, I'm pretty excited about USB 3.0.

Re:eSATA is here already

[simcop2387]

http://en.wikipedia.org/wiki/SATA#Features Actually SATA is supposed to support hotswapping of drives, i've never actually tried though.

Re:eSATA is here already

[neokushan]

I can confirm it definitely does, but Windows (not sure about other OS's) is a little picky about when it decides to detect if that device has been connected or not.
It's not a major issue, a quick device manager refresh is all you need to do, but it's still an extra step current USB drives don't have.
Of course, it's probably windows' fault more than anything, but still.

Re:eSATA is here already

[Anonymous Coward]

This def. sounds like an issue with your particular BIOS, not with eSATA in itself.

Re:eSATA is here already

[nabsltd]

Hot-swapping of SATA works just fine (at least for me).

There are some SATA controllers that specifically report the attached drives as being "removable", and these are easiest to work with, since any modern OS will optimize for removal.

I have a cheap add-in SATA card where the drives don't show up as removable, and all I do is make sure I manually flush the disk buffers (using "sync" from SysInternals) before removing it. I have the drives for this card hooked up using this Kingwin hotswap bay and have had no issues whatsoever in Windows 2003 Server.

Re:eSATA is here already

[Britz]

I was wondering the same thing. Why not just use eSATA? Can MP3-Players and Digital Cameras get this Chip? And why no eSATA flash memory modules?

Re:eSATA is here already

[qupada]

Had you considered Sandisk's (I believe there are other manufacturers with similar devices, but I can't remember which right now) Firewire CF reader?

There is something about the simplicity of the single-slot card reader and nice fast FireWire 800 interface, rather than some lambda-in-one usb monstrosity. My only complaint is it has no daisy-chaining port, the bus certainly has bandwidth to spare.

I have a certain curiosity about this card reader too, no sense in laptop users being left out of the fast card reader party (providing power to the Sandisk FireWire reader is the issue here - Apple's laptops are the only models I can think of with powered FireWire ports). I'd hope most operating systems these days wouldn't spit the dummy when presented with a hot-plug PCI-e device, but you never know.

Re:eSATA is here already

[upuv]

Like others have mentioned. eSATA is the annoying spec.

1. eSATA is just like SATA. The OS does like to treat devices as removable. Often resulting is a manual reboot of OS to free drive.
2. The eSATA spec didn't have power on the connector! What were they thinking?
3. The physical connector format is not as robust as the usb. Highly prone to just plain wearing out.

When I first heard about eSATA I was very excited. As firewire already had it's gravestone made. Finally a method of transferring huge files fast from a portal device. Only to find out that it was a pointless useless spec.

Re:eSATA is here already

[DiegoBravo]

But faster USB = less holes, less chipsets, less price, less drivers, less trouble... for future machines.

Re:eSATA is here already

[petermgreen]

Esata has two main advantages over interfaces like USB and firewire.

* Fast
* No bridge board needed at the drive end

And several major disadvantages

* Doesn't carry power
* less common than either USB or firewire (and not backwards compatible with either)
* Only one level of port multipliers allowed and some controllers don't even support that.
* While in theory it should be possible to make a wide variety of eSATA devices the only devices on the market seem to be hard drives and optical drives
* Neither eSATA ports nor anything they are backwards compatible with are very common (I think i've seen them once or twice on a desktop and never on a laptop). You can add them easilly enough to a desktop with an adaptor bracket but that means getting permission to open the case and using an adaptor bracket rather than a port intended to be eSATA may cause hotplugging issues.

So I think it's destined to remain a niche interface.

Re:eSATA is here already

[lysergic.acid]

if you're transferring that much data with a card reader, it might be worthwhile to get one that supports FireWire. even a FireWire 400 CompactFlash card reader in PIO mode beats out a USB 2.0 card reader in UDMA mode. but a FireWire 800 CF card reader in UDMA mode absolutely smokes USB 2.0 in UDMA mode.

Synchrotech has performed some FireWire vs. USB 2.0 UDMA CompactFlash benchmarks. this is their conclusion:

While USB 2.0's theoretical 480Mbp/s (60MBp/s) throughput should be sufficient for UDMA 4 CompactFlash, real throughput is significantly less. Top hard drive manufacturers typically cite USB 2.0's best speed at 33MB/s, or about half the speed of UDMA 4 CompactFlash, or 25% of UDMA 6 CompactFlash. There are myriad reasons for USB 2.0's 'real world' speeds including: CPU overhead from its master/slave arrangement, NRZI encoding, and inexpensive chipset implementations. The USB 2.0 UDMA reader used in the benchmarks above uses one of the latest USB chipsets from Genesys Logic. While a new generation of that chipset should soon be available, we don't foresee it providing throughput close to half of that of FireWire.

The above tests demonstrate both FireWire 800 and 400 readers are significantly faster for reading CompactFlash cards by orders of magnitude. When card to computer speed in crucial, always choose a FireWire based CompactFlash reader or a reader with a comparable bandwidth.

so if you don't want the bus interface to be a bottleneck preventing your CompactFlash cards from realizing their full performance, then it might be worthwhile to invest in a slightly more expensive FireWire 800 card reader.

of course, if you have a laptop with an ExpressCard 34/54 slot, then you could go with a CFExpressPro+ PCIe ExpressCard to CompactFlash Memory Card Adapter. though CF rev. 4 itself only supports transfer speeds of up to 133MB (UDMA 6), so any bus speed beyond 1 Gbps would be overkill.

Re:eSATA is here already

[Trixter]

eSATA is great for external drives that stay connected and turned on. But for removable (i.e. flash) drives they can be a pain. Every time you pop a card in or out and then reboot the BIOS makes you redefine Boot Order, eSATA drives are just like regular SATA drives, not a "removable device".

So don't make your primary boot device a removable drive. Make it the internal drive and you don't have a problem.

As a photographer who unloads about 20-30GB of raw files every week from CF cards in multiple readers, I'm pretty excited about USB 3.0.

As someone involved with a Blu-Ray project with 3 terabytes of footage, my life has been made much easier by eSATA. I can pop a drive into a $40 hard drive "dock", turn it on, dump footage onto it at speeds exceeding 90MB/s (5.4G/minute), pop it out, and give it to a client. And the dock even came with a SATA-to-external-bracket so I didn't need to buy anything else.

Re:eSATA is here already

[Brain_Recall]

No, not really windows, but more of a hardware issue. The SATA spec doesn't require manufatures support hot-swap, so, some don't fully support it. nVidia's SATAs always have had hot-swap, so thats why you would see the main hard drive in the "safely remove" list. But, for example, my Marvel and Intel controllers don't support hot-swap. My Marvel controller handles the eSATA ports, and I've tried adding a drive while it was powered up. Windows detected it and everything work, but there is no safe method to disconnect the drive (the write-cache buffers need to be cleared out before you can yank it off safely). The best solution I could come up was force the computer into S3 standby, and then pull the drive. That seems to work good enough.

Re:eSATA is here already

[Siffy]

USB wins in the convenience over performance game. eSATA in a SATA 3.0 package will get a little closer once the SATA controller can provide power to the device over the data cable, but by then it will probably be too late. Not having to carry around a separate power brick for each and every device you plan to plug in is an extremely nice feature.

Re:eSATA is here already

[Siffy]

Like this? http://www.ocztechnology.com/products/flash_drives/ocz_throttle_esata_flash_drive http://www.newegg.com/Product/Product.aspx?Item=N82E16820227390&Tpk=N82E16820227390 It's about twice the cost of most USB flash drives the same size and it still requires a USB connection to power the device on most computers. You get to carry around a cable to use it... Awesome!

Re:eSATA is here already

[Agripa]

So don't make your primary boot device a removable drive. Make it the internal drive and you don't have a problem.

Some BIOS implementations for whatever reason mangle the boot device order when enumeration after POST shows a change in possible boot devices on USB or SATA.

Re:eSATA is here already

[Hurricane78]

As far as I know, this information is wrong. SATA is completely hot-pluggable, and was defined as such.
Because of this, eSATA is of course hot-pluggable too.

My guess is that you should fix your OS. :)

Re:eSATA is here already

[Anonymous Coward]

But faster USB = less holes, less chipsets, less price, less drivers, less trouble... for future machines.

It isn't an either/or choice. Faster USB is fine and dandy, but since eSATA is simply using existing SATA support there's generally little extra involved besides adding the connector. The functionality is already in a multifunction chip in many motherboards. For example the Intel ICH7 chip in the desktop I built 2 1/2 years ago provided 4 SATA ports. The early Macbook of about the same vintage used the ICH7m which similarly supported the internal SATA HD (and perhaps the optical drive too?) in that machine.
Obviously driver support is already there too. USB doesn't replace SATA.

For those needing external drives that can perform as well as the internal drives do, eSATA could be very helpful and cheap too. Once you've experienced file copy between drives at close to 6 gigbytes a minute over SATA (with some of the recent 1 and 1.5 TB drives), getting 1 gigbyte a minute over USB 2 feels painfully slow. I think the speed is needed even in low-end machines. For example anyone using a Mac Mini as a PVR would easily appreciate the difference.

Re:eSATA is here already

[Lumpy]

As a photographer who unloads about 20-30GB of raw files every week from CF cards in multiple readers, I'm pretty excited about USB 3.0.

Why? you dont want to buy the right gear now? My Firewire CF reader is faster than any usb 2 reader I have ever tried, and is over 3 years old now.

My Firewire 800 Cf reader from lexar is insanely fast (as fast as the card let's it read)

I'm confused by photographers that do not buy the correct gear or even go looking for it.

Re:eSATA is here already

[Anonymous Coward]

* Neither eSATA ports nor anything they are backwards compatible with are very common (I think i've seen them once or twice on a desktop and never on a laptop). You can add them easilly enough to a desktop with an adaptor bracket but that means getting permission to open the case and using an adaptor bracket rather than a port intended to be eSATA may cause hotplugging issues.

So I think it's destined to remain a niche interface.

Huh?

My two years old Athon64x2 motherboard had a built-in eSATA connector as well as firewire. It also has a plethora of USB ports, something like 10, but some are not really accessible, especially if you fill the expansion slots, so I have only something like 6 brought out of the box.

On my 3 month old Dell laptop, one of the 4 USB ports is actually a hybrid USB/eSATA connector.
It is very frequent among Dell laptop, at least the ones considered "professional".

Re:eSATA is here already

[iphayd]

Maybe they're too busy taking pictures? That's why forums are helpful - they allow busy people to share knowledge with other busy people.

Re:eSATA is here already

[Anonymous Coward]

at 30Gb a week? not a chance. he's spending less than 8 hours a week shooting photos.

I shoot 3X that a week and I only work 12 hours a week shooting. granted, I shoot with low res 24megapixel Hassleblad and not any of the pro level cameras that produce 2X the size photos...

Re:eSATA is here already

[Anonymous Coward]

I have more card readers than firewire ports.
I can hang the readers off my monitor, or anything else that has USB ports in it.
Sandisk "Extreme" readers are much faster than other USB readers I've used.
I can take them and plug them in anywhere. Like an associate's laptop for instance.
So for me USB is the correct equipment.

I use DSLRs. Using a Hassy MF back for weddings would be idiotic.
At least half the people using MF backs are wasting thousands and thousands of dollars.
How many tack sharp pixel-peeping prints do you need to print at 40x60? Probably not too many.
A 21mp DSLR will print full bleed magazine spreads just as well.
But it sure feels "superior" using the Hassy huh?

One cable to rule them all

[Anonymous Coward]

If all these things have to

USB3 whitepaper

[whyde]

Most of the replies so far show a glaring lack of knowledge of what USB3 really is. Honestly, it only bears a passing resemblance to its predecessors, and is a closer relative to PCIe. If you want more technical information, Denali has a good whitepaper (registration required):

http://www.denali.com/en/events/usb3_whitepaper/?EB20090105

Re:USB3 whitepaper

[legirons]

Most of the replies so far show a glaring lack of knowledge of what USB3 really is. Honestly, it only bears a passing resemblance to its predecessors, and is a closer relative to PCIe. If you want more technical information, Denali has a good whitepaper (registration required):

And the real info can of course be found at https://secure.wikimedia.org/wikipedia/en/wiki/Usb3#USB_3.0

weirdly, wikipedia doesn't seem to include any of that stuff about upgrade to our standard or your company will die - you might need to turn to an industry-funded news source for the full story ;)

Re:USB3 whitepaper

[LordMyren]

Fantastic link, thanks.

Given that ExpressCard already has both PCIe and USB connectors, and that the spec you linked states,

Both the SuperSpeed USB and the PCIe specifications, therefore, are derived from the basic
OSI layered architecture. Both protocols look very similar in terms of layer architecture, and their physical layers share many common functions,
as well as similar concepts for other layers.

it'll be interesting to see if the confluence comes to a head and the two specs gain some kind of genuine interoperability. Afaik the current ExpressCard implementation works by having two sets of connectors; if USB 3.0 really is PCIe dervied, it would be great to collapse it to using the same PCIe interfaces.

The other two outstanding questions I have are:
1) how much the new architecture will alleviate latency?
2) is the time quantization better than the old 1ms standard?

Both of these prevent USB from being usable in real time contexts, contrary to evidence of the massive number of craptacular web cams sold.

Re:USB3 whitepaper

[Anonymous Coward]

Well, wikipedia has no original research. (Well, shouldn't have). As such, you always need to

a) spend a lot of time reading all the reference material yourself or

b) trust that whoever wrote the article understood the source material well and was able to write it down well enough again to not lose anything important or add any misinterpretations.

While I love wikipedia as much as the next guy, when presented options like that I would rather choose to read the industry funded news source directly myself and try to filter the "upgrade or die" parts away on my own.

Not necessarily the superior choice but not inferior either.

Re:USB3 whitepaper

[Anonymous Coward]

Both the SuperSpeed USB and the PCIe specifications, therefore, are derived from the basic OSI layered architecture.

osi's model has so many layers, i really doubt there's been a serious protocol that doesn't "fit" osi's model.

if you're abstract enough, you're can't be falsifyable; unfortunately the quoted statement can't even be wrong.

Faster data is great, but...

[exploder]

...does it have any greater power capacity?

Re:Faster data is great, but...

[Vectronic]

'exploder (196936)'
'...does it have any greater power capacity?'

Just what do you have planned?... :P

Re:Faster data is great, but...

[Anonymous Coward]

From my perspective, the biggest annoyance about USB 2 is that you can't power a 3.5" disk from it.

So, roughly double the power that 2 has would be useful.

Re:Faster data is great, but...

[Anonymous Coward]

Charging a plug-in hybrid off the USB port, while uploading songs to the built-in mp3 player.

Re:Faster data is great, but...

[neokushan]

You'll be glad to know that it does, but I'm not sure if it's enough to run a 3.5" Magnetic Hard drive.

"Maximum bus power is increased to 150mA per unit load (+50% over USB 2.0)."

A solid State drive, on the other hand...

Re:Faster data is great, but...

[pchan-]

You'll be glad to know that it does, but I'm not sure if it's enough to run a 3.5" Magnetic Hard drive.

"Maximum bus power is increased to 150mA per unit load (+50% over USB 2.0)."

A solid State drive, on the other hand...

Power is measured in Watts, not Amps. USB3 is still at 5V, but now lets you negotiate up to 1 Amp of current (USB2 limits at 500 mA). So, that's 5 Watts of power. the 150mA draw is the maximum current you are allowed to draw in before negotiating up to verify the host supports more.

Re:Faster data is great, but...

[neokushan]

Don't blame me, if you'll notice that was a quote, taken from the wikimedia article linked above.

Re:Faster data is great, but...

[sjames]

It's raised from 500mA to 900mA.

Re:Faster data is great, but...

[Hal_Porter]

You need to read a few words further in wikipedia

http://en.wikipedia.org/wiki/USB#USB_3.0
Maximum bus power is increased to 150mA per unit load (+50% over USB 2.0). An unconfigured device can still draw only 1 unit load, but a configured device can draw up to 6 unit loads (900mA, 80% over USB 2.0).

USB 1 and 2 had a maximum current of 500mA for a configured device USB 3 has 900mA.

Looking here

http://www.samsung.com/global/business/hdd/productmodel.do?group=72&type=62&subtype=67&model_cd=325&tab=spec&ppmi=1159

The 250GB HM251JI has a 900mA spin up current so it will work. The 500GB HM500LI has a 1000mA spin up current. Slightly over the limit but it will probably work.

Actually being able to run a 250GB 2.5" drive from one port rather than needing a Y cable to power it off two is quite handy.

Re:Faster data is great, but...

[petermgreen]

Afaict current USB can barely power a laptop HDD and has nowhere near enough power to supply a desktop HDD.

It seems USB 3 is increasing the max power but not by enough to run a desktop HDD.

Re:Faster data is great, but...

[Anonymous Coward]

5 Watts is not enough to spin up a 3.5" drive though...

Re:Faster data is great, but...

[moonbender]

When USB3 is prevalent (2010/11), 3.5" HDD will be fairly exotic beasts anyway, certainly for external storage.

Faster speed? Really?

[maugle]

According to Symwave, this will result in 'speeds previously unattainable with legacy USB technology.'

New technology will be faster than old technology? Impossible!

Comment removed

[account_deleted]

Comment removed based on user account deletion

Missing a tag in the summary...

[Anachragnome]

..."Advertising".

Yeah, but we all know that...

[denzacar]

... The Ultimate speed is 88 mph.
Once a device hits that the Flux Capacitor kicks in and it goes back in time.

There is probably a huge prehistoric dump of USB sticks from the future somewhere in California.

Re:Yeah, but we all know that...

[Anonymous Coward]

Dude, you never go FULL retard.

Re:Yeah, but we all know that...

[Laser_iCE]

Just make sure it doesn't drop below 50 or we're all FUCKED!

PR

[Dgawld]

"This will be the best uSb Evar!!"

Considering what a processor hog USB2 is...

[Lord Byron II]

...I'm not looking forward to USB3 all that much. With USB2, copying from one flash drive to another takes my CPU utilization to 100%. My speeds are constrained by my processor, not the USB bus. Does anyone know if USB3 is less dependent on the processor? Someone else posted about one using PMI and the other using DMA, but would someone like to break that all down?

Re:Considering what a processor hog USB2 is...

[aksansai]

With USB2, copying from one flash drive to another takes my CPU utilization to 100%.

Please upgrade from Pentium era processor.

Re:Considering what a processor hog USB2 is...

[Wildclaw]

Strange, since my external USB connected hardrive uses 3% of the CPU. Not what I would call a processor hog.

Re:Considering what a processor hog USB2 is...

[MBCook]

The 100% is a little nuts (bad driver?), but USB is definitely a CPU hog. I can move files between a computer and an external drive about twice as fast most of the time using FireWire 400 than USB2. FW800 is even faster.

The annoying thing, though, is the CPU usage. On my Mac (2.4GHz Core2Duo) I can copy large files (such as video or 8MP RAW images) as fast as the disk can take them, no problem. With FW (either 400 or 800) it barely shows up on the little CPU graph I use. It's easily under 10%.

USB2? Usually up to 20%.

It's been like this on every computer I've done this on. It's the same on my parent's iMac. It was the same of my PowerBook G4. It was the same on my old 933MHz Dell laptop.

The CPU thing isn't as much of a problem now, but it's still a problem. USB just wasn't designed for large transfers like HDs sometimes demand. It's nice they've improved things, and hopefully the fact that polling is no longer necessary will fix the CPU problem... but I'm not too enthusiastic.

Re:Considering what a processor hog USB2 is...

[Anonymous Coward]

Mobile phones use USB as well. The fact that USB requires the host processor to activate all the time is a huge impediment to lowering power consumption when streaming through that interface.

Re:Considering what a processor hog USB2 is...

[drinkypoo]

Core Duo T2600, Intel 82801G chipset, 2GB DDR2, copying from one typical disk to another on USB2 takes 22% CPU and runs at about 17 MB/sec, on 1394 takes 1% CPU (background noise) and runs about 23MB/sec (these are not fast disks, just ordinary 7200 RPM 3.5" WDs with 2MB caches, 160 and 250 GB.) This is on Linux, haven't done the test on Windows, sorry. But let's face it, USB2 is a lemon. It's okay for short-run tasks, but you don't want to actually USE data connected to USB2, you want to transfer it someplace else and use it. It's fine for flash drives and digital cameras, but not okay for scratch disks or even pro sound cards! If you're playing a game, and there's a dropout, you're annoyed. If you're recording a song, and there's a dropout, you start over. USB has its uses, but if you're connecting hard disks to it for long periods of time, and you care about performance, you're an idiot. Kind of puts ReadyBoost in perspective - I'm going to piss away CPU time for disk caching? By the same token, you can build a better filer by doing software RAID, because a dedicated RAID controller generally has a not-incredibly-impressive CPU, but today's general purpose processors have vector coprocessors in them to do all those fancy things that the RAID controllers are actually good at, so you can calculate parity quickly etc.

Don't touch type?

[spineboy]

Obviously you don't, but with my blazing touch typing speeds, I can take advantage of the new 3.0, and not have to worry about the lag any more.

Re:Don't touch type?

[martinw89]

True, and as an extreme gamer, I need the extra bandwidth for my 37 megapixel sensor blue-laser mouse. Red lasers just never felt accurate.

Re:Don't touch type?

[Lumpy]

Blue? that's old tech. burning green laser mice are what work best.

Ahh nothing like a game of UTIII and the smell of burning laminate to get your game on...

Ten times the speed!

[David Gerard]

This means my USB Christmas tree would flash ten times as fast!

Not to mention my USB chicken foot or USB mouse with a real scorpion in ...

Re:Ten times the speed!

[Joe The Dragon]

There only use usb for power and don't even have the data lines linked to them.

Re:Ten times the speed!

[David Gerard]

USB 3 loses then - it increases the current, but drops the voltage to four volts!

Re:Ten times the speed!

[petermgreen]

Thats only the minimum in-spec voltage, real voltages will most likely be much the same under the same load.

I suspect the lower minimum in-spec voltage is a result of higher volt drops at the higher currents they are using.

According to Wikipedia

[XanC]

The total amperage you can push through a port has gone from 500ma to 5400ma, or from 2.5W to 27W. That's a lot of power. Not sure that it's enough to spin up a magnetic hard drive though.

Re:According to Wikipedia

[tabrisnet]

I think you misread, it's not (900mA/unit)*(6*unit), but (150mA/unit)*(6*unit). USB 2.0 had 100mA/unit and a maximum of 5*unit.

Re:According to Wikipedia

[XanC]

Whoops, sure did! Thanks.

Ten Times Faster, Eh?

[dangitman]

If this is true, then USB might finally be usable for copying files and such.

Re:Ten Times Faster, Eh?

[Spatial]

Don't worry, it'll quickly become obsolete just like last time. Transfer speeds always seems to lag behind storage sizes.

Re:Ten Times Faster, Eh?

[symbolset]

If this is true, then USB might finally be usable for copying files and such.

... In Linux, XP or OS-X. On Vista and W7, not so much.

Not what the phone companies want

[Anonymous Coward]

The last tidbit at the end -the Chinese government wants all cell phones to be chargeable on USB connectors- is the most interesting part of the article. The phone companies may be fighting it tooth and nail, but its not up to them to decide. If they don't like it, then not only do they lose at least 1/5 of the worlds cell phone users, but because they lose a good chunk of the US and European users (using the cheap Chinese cell phones), then they lose roughly half their business. Whining about it won't get it done. Its good for everyone except the phone companies. Standards are monopoly breakers, its true. But China has a bigger stick than the phone companies, and not just that: consumers and environmentalists are happy with it too. This is so NOT A GAME THE PHONE COMPANIES CAN WIN.

Re:Not what the phone companies want

[nprz]

I am so happy that my motorola is charged by USB and I can hook it up to my computer, and copy files back and forth with the bluetooth.
This is one thing I wished the Japanese phones started doing.

Having one cable that can plug into any of my devices is great when traveling.

PIO vs. DMA

[Alwin Henseler]

You understand that PIO/DMA transfer modes only meaningful for Parallel ATA devices?

That's the meaning in traditional sense. But you can also use this distinction in a wider sense:

• PIO: The CPU has to manage / monitor / do every little step in the process.
• DMA: The CPU sets parameters, give a start signal, and then just waits (ehm, can do something else in the meanwhile), while dedicated hardware does all the boring work, like tranfer individual bytes / words of data to main memory. When ready, the CPU gets a signal (for example: an interrupt) that the transfer is complete. This may be used to describe many hardware-supported tasks, not just IDE harddisk controllers.

How much of an advantage this is, depends on how complex the initial parameter setup is, how much of the work is done by hardware vs. CPU, transfer speed, how large transferred blocks are, how often transfer occur, etc. etc. Besides overall speed, a big advantage is that the CPU can do other things (like decode a video stream, respond to keyboard / mouse input) while a tranfer continues in the background. This allows a system to feel much more responsive.

You state that USB controllers use DMA, parents says not. I don't know which is true. Perhaps there is DMA support for USB controllers, but the packets are small enough and flowing at a high enough rate that it feels like the CPU is doing all the work?

Re:PIO vs. DMA

[petermgreen]

You state that USB controllers use DMA, parents says not. I don't know which is true. Perhaps there is DMA support for USB controllers, but the packets are small enough and flowing at a high enough rate that it feels like the CPU is doing all the work?
That was the impression I got from the various sources I have read. I also think I remember reading that they are planning to fix this with USB 3.

Re:PIO vs. DMA

[Agripa]

That was the impression I got from the various sources I have read. I also think I remember reading that they are planning to fix this with USB 3.

I read through Intel's EHCI specifications and from what I understood, while DMA is supported it still requires double buffering because of alignment issues. The interrupt rate and required CPU supervision were probably not a problem until the 480 Mbits/sec transfer rate was implemented.

Re:PIO vs. DMA

[petermgreen]

hmm, reminds me of the description in the comment in http://www.freebsd.org/cgi/cvsweb.cgi/~checkout~/src/sys/pci/if_rl.c (just below the license header)

"The 8139 supports bus-master DMA, but it has a terrible interface that nullifies any performance gains that bus-master DMA usually offers."

Re:PIO vs. DMA

[markhahn]

PIO means that each byte/word of data requires an in or out instruction. this was common in older peripherals, including parallel ports, ancient ATA, even some NICs.

DMA simply means that the device controller writes into a main-memory buffer (or reads from).

Re:PIO vs. DMA

[swarsron]

we have more than enough cores available as soon as USB 3 will be available so i don't think that it'll be a real problem for the vast majority of the users. At least this way some of the cores are used and don't idle away ....

Previously unattainable with legacy technology?

[noidentity]

According to Symwave, this will result in 'speeds previously unattainable with legacy USB technology.'

Huh, so USB 3.0 peripherals working over USB 2.0 will be faster than USB 2.0? I kind of doubt that speeds attainable with legacy USB technology will be affected by this.

Re:Previously unattainable with legacy technology?

[Surt]

The result will be that usb3 devices connected to usb3 hosts will attain speeds that are unattainable by usb2 devices connected to usb2 hosts. Is it not clear that was what they meant?

To paraphrase...

[Colourspace]

..5.0 Gbits/s should be enough for anyone. We'll never see USB 4.0 in our lifetimes, let alone need it. /sarcasm

But what happened to USB2.1?

[Narnie]

But what happened to USB 2.1? Isn't there naming conventions to follow?

A better reference

[EdelFactor19]

An excellent reference can be found at Wikipedia. http://en.wikipedia.org/wiki/Direct_memory_access

You clearly have no idea what you are talking about. I did a co-op during college involving real-time OS, highly parallel processing, etc; primarily on embedded systems. (boards w/ lots of 500-800mhz powerPC chips and ram) Still rather than try to communicate it or take my word for it, take a look at this except from wikipedia...

"Direct memory access (DMA) is a feature of modern computers and microprocessors that allows certain hardware subsystems within the computer to access system memory for reading and/or writing independently of the central processing unit. Many hardware systems use DMA including disk drive controllers, graphics cards, network cards, sound cards and GPUs. DMA is also used for intra-chip data transfer in multi-core processors, especially in multiprocessor system-on-chips, where its processing element is equipped with a local memory (often called scratchpad memory) and DMA is used for transferring data between the local memory and the main memory. Computers that have DMA channels can transfer data to and from devices with much less CPU overhead than computers without a DMA channel. Similarly a processing element inside a multi-core processor can transfer data to and from its local memory without occupying its processor time, overlapping computation and data transfer.

Without DMA, using programmed input/output (PIO) mode for communication with peripheral devices, or load/store instructions in the case of multicore chips, the CPU is typically fully occupied for the entire duration of the read or write operation, and is thus unavailable to perform other work. With DMA, the CPU would initiate the transfer, do other operations while the transfer is in progress, and receive an interrupt from the DMA controller once the operation has been done. This is especially useful in real-time computing applications where not stalling behind concurrent operations is critical. Another and related application area is various forms of stream processing where it is essential to have data processing and transfer in parallel, in order to achieve sufficient throughput."

One of the first learning activities I did was to compare the speed of computing a matrix multiplication in parallel (for an 1000x1000 matrix of floats and then ints). Comparing: programmatically on one node to: 4 nodes using sockets (PIO), to DMA on 4 nodes. DMA is significantly faster than PIO because it avoided context/thread switches and more importantly interupts. This results in minimizing misses on the L1 cache. There was zero file IO at all, in fact there wasn't even a hard drive attached to those boards.

You make in interesting point; but PIO/DMA has implications FAR beyond just Parallel ATA devices.

Re:A better reference

[commodore64_love]

>>>Direct memory access (DMA) is a feature of modern computers and microprocessors

Eh. DMA is nothing special. My Commodore Amiga 500 has been doing DMA since 1987. Old news.

USB Video?

[nurb432]

Will i finally get a monitor that only needs USB? One less port to deal with.

Next, a wifi or bluetooth monitor.

Re:USB Video?

[threephaseboy]

If only such a thing existed.

Re:USB Video?

[nurb432]

Cool.

I stopped looking for them a while ago when all i would ever see were big bulky adapters that needed their own power.. ( sort of defeated the purpose ).

What the goal would be is to have a pretty powerful PDA that would use these devices either via usb or bluetooth so you can take your 'pc' with you, and if you stop somewhere just associate with the display and power where you are sitting for the day.

Add wifi + remote desktop ability to a server somewhere and you can have extra computing power/storage too.

Now we need to find a way to loop the video feed s

[Joe The Dragon]

Now we need to find a way to loop the video feed so we can get the people off of the usb bus.

Firewire isn't dea

[trolltalk.com]

Gee, don't say that to the aviation industry - they've standardized on Firewire because it saves weight in cabling.

The F-22 Raptor, the A380 Airbus, etc use firewire and gigabit ethernet to save weight. With over 300 miles of wiring an each A380, cutting the weight even in half makes a big difference with an A380.

http://74.125.77.132/search?q=cache:L96bOxSv3V8J:www.critical-embedded-systems.com/meecc/2005/presentations/Keller.pdf+army+tank+firewire+combat+electronics&hl=en&ct=clnk&cd=6&gl=ca&client=firefox-a " JSF Avionics snapshot

Distributed avionics: display- management computers, integrated core processing, and flight subsystems

IEEE 1394 FireWire network links core processor and display processors

Fibre Channel links core processor modules and sensor subsystems "

The military will be saying "You can have my Firewire when you pry it from my cold, dead hands." They have the bigger guns, so I think they'll win any argument.

Re:Firewire isn't dea

[Anonymous Coward]

I'm been writing avionic software for the military for over 15yrs, and I've never had to been with firewire. Ethernet, MIL-STD-1553, ARINC-429, RS-422, RS-232, Fibre Channel, PCI, VME, and, ...USB. Nope, no firewire.

Re:Firewire isn't dea

[trolltalk.com]

I'm been writing avionic software for the military for over 15yrs, and I've never had to been with firewire. Ethernet, MIL-STD-1553, ARINC-429, RS-422, RS-232, Fibre Channel, PCI, VME, and, ...USB. Nope, no firewire.

So what you're saying is that either:

1. you haven't been given the opportunity to work on the latest and greatest
2. the software you write doesn't interface directly with the avionics hardware, so it doesn't need to know the specifics

The F-22, the F35, the A380, the 747-8 - they all use firewire. Even unmanned military aircraft now use it http://www.google.ca/search?hl=en&client=firefox-a&rls=org.mozilla:en-US:official&hs=mcT&q=related:www.moog.com/media/1/LOA_08_News.pdf

Correction

[sjames]

Now that I'm reading the actual spec, USB 3.0 DOES support async notification from device to host (that is, interrupts). So a 3.0 device on a 3.0 port may place less load on the CPU than a 2.0 device while transferring data much faster. The device and port auto-negotiate that when the device is plugged in.

Re:Correction

[Tiger4]

So on the bottom line, is the a CPU penalty of significance with this new spec?

And while we are at it, what would the theoretical maximum USB rate be, using the same basic architecture? At some point it would be impossible to push bits down the bus any faster, right?

Re:Correction

[sjames]

I'll have to look at the host controller specs (which I don't have currently) to be sure defeat wasn't snatched from the jaws of victory, but it looks like there won't be a significant CPU penalty with 3.0

I'm not honestly sure what the max would be for a serial, but there is always adding extra serial pipes (3.0 adds one extra). The same tactic is used for 10Gig Ethernet, PCIE and Hypertransport.

There certainly are limits to how fast you can signal. You're limited by how quickly the level can be stabilized on the line and how quickly the receiver can reliably detect that level. That includes handling external noise.

Re:Correction

[reiisi]

What bugs me is that it seems to me that intel is trying to re-invent USB to do what firewire already does.

Compatibility here is an illusion. You're not going to hook a USB 3 device to a USB 2 port (or vice versa) on a regular basis. Most people simply will never do it, and those who actually need to do it will buy some special controller that makes it work for real, rather than relying on the half-baked compatibility in the spec.

Why bother? We already have firewire to USB converters.

Re:Correction

[sjames]

But you WILL likely connect USB2.0 and 1.1 devices to a 3.0 port. No keyboard or mouse needs to transfer gigabytes of data, why not use a cheaper and lower power chip that matches it's actual requirements? I have a keyboard, mouse, tablet, and parallel port (all 1.1 devices) hung off of a 2.0 port now through a hub. There's nothing half baked about it, it's actually well thought out and it just works. When I eventually upgrade and have USB 3.0 ports, I'll plug the very same hub in and it'll still just work.

USB was designed specifically to allow for low cost implementation. You will never see a firewire mouse because the interface would exceed the cost of the mouse.

I don't think illusion is the right word anyway since if you DID plug a 3.0 device into a 2.0 port, it would actually work just as well as any 2.0 device would. While that seems unlikely for a permanent connection, I can easily imagine using a portable 3.0 device on an older machine once in a while for backups (for example). I have connected a 2.0 device to a 1.1 port before to get 1 last backup before decommissioning the machine. There are other ways I could have done it, but that way was simple and easy.

Meanwhile, I have a firewire port and zero firewire devices. I have several USB ports and devices.

USB is being extended to take over the functionality of firewire as well as remaining affordable when doing what it already does well.Why in the world is that a bad thing?

Re:Correction

[reiisi]

But you WILL likely connect USB2.0 and 1.1 devices to a 3.0 port.

At this point in time that sounds more like posturing or threats than promises of a rosy future.

You do understand the logical fallacy of arguing from assumptions about the future? Six years ago, there were a lot of intel UWB camp guys claiming that, by this year, consumer devices would no longer use cables at all. Great silliness, especially considering intel's UWB spec.

No keyboard or mouse needs to transfer gigabytes of data, why not use a cheaper and lower power chip that matches it's actual requirements? I have a keyboard, mouse, tablet, and parallel port (all 1.1 devices) hung off of a 2.0 port now through a hub. There's nothing half baked about it, it's actually well thought out and it just works.

Bully for you. I have the same (well, no parallel port) and it likes to die when I hang a USB 2.0 disk drive on that hub. Admittedly, not the most recent devices (which benefit from a lot of consumer debugging and a lot of die-shrinkage that allows building more of a proper DMA interface into the same amount of chip real estate), but not first generation devices, either. Five year old tech.

I'm pretty sure that one of the reasons that hub hangs is the physical construction of the plug. It's really easy to lose electrical connection. And that's one of the glaring errors in the spec.

It's not a bad design for dongles and flash storage, makes for a great substitute for floppy disks. I'll give it that. (Unadorned flash RAM is not a good substitute for floppy disks. Physically too small.)

I have no reason to expect that USB 3 won't take ten years to work the bugs out of, or that we will be able to see the full effect of DMA at current chip design rules, or that we will ever actually see a decent, affordable implementation of both sides of the protocol in a single port, so that ordinary USB 3 devices can actually do proper peer-to-peer transfers like firewire devices can.

Not from intel's track record. They may have a comparatively decent track record on CPUs for the last three years, but they just don't get things. Things like, there's only so much trashing your own market you can do.

When I eventually upgrade and have USB 3.0 ports, I'll plug the very same hub in and it'll still just work.

You hope. Yeah, if it works, I'll eat some of my doubting words without too much complaint. But from this side of the future, I see no reason to make any business decisions relying on your assurances or intel's.

USB was designed specifically to allow for low cost implementation.

To a certain extent, designing down to the expectations of the current market is a good thing. But that's not what intel is doing, or they wouldn't be trying so hard to push perfectly good specs out of the market and replace them with their own.

You will never see a firewire mouse because the interface would exceed the cost of the mouse.

Huh?

At this point in time, chip design rules and available micro-controllers are just fine for packing a mouse or keyboard controller and firewire controller into a single chip at sub dollar prices. Maybe not pennies, but easily low enough prices to sell into the higher-end mouse market. By the time USB 3 is usable and you can plug a USB 1 mouse or keyboard into a USB 3 hub without degrading the performance of all the other devices on the hub, price parity with a cheap USB 3 keyboard or mouse controller will easily be possible.

And if that's not what you mean, the question is moot. 2 USB 2 ports + 2 USB 3 ports on a lightweight laptop is not substantially different from 2 USB + 2 Firewire. Temporary hook-ups mean moving cables anyway.

I don't think illusion is the right word anyway since if you DID plug a 3.0 device into a 2.0 port, it would actually work just as well as any 2.0 devi

Re:Correction

[sjames]

The first thing to realize is that USB is not owned by Intel and never has been. They are a more prominent member but they don't own it.

Bully for you. I have the same (well, no parallel port) and it likes to die when I hang a USB 2.0 disk drive on that hub. Admittedly, not the most recent devices (which benefit from a lot of consumer debugging and a lot of die-shrinkage that allows building more of a proper DMA interface into the same amount of chip real estate), but not first generation devices, either. Five year old tech.

You're probably running up against bus contention. The point is that the usage pattern is quite common. You were speaking as if nobody would ever have any reason to plug a 1.1 or a 2.0 device into a 3.0 port. Both you and I are now plugging 1.1 devices into a 2.0 port. I argue that that pattern is likely to continue.

You don't understand that you're arguing from intel's forward looking statements? These are statements that they themselves admit depend heavily on a market that we now know is not likely to appear, given the current direction of the economy. Right now, the expense of yet another standards war hitting the market place is not worthwhile, and I would go so far as to accuse intel of fiscal irresponsibility if they pursue this war.

Talk about forward looking statements! You figure the economy problems mean nobody will buy new computer hardware ever again? The standards war is over for the most part. Everyone has a firewire port and yet the vast majority of external drives are USB 2.0 because that's what the market wanted. None of those devices are slated for obsolescence by the new spec. MEanwhile, I'm looking at the spoec and saying that a compliant device will by definition interoperate with a 2.0 port and a compliant 2.0 device will work in a 3.0 port. That's not forward looking statements, that's by definition true.

I'm not trying to cram anything down your throat here and neither is Intel. If you like firewire that much, by all means use it.

At this point in time, chip design rules and available micro-controllers are just fine for packing a mouse or keyboard controller and firewire controller into a single chip at sub dollar prices. Maybe not pennies, but easily low enough prices to sell into the higher-end mouse market. By the time USB 3 is usable and you can plug a USB 1 mouse or keyboard into a USB 3 hub without degrading the performance of all the other devices on the hub, price parity with a cheap USB 3 keyboard or mouse controller will easily be possible.

Considering how long 2.0 has been out there and mice are still 1.1 devices, it seems likely that will continue to be the case. It's a question of volume and design costs. Why make a new rev when the old one works fine? It's a matter of volume manufacture. Firewire existed well before USB. The expanding use and popularity of USB mice and keyboards shows that the one wire just hot-plug it in was desired. No firewire mouse ever materialized. Apparently it either can't be done as cheaply or the market just doesn't find firewire attractive. Perhaps if MS had demanded multiple firewire ports on PCs, things might have gone different, but it didn't. These days people know what a USB port is and think of firewire as that radiation thingy port nobody uses.

Killing firewire and make everyone buy new devices when the old ones work just fine is only good for an economy in which resources are infinite. That's one.

I missed the paragraph in the spec where firewire devices are banned or actively blocked from working, where was that? The fact is that firewire has never managed to become mainstream. It remains a niche spec. I suspect that those who do use firewire will probably be able to get firewire ports and USB<->firewire adapters for longer than the useful life of their firewire devices.

And so intel uses USB 3 to corner the market on intercon

76 Virgins!?!?!?

[HRbnjR]

then something where they claimed you'd get 76 virgins for using USB... ...I think even some of the geekiest computer users start to not care when bombarded with all this

You must be new here.

Re:76 Virgins!?!?!?

[Jafafa Hots]

There has to be at LEAST 76 virgins on slashdot.

Re:76 Virgins!?!?!?

[Jafafa Hots]

OK, I'm cracking up that someone modded that as "informative."

Re:76 Virgins!?!?!?

[Anonymous Coward]

... at any one time.

Chinese mandate USB charging for mobile phones

[AlXtreme]

A more interesting bit from TFA:

Not connected with CES but related is the fact the Chinese government has declared its intention to force all digital phone makers to use a standard USB connector from the charger. That would mean that a single charger would do for all of your devices and would save an immense amount of wastage and frustration.

Good call, I hope to finally ditch those dozens of different chargers in a couple of years.

Re:Chinese mandate USB charging for mobile phones

[glassware]

See, not all government intervention is bad :)

Re:Chinese mandate USB charging for mobile phones

[man_ls]

Likely means we'll be seeing something similar on U.S. phones. Mainly because they're all made in China, and there's no way the companies would maintain two different production lines. (After all, both board layout and the casing would be different for a different charger.)

Re:Chinese mandate USB charging for mobile phones

[drew]

Most of the major cell phone manufacturers in the US and Europe had already agreed to do this over a year ago: http://news.cnet.com/Pros-seem-to-outdo-cons-in-new-phone-charger-standard/2100-1041_3-6209247.html

Perhaps China is looking to ensure that their local manufacturers get/stay on the boat. Or maybe they are hoping to speed up the process- It's been a while since I got a new phone, so I have no idea whether the standardization mentioned in that article has arrived yet, or when it might.

Re:Chinese mandate USB charging for mobile phones

[drinkypoo]

See, not all government intervention is bad :)

In further news, the Chinese government has announced its intention to murder and harvest the organs of anyone caught selling a new or used cellular telephone without a mini-USB jack for charging and data interchange beginning July 15, 2009.

Yay government intervention!

You did catch that article on the Chinese death vans, right?

Re:Chinese mandate USB charging for mobile phones

[badkarmadayaccount]

Motion seconded!

What about eSATA III?

[MartinSchou]

Am I the only one looking forward to eSATA III (or whatever they will call the eSATA standard that delivers power)?

I keep having this vision of a memory stick that I plug into my eSATA ports, gets SSD speed and doesn't bog down the CPU.

Finally

[hwyhobo]

It only takes the first hard disk backup or recovery to understand how frustratingly slow USB 2.0 really is. If USB 3.0 arrives soon, and if it delivers on its promise, it could also mean an untimely demise of eSATA, just as those devices started to show up.

I have had to perform laptop drive recovery a couple of times in the field, and having a way to do it without losing half the night before an important presentation would be a huge win. In fact, it will be a since qua non condition for my next laptop purchase.

Re: Firewire isn't dead

[Alwin Henseler]

The military will be saying "You can have my Firewire when you pry it from my cold, dead hands." They have the bigger guns, so I think they'll win any argument.

You mean a market niche just opened up for USB-powered Hellfire missiles? Count me in! Or should I order one of these ?

Latency?

[Anonymous Coward]

Let's hope it will be better than USB 1 and 2 wrt latency (which is very different from speed) Today Firewire spanks USB ass in everything remotely related to multimedia. USB based sound and MIDI cards, even the professional ones, suffer from high latency that limits a lot their use, meanwhile PCI connectors and game adapters that musicians use as MPU401 MIDI interfaces have completely disappeared from mainboards, and PCIe audio cards are still rare beasts.
I really hope USB3 won't be the joke USB 1 and 2 were, at least when doing music and multimedia stuff.

ps. Nope, I'm not a Mac user. USB is really, really, really inferior to Firewire in this field.

Re:Latency?

[petermgreen]

meanwhile PCI connectors and game adapters that musicians use as MPU401 MIDI interfaces have completely disappeared from mainboards,
Hmm, I think i've only once seen a board in person with no PCI slots (though I think it had one PCI-x slot which should work with double notch PCI cards)

I just went to dabs.com, looked at motherboards, sorted by popularity and looked through the first couple of pages, all of the ones for which the pic was clear enough to see the slot types seemed to have PCI slots (unfortunately DABs don't list PCI slot count in theier specs list so I had to go by the photos, if I was actually buying the boards I'd check the manuals but i'm just doing a quick survey).

Re:Latency?

[im_thatoneguy]

I just bought a Motherboard today. All of the ones I looked at had standard PCI slots (All i7 motherboards so top of the line latest chipsets). Considering the number of WIFI cards floating around that are still being released that are standard PCI I HIGHLY doubt that we'll see an end to PCI any time soon.

It wouldn't seem the musicians have anything to worry about. :D

Re:eSATA is hot pluggable if you use AHCI drivers

[cyclocommuter]

If you installed Intel's AHCI drivers during the OS install, eSATA will be hot pluggable. That means you can plug an eSATA drive to a PC, use it, then, unplug it without the need to reboot the PC.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Re:eSATA is hot pluggable if you use AHCI drivers

[ion.simon.c]

Are you unable to install Intel's AHCI driver after the OS is installed?

Re:Faster speed? Really?

[pizzach]

Impossible, but not improbable.

That's fine

[Sycraft-fu]

But just because the military uses it doesn't mean anyone else will. Computers may well get rid of it before long (Apple certainly seems to be doing that). I mean the military also likes Ada, but you don't see it being used to develop desktop apps often (or at all really).

By "dead" I don't think the grandparent means "Gone form the world," they just mean "Has no future in desktop PCs."

Re:That's fine

[lysergic.acid]

eh, Apple hasn't gotten rid of FireWire. they're just not including FireWire with their low-end products because Macs are now being used by more and more casual computer users. the MacBook Pro has FireWire 800, as does the Mac Mini, iMac, and Mac Pro.

FireWire certainly isn't something that is needed by the average computer user (though more and more people have large 1 TB+ external HDDs these days, and they would certainly benefit from FireWire) so to save cost, it makes sense to simply install USB 2.0 ports for peripheral devices.

saying FireWire is dead is like saying DSLRs or high-end video cards are dead. sure, the average person doesn't need that level of performance, but it's rather arrogant (and ignorant) to think that just because you don't have a use for something that it must have no future. i mean, i may not have a use for pro audio equipment, and neither do most people, but i'm not going to claim that pro audio equipment is dead.

FireWire still pretty much comes standard on most mid-range consumer desktops and laptops. HDMI and DVI interfaces are much more rare than FireWire, but neither of those interfaces are even closed to being dead. all this melodrama about Apple "abandoning" FireWire or FireWire "being dead" typically comes from poorly informed casual computer users. because the idea that everyone is suddenly going to abandon FireWire for USB 2.0/3.0 is laughable.

Windows Device Manager SATA & DMA

[AmigaHeretic]

Windows device manager shows my SATA HD using DMA.

Re:Windows Device Manager SATA & DMA

[jnork]

If it weren't, it would be slow as molasses.

One of the first things I do when setting up a new Windows system is set the optical drive to use "DMA if available". Windows seems to like to default those to PIO. Then when you try to burn a DVD it takes forever and requires using the buffer-underrun feature to keep from writing nothing but coasters. Reading is slow, too.

YMMV.

But I suspect that a SATA drive running in PIO mode would be slower than an IDE-33 drive running in DMA mode.

Re:Faster speed? Really?

[Anonymous Coward]

I guess you've never had the pleasure of flying with the Concorde.

Re:Now we need to find a way to loop the video fee

[Anonymous Coward]

Congratulations sir, this is the first Speed reference that I've seen on Slashdot!

Re:eSATA is hot pluggable if you use AHCI drivers

[Anonymous Coward]

I can hotswap it just fine. It's changes to the drive arrangement that cause the BIOS to make me redefine the boot order whenever I reboot (for whatever reason). The j-micron controller is the same SATA controller that hosts my Boot/OS drive. I guess it's just me. Still USB 3.0 will be great for card readers.

Re:eSATA is hot pluggable if you use AHCI drivers

[ion.simon.c]

TBH, I'd rather Apple drop the licensing fee requirement and see Firewire in more devices. Anything to get IO work out of my CPU!

You better believe

[symbolset]

Even my flash drive probably won't benefit without faster memory.

You better believe this is coming. Oh, my. You have no idea.

This is not true

[symbolset]

As another has pointed out here: support for hot swap is optional, not a required part of the SATA specification. Some controllers don't support it.

So the answer is.... no.

[symbolset]

5 Watts is still not enough for most 3.5" HDDs, even before you add the interface overhead.

Answer me this

[Seraphim_72]

Why is USB Handed? What I mean is why does it have a 'direction' that it must be plugged in? Yes I know there is contacts to electricity involved, but still and all, why for the spec didn't someone step up to the plate and say "It should be so you just plug it in in any direction, you know, like almost any electrical chord".

Re:Answer me this

[Anonymous Coward]

Might have something to do with the fact that USB power is DC and your typical wall outlet is AC.

Re:Answer me this

[duguk]

Why couldn't they put the same connectors on the other side of the inside of the plug, but the other way around? So it doesn't matter which way up it went? ;o)

I say make them triangular or something. That'll really make things confusing.

Just in time

[Fengpost]

to transfer my HD pr0n to my mobile hard drive!

Re:Just in time

[Anonymous Coward]

The trick is to put it on your mobile hard drive in the first place.
It isn't like you are downloading it faster than USB2 anyway, and you likely only watch 1 movie at a time, so you don't need more than that for bandwidth.

But now I can transfer it from my 2TB drive to my 4TB drive to finish my collection...

Re:Now we need to find a way to loop the video fee

[canonymous]

And this is why we will still need Firewire for use with video cameras!

one standard?

[reiisi]

Just for the record, I don't want a one-cable-fits-all universe.

For instance, maybe I have a luncbox computer with an internal HD and I want to use it as a server. The internal drive would be great for logging, but running a notebook hard drive as a server driver is not a guaranteed proposition. With a firewire port, I'm willing to boot and run it from an external drive. I could theoretically put / and /boot on the internal drive, and put /bin, /etc, /home, /sbin, /tmp, /usr, /var, and whatever on a USB drive, but no.

USB connectors really aren't up to the job -- slip out, corrode, just not really good for a semi-permanent connection. Also, while booting a USB thumb drive may be useful for some purposes, I don't think I want a server cpu having to work that hard, even if the server is just a home NAS.

It sounds like an extreme example, but we really need to be throwing less of our electronics devices away all the time. USB is great for keyboards and mice, but for storage beyond flash, it's mostly throwaway tech. A firewire port is going to better for extending the useful life of a device. SAS or (uggh) eSATA would be better, of course.

The point here is that there are good reasons for having multiple standards, and this "Improve the cheap solution and sell it hard!" attitude tends to me to be a good path towards polluting and devaluing the market.

Might sound a bit extreme, but I'm going to go out on a limb and say part of the blame for the current economy lies with this insistence on capturing the market with trash.

Re:eSATA is hot pluggable if you use AHCI drivers

[Anonymous Coward]

Watch out with changing the BIOS setting for the drive interface type from to AHCI or RAID if your OS is Windows and resides on that drive. You'll get a blue screen on boot as it can't find the system disk. If you want to change this without reinstalling Windows, it's usually a hack of the driver .inf files to get them installed under the old setting, then a registry hack and a reboot or two.

I was not brave enough to try it.

whoosh

[reiisi]

I know I'm letting the joke fly over my head, but UWB, until iNTEL killed it by insisting it was their tech or none at all, was supposed to get rid of the cables entirely for most consumer devices. No more plugs.

I'm becoming a sexist pig in my old age.

[reiisi]

When I was young and innocent, I wanted a wife like Marie Curie.

The longer I live, the more I am convinced that most woman demonstrate an almost allergy to technical reasons for things being other than they think they should.

By the way, there was once, about six or seven years ago, an opportunity for the industry to do away with the wires almost entirely for most consumer grade devices.

You probably wouldn't have wanted to use Freescale's UWB for mounting the drive containing your /usr and /home, but it would have been fast enough for watching video in real-time, for moving files between your camera and your laptop, all of that. And it would have been about as secure as wire, as well.

intel and their group have done their best at erasing the history from wikipedia, so, no, you'll never read most of the lurid details.

Integrated? what was integrated?

[reiisi]

Lousy Xebec controller that started it all. I think it was designed by someone who didn't know that the 8086 had a lot more port addresses than the 8080.

And, yes, the Xebec controller was integrated, so to speak.

point of view

[reiisi]

Isn't your description from the user side of the driver?

Re:point of view

[Hal_Porter]

I'm not sure what you mean. I'm talking about USB 1.0/USB 2.0 devices in Windows upto Vista with no UMDF. In this case the USB device driver and the USB host controller driver are both in kernel mode.

It wouldn't change much if user mode code called the USB device driver though, then you'd lock the user buffer into memory to stop it being paged out and build a scatter gather list, i.e. a list of physical pages covering the user buffer. The USB host controller driver would put these addresses into transfer descriptors and the host controller would DMA the data from the device right into the user buffer.

In fact UMDF does pretty much this as far as I can tell. There's a generic USB kernel mode driver for UMDF devices. User mode code can issue DeviceIoControl calls down to that.

The point I was trying to make is this is all driven by PCI/PCI Express bus master DMA.

get that

[topanz]

i think i must to be get that

(obligatory) netcraft confirms it

[reiisi]

Or should we be quoting Monty Python, from Holy Grail?

hmm. I think intel is the one with the mallet.

By "they", you mean intel, I think?

[reiisi]

intel being the owner of the USB spec, we would suppose they think they stand the most to profit if they corner the markets on interconnect.

And that is precisely what they are trying to do. They killed UWB because they couldn't control it.

But they're just fouling their own market. If that market didn't have a huge overlap with the rest of the computer industry, I'd say let them commit technical suicide, and be done with them.

dual processor

[reiisi]

Which is why intel is so enamored of dual processors.

netcraft confirms it ...

[reiisi]

... desktop PCs are dead.

lack of security

[reiisi]

you mean, lack of security, like SCSI? Hmm, sort of like ethernet, too?

not surprised

[reiisi]

WOW!

intel chips don't fully support an intel spec?

Didn't think about power

[Britz]

Why not just extend the specs and put a little power connector at the side or on top? A power connector would be easier to implement than a whole new standard (USB3).

Re:Didn't think about power

[Siffy]

They are. Well, not extend but revise. Sata rev 3.0 should support supplying enough power to run a 2.5" hard drive.

http://www.engadget.com/2009/01/07/faster-sata-standard-coming-in-second-quarter-of-this-year-says/

But there's no telling when it'll actually hit market given they've been officially talking about for a year.

http://www.engadget.com/2008/01/15/sata-io-cranks-up-power-over-esata-initiative-not-a-moment-too/

Until I actually see it I'll think of it as vaporware and pretty much half-assed and almost useless until it can run a 3.5" drive.

acronyms 101

[Anonymous Coward]

PIO: polled input/output. The CPU has to sit there and shuffle data in and out of an I/O port. If it's lucky it'll get an interrupt to let it know when the device is ready for more.

DMA: direct memory access. The device can take control of the memory bus and read and write data itself, without the CPU necessarily having to do anything.

These are basic computing concepts. The acronyms certainly pre-date ATA/IDE or the IBM PC.

Get it in 2010

[Anonymous Coward]

along with Linux for the desktop

Firewire gives you 77 movies of virgins.

[Organic Brain Damage]

Proving yet again, Firewire is Betamax to USB's VHS.

Re:Faster speed? Really?

[TheVelvetFlamebait]

Joke goes horribly wrong: opportunity to use stale meme missed

News at 11

multi-interface

[Anonymous Coward]

I tried carrying a Firewire external drive a few years back. I ended up selling it because I almost never got a chance to use it.

All current external drives have at least USB, but also optionally are available with FW as well. This way you have the ubiquity of USB, but if you have access to a machine with a FW port you can advantage of the extra speed.

My external drive has USB, FW400, and FW800. For Time Machine on my iMac I use FW800 (faster, lower CPU), but in an emergency I can connect it to any machine because of the USB.

Best of both worlds.

Vista

[Anonymous Coward]

According to Symwave, this will result in 'speeds previously unattainable with legacy USB technology.'

New technology will be faster than old technology? Impossible!

Well, there's always Vista...

The question is now...

[Jorophose]

Do you wait until 2010 to upgrade an old PC?

Or do you buy now and screw USB3.0? Do you invest in AM2+ or LGA775, or do you wait Core i7? Do you consider an ARM solution or Atom to hold you over?

I've got a really old PC. And USB3 was one of the last things I was waiting for to upgrade... But at this point is it even worth bothering? What the heck is going to need USB3, other than maybe external hard drives?

Re:Faster speed? Really?

[YourExperiment]

New technology will be faster than old technology? Impossible!

That's right Mr. Ballmer!

user software vs. user mode

[reiisi]

I mean, from the point of view of the software using the USB port, not the software running in OS user mode, it looks like that.

But from what I've read, actual implementations tend to push most of the handshake onto the (software) drivers. That means the CPU has to do a lot of, you guessed it, polling states. Even USB hardware that supports interrupt mode tends to be so kludgy that it is often more effective to use low-level polling. And the standard allows that.

From what I've read.

And it shows. USB devices, especially cheap ones, tend to consume an awful lot of CPU time compared to proper DMA implemented interfaces like SCSI or firewire. (Although I'll acknowledge that there were a number of cheap Macs that did that with the SCSI interfaces. Not something they could have gotten away with at all with x86 chips of the same era, but not something they should really have done, either. With the SCSI interface, you could at least say that the design did not meet standard.)

Maybe you don't notice it so much when all you have is the old IDE/pATA interfaces to compare it with.

And maybe you don't notice it so much when the CPU is actually two, so that one CPU can be busying itself with polling the interface while the other is doing your work. (Which would be one reason intel is so enamored of multiple cores.)

But it's there.

Re:user software vs. user mode

[Hal_Porter]

I mean, from the point of view of the software using the USB port, not the software running in OS user mode, it looks like that.

But from what I've read, actual implementations tend to push most of the handshake onto the (software) drivers. That means the CPU has to do a lot of, you guessed it, polling states.

That's not true. On a desktop machine it works like this.

For USB 1.0 there are two types of host controllers in use - OHCI which is backed by Microsoft and UHCI which is a backed by Intel. UHCI is slightly simpler hardware than OHCI. Both of them use DMA and are not PIO based though. For USB 2.0 there is only one host controller standard called EHCI which is DMA based.

For USB interrupt pipes, which are actually polled, the polling is done by the host controller. The host controller driver sets up a structure in RAM to list the endpoints that need to be polled. Only when one of them replies with data will it generate an interrupt for the CPU.

Check out the specs for the OHCI, UHCI and EHCI controllers. All of them are use DMA to read from a structure in memory. The host controller puts isochronous endpoints at the start, then interrupt and finally and bulk endpoints that need to be read. The actually polling on the wire is done by the host controller.

Of course the standard allows for any kind of host controller. This is so embedded systems can act as USB On The Go using FIFOs rather than bus master DMA. I've never seen a PC host controller work like this though, and it probably isn't possible because desktop OSs like Linux or Windows will usually have poor interrupt latency compared to an embedded system.

Re:user software vs. user mode

[reiisi]

LOL

You realize something?

Your post explains precisely why I don't like USB.

Have you read comments from some of those who have been involved writing USB drivers for Linux?

But, yeah, just so you know, the CPU can actually handle polling. Plenty fast enough, as long as you don't mind the cycles it takes away from whatever you were wanting your computer to do.

A Little Reality Check

[Anonymous Coward]

1. USB 3 will require an entirely new cable, not compatible with USB2 etc. The connectors are the same, but the cables will all have to be replaced.

2. Like its predecessor, USB3 is not anywhere near as fast as advertised. Real-world performance (among vendors, mind you!) has been measured at about 40-60% of advertised top speed. Yes, that's still faster than USB2, but not as fast as Firewire S3200 or eSATA.

3. USB3 is already a year late -- I refer you to this slashdot story which promised USB3 for early 2008.

Interesting tidbit FTA

[Peepsalot]

Not connected with CES but related is the fact the Chinese government has declared its intention to force all digital phone makers to use a standard USB connector from the charger. That would mean that a single charger would do for all of your devices and would save an immense amount of wastage and frustration.

I never thought I'd say this, but: Three Cheers for the Chinese Government!