eSATA on desktops is cheap

Almost all desktop PCs have spare SATA ports on the motherboard and a bracket with an e-SATA socket and an internal SATA lead is available for very little money. (Aria sells the Darkstar E-Sata Bracket and Cable for £1.44 inc VAT.) Many motherboard manufacturers include such brackets with their motherboards.

As USB3 is not likely to be common for 2 or more years and will NOT be transparent to the OS (unlike e-SATA) and WILL require an additional controller board for all existing PCs, I think that e-SATA will have a brightr future than implied in your article.

Doesn't USB still suck ?

USB's theoretical speed is well and good but:

1- I've never reached it. Not even close

2- When doing file transfers over USB 2.0, my PC suffers from mini-freezes

3- Some of my USB 2.0 periperals won't work, or will even hang my PC, when connected at 2.0 speeds, I have to force 1.0 to get them to work.

On the other hand, I never had a problem (neither speed nor compatibility nor sluggishness) with FireWire.

My take on it is that early USB was designed for slow periperals: kb, mouse, printer... and specifically not HD. There must be something in the way USB 1.0 and 2.0 were done that made them unreliable ressources hogs for high speeds/big transfers.

Do you know if USB 3.0 fixes that, or is it just a faster release of them same very touchy s**t ?

Mousing

This is good news for us professional mousers. I can whizz around my multiple monitors so fast I find there isn't enough USB 2 bandwidth for my mouse.

It's academic until solid state drives are everywhere

When you're copying huge files then burst speed's irrelevant anyway.

When you're read/writing all over the shop the same is true.

eSata needs more support

I bought a portable drive enclosure with USB and eSata from Overclockers; coupled with a 7200 rpm 2.5" hard drive it makes a great home for my virtual machines (try copying VM disk files over USB). In the office the drive slides into a slot in my desktop machine (a 3.5" drive bay fitting that came with the enclosure) and the drive works as good as an internal drive. When I'm out on the road the drive slides out and works as a USB drive with my Dell laptop.

As for eSata support, it is bad, but "professional" laptops like the Dell Latitude E6500 have eSata. A big problem is the storage vendors (WD, Seagate etc) aren't supporting eSata, so I had to go an build my own.

Real world speed?

It is all very well talking of the raw Gbit/sec speeds, but the whole USB standard sucks from a technical point of view, and often firewire disks at 400Mbit/sec would comfortably out-perform USB 2.0 at 480Mbit/sec. What will USB 3 *really* deliver?

The only real attraction of USB 2 external HDD is the small ones can be powered from the USB port (so long as you don't try a long extender cable). Will that still work with the interfaces needed for Gbit/sec USB 3?

USB vs eSATA vs FW

USB relies on the host CPU (unless there's some fancy accelerator tech I haven't heard about though I am aware not all USB motherboard controllers are created equal), but eSATA is presumably DMA - using the same motherboard controller as the other internal SATA ports and FW has it's own processor.

However, since USB can be used for _everything_ wired to your PC, including your coffee cup warmer, it will be used for everything, because the interface will be extremely cheap. If you want real throughput you probably won't be using USB, but since not everyone needs real throughput, those people will make do with USB.

FW can also be used for more than just HDDs, so it might be cheap, but it might get abandoned in the face of USB 3. That would leave POeSATA to be the storage choice. BUT we're forgetting power over ethernet might be just as appealing.

Then there'll be short range high performance wireless links coupled with wireless induction for power.

'ang on a tick...

If the new SATA spec is 6Gbps, then the eSATA version will doubtless be identical since eSATA is just SATA piped to a posh socket on the back panel of the PC. And if it's already built into the chipset on the MB, how hard can it be to bung in a couple of lines on the board to connect it to the socket?

Maybe nobody needs the extra 1.3Gbps of SATA over the USB 3? Actually, I wouldn't mind having a bit more throughput since I run other OSes from eSATA externals now, but I'm probably the only person on Earth doing so.

And I guess the USB overheads will mysteriously vanish in USB 3 as well.

Mmm, right-oh.

As you were then.

raw speed versus real performance - long live esata

Raw speed is not what most people really want to know, just the only info easily available info

My experience with USB disc conenction is the performance is usually way below what you would expect, and is also very eratic - just watch the time remaining when copying large files on windows.

Also USB often puts a fairly heavy load on the CPU which slows down everything.

I've found that on many machines, trying to run a Virtual Machine from a USB disc is a non-starter - it will often crash within 5 - 30 minutes

No such problems with eSata connected discs - fast, reliable consistent, and well worth the small inconvenience of the PC card and cable. Oh yes - and for real speed I've found that an express esata card normally knocks spots off a cardbus card

Native USB 3.0 drives?

It's possible to imagine that USB 3.0 will be integrated direct into drives. In that case it would be possible to eliminate SATA from future motherboards provided that there is BIOS support for USB 3.0. Of course USB has higher CPU overheads than SATA, but processor power is not in short supply these days.

For the very fastest access, SSD cards directly plugged into PCI-X slots would do the trick, although that, again, needs BIOS support for boot disks. Future PCs with SSD cards for the system and hard disk support via USB 3.0 for bulk data would make a lot of sense. An intelligent operating system could even automatically place data onto the right sort of location based on access patterns.

USB is consumer level

Surely the problem with USB is that there's no guaranteed throughput - all devices compete, and the latency can be quite bad. This is why, for instance, firewire audio cards are much better at recording in real time than USB audio boxes, which have a significant delay.

This doesn't matter so much for bunging a hard drive into your desktop/laptop (so long as you're not doing really high end video work), but it does matter for servers. eSATA is particularly good for expanding either your server or your existing disk array with another large box of disks (and it's a lot cheaper than fibre channel) - I can't imagine USB 3 making many inroads there.

4.7GBps

Hopefully USB 3 will not require so much CPU intervention as USB2? 480MBps was never achieved in practise and Firewire was much faster.

The clue is in the name

The other reason USB 3 will be king, is that USB is Universal for all types of devices. Fireware and eSATA have limited device support, almost always high capacity storage devices. The average user just wants one set of ports to plug they low def webcam/ printer/ 1TB external hard disk into.

Delayed Write Error

Don't get those with eSata, shame if it does die a death

Killer

USB's always going to be around 'cos it's a general purpose interconnect, and so can be used for hundreds of different things.

Personally I reckon eSATA will kill off Firewire (as the connector of choice for discerning external storage bods), for all but niche video editing stuff.

eSATA/USB combo on 250quid netbook by Toshiba NB100

Bought a 250quid inc vat NB100 from staples last week during their sale. Price now gone up to 294.

This netbook has eSATA/USB combo ports. So market penetration of eSATA is enabled by 1) dual-purpose ports (eSATA and USB) - no extra ports needed, form factor and some costs kept down. 2) Cheap hardware - and netbooks are the fastest selling notebooks.

On the other hand I can see USB 3.0 winning the day as this ought to provide full backward compatibility where as eSATA isn't compatible with USB - unless your USB port is a eSATA/USB combo. It all depends on if USB 3.0 hardware at launch is cheaper than eSATA. eSATA is here now and already has a head start recouping the R&D costs to progress to a commodity item.

I suspect a stalemate of co-existence much like the DVD +/- formats.

eSATA expensive?

I've seen cheap nasty internal card readers with eSATA ports. Simply plug them into a regular SATA port on your motherboard and you're ready to go. Throw in an eSATA hard drive caddy from ebuyer or the like and your choice of drive and you've got an eSATA system for not a lot more than a USB external hard drive.

Of course anyone who thinks that's too complicated isn't going to appreciate the extra 20MB/s* transfer speeds anyway.

*guestimation.

Can we standardize on an external bus, please?

This entire USB/Firewire/eSATA business has been a big waste of time, IMO.

What I would like to see is more of a hybrid use of the Ethernet port. Put it into true 'network' mode for talking on a net, but have the option to put it into a 'peripheral bus' mode.

eSata != USB 3.0

There are distinct advantages to eSata vs. USB 3.0. Speed is not the only criteria. Besides the fact that the internal drive cannot sustain either eSata or USB 3.0 data rates, eSata is IMO much better suited for external drives that are continuously (mostly) attached to a system, and especially for external RAID arrays. You can add a nice 4 port eSata RAID controller to a system for under $50 USD these days (25 quid?) that will provide you with a very nicely performing backbone for RAID 5/10 enclosures. So, for a total of under $1000 USD you can put together a 4.5 TB array (6 TB including parity) that takes up about as much space as a shoe box.

I'll admit, I do like USB drives for data that I want to sneaker-net between systems. If it's under 16GB, I'll use a thumb drive. If more, then a nice USB enclosure with a high capacity HD is ideal. Unfortunately, it will be some time before USB 3.0 is as ubiquitous as USB 2.0, so even if one has a USB 3.0 drive, it still be limited to the speed of the system to which it is attached.

That said, I will admit that I will likely be the first person on my block to buy a USB 3.0 drive and controller for my workstation, and possibly a controller for my laptop as well.

eSATA == SATA, just another connector

From reading the comments, I see that there is a fact some people are missing: eSATA is identical to SATA, the only difference is the shape (and quality) of the connector. A SATA connector is rated for about 40 reliable plug/unplug cycles, eSATA several thousands. You can also buy (cheap) SATA to eSATA cables, brackets,...

USB 2.0 and Firewire are limited to 2TB ...

One issue that hasn't been covered explicitly here is that USB 2.0 and Firewire are both limited to 2TB (32-bit block numbers x 512-byte blocks == 2^32 x 2^9 == 2^41 == 2 x 2^40 == 2TB). Meanwhile eSata supports 128PB (48-bit block numbers x 512-byte blocks == 2^57 == 2^7 x 2^50 == 128PB). What is the size of logical block numbers for USB 3.0? I haven't found that information in a few minutes of google'ing.

In any case, the absolute size of disks that can be addressed is going to become more and more important as giant video files become popular.

Speed hype vs real speed

One thing nobody has pointed out yet is that USB is a shared bus. Actually, one other commenter did point it out, but they pointed it out as an advantage. It's not. A shared bus is a disadvantage when you're talking about storage devices. With my USB2 connection, I have an approximate bandwidth of 48MB/sec. That 48MB/sec is shared between my external hard drive, external DVD rewriter, UPS, mouse, keyboard, and audio device (Zoltrix 5.1-channel). It may not sound like much, but trust me, it is. Without music playing, I can burn an ISO to CD at an average speed of 48x. With music playing, burning that same ISO reaches a maximum of 42x.

Will that really make a big difference? It depends on what you're trying to do. If you hook up multiple external hard drives, then USB will most definitely slow down transfers when both drives are transferring simultaneously. With an eSATA connection for each drive, each drive would have its own channel, thus its own dedicated bandwidth, so it wouldn't be slowed down.

Also, as others have mentioned, USB is *VERY* CPU-intensive. When playing music from WinAmp through my laptop's internal Conexant HD Audio output, my average CPU usage is 1% (as you would expect when it's not doing anything else). When I switch the output to the Zoltrix 5.1-channel USB audio output, the CPU usage averages 14% (it averages 14% whether I set it for 5.1 or 2.0 channel mode). When I look at the processes, the System Idle Process averages 97% and the System process averages 3%, yet the total CPU Usage shown at the bottom of Task Manager shows 14%. That very clearly illustrates that the USB bus (specifically, sending the audio data to the Zoltrix) is taking up a lot of the CPU's time.

USB is a good idea for general connectivity, though something does need to be done about the CPU usage. But no matter how you look at it, USB is not the bus to use for storage.

@The clue is in the name

"The average user just wants one set of ports to plug they low def webcam/ printer/ 1TB external hard disk into"

You seriously want to move 1TB of data through USB?

I've just spent almost 2 hours trying to transfer 57GB - and it failed - twice.

Why is it impossible (in the UK at least) to buy a SATA drive enclosure with FW ports?

This again?

Firewire has, yet does not need hubs, since you can chain devices together like with SCSI. Most devices have 2 ports, and you can just keep plugging them into eachother, up to 127 or something. And it's rather decent about sharing bandwidth and keeping latency low.

I don't know who in their right mind would do a 4 disk external raid. If you have data important enough for redundancy, then spend a couple extra bucks and get a case with extra internal drive bays. Or get a hotswap deal to stick into your 5.25" bays.

With reasonable software raid (Those $50 raid cards do software raid too, don't be fooled,) you can raid anything you want. I'm referring to md in Linux and whatever OSX uses. I've not used OSX's version much, but software raid in Linux has most of the features of a high end hardware raid card (online capacity expansion, background rebuilding / consistancy checking, shared hot spare, etc) and it's faster than a couple of the low end hardware raid cards that I've tried. And you could make a raid out of a network disk (AoE), a sata disk, a usb disk, and a firewire disk, and a disk image file. It'd probably be slow as balls since that USB disk would probably run slower than a network disk using 100mbit ethernet, and it'd introduce all kinds of rampant latency, but you get the idea. I've also heard of someone using OSX to raid together 8 usb floppy drives. It's not a good idea, but it illustrates the flexability of the system.

Yes, USB sucks, especially the Intel and VIA controllers, since they do even less of the work than a regular controller, leaving more up to the cpu to deal with.

As to the math, not every single bit in the GBit rating is actually used for data, they typically do some sort of 8 in 10 encoding to get some error checking and data resilliency, so you actually just divide the rated speed by 10 to get the actual speed in bytes.

Don't get me wrong, I've used external USB drives, but I typically do it when I'm not even kind of in a hurry, mostly when backing up users' machines before I reformat or whatever. Then again it was hard to tell whether the shitty computer, shitty drive, shitty enclosure, or USB that slowed things down.

I mostly hope that there's always some alternative to USB that isn't too expensive, since I may need an external drive in the future, and I don't really want it plugged into my mouse port.