Serial ATA and USB 2

An anonymous reader pointed us to an anandtech article that discusses the future of device attachment. Specifically USB 2:Electric Boogaloo and Serial ATA. Definitely 2 standards will likely matter a lot in the future. That Serial ATA stuff looks interesting...

Why reinvent the wheel?

As far as Serial ATA goes, why reinvent the wheel? Why not just adopt the SSA protocol for drive attachment? If it was widely adopted the price would come down significantly. As far as speed and flexibility, it is already further along that Serial ATA is attempting to be. I guess that since it comes from IBM, the odds of it being adopted elsewhere (despite the published spec) are slim, though.

post firewire flames here

fellow karma whores:
in the interest of space conservation and sorting, and so that the rest of the discussion can be devoted solely to discussion of USB 2.0 and Serial ATA themselves, i would like to request that all Firewire-related flaming be posted as a reply to this message.

suggested topics:
-USB 2.0 is a pathetic excuse to destroy Firewire, designed solely to prevent a standard not controlled directly by Intel from gaining importance.
-Firewire/Serial ATA/USB 2.0 is not open enough in the sense that the open source movement would like it to be.
-Firewire is outrageously liscenced and expensive.
-USB 2.0 and Serial ATA will be far more widely used and supported than Firewire.
-Firewire will be far more widely used and supported than USB 2.0 and Serial ATA.
-WH04 D00D 1 \V4N7 4 830\VULF CLU573R 0F 1B00KZ
-[Firewire, Serial ATA, USB 2.0] is far technically superior to [Firewire, Serial ATA, USB 2.0]
-[Firewire, Serial ATA, USB 2.0] is vaporware.
-by the time [if ever] USB 2.0 comes to market, Firewire will already be in its second generation and far superior to the USB 2.0's first generation in every way USB 2.0 defeats Firewire now.
-Firewire is already widely supported.

OK, have fun!

In depth reporting... not

Oh boy, USB 2.0 will be 480 Mbps. It will be forward and backward compatible -- wowee.

Their treatment of USB 2.0 was severely lacking. I wish Apple would drop its firewire tax (if not completely, then at least a few notches) so we could use a technology that was designed for things like this. This is the second time I've linked to this article today, but it's worth a read.

• http://www.mackido.com/Hardware/USB2.html [mackido.com]
• http://www.mackido.com/Hardware/USB20.ht ml [mackido.com]

It's easier to just link to these articles than just rehash the same old arguments every time USB 2.0 is mentioned. Suffice it to say it'll be more than a day late and a dollar short, but, sadly, it will probably catch on due to Intel's stranglehold on the chip market. Once again, if Apple would ease up on the 1394 tax, we would all be better off (and Apple would have at least one victory against Wintel under its belt). Alas, they will not do that, so tomorrow's iMacs will probably ship with USB 2, 3, and 4, and FireWire 2, 3, and 4 will be quietly brushed under the rug.

_________________

Serial ATA: The Unnecessary Standard

The ideas behind Serial ATA are quite sound. The problem is that there is already an existing standard which accomplishes all that Serial ATA does, and much more, and is not inherently any more expensive.

That standard is IEEE-1394, also known as FireWire or i-Link.

Like Serial ATA, 1394 runs at very high speed over inexpensive cabling consisting of only two pairs of signal lines and one pair for power. The cabling does not impede airflow like parallel ATA cables.

Like Serial ATA, 1394 can be used to transfer data to or from ATA (IDE) interfaces, completely transparently to the host.

Like Serial ATA, 1394 can be (and is) implemented in extremely cheap chips. In fact, there are sub-$5 "tailgate" chips which provide a single-chip 1394-to-ATA adapter. If 1394 was integrated into the drive instead of Parallel ATA, the drive could actually be cheaper than it is now.

UNLIKE Serial ATA, 1394 can also be used to connect SCSI devices, digital cameras, digital audio and video, TCP/IP networking, and many other categories of devices. 1394 already appears to be the interface of choice for most upcoming consumer electronic gear.

UNLIKE Serial ATA, 1394 is already a recognized standard, and work is already underway to extend it to speeds of up to 3.2 Gbps.

UNLIKE Serial ATA, 1394 supports complex topologies: devices with multiple 1394 ports for daisy chaining and hubs. This provides considerable flexibility in how devices are hooked up.

UNLIKE Serial ATA, 1394 supports multiple masters on the same bus.

UNLIKE Serial ATA, 1394 is available NOW, and is already built into some computers.

UNLIKE Serial ATA, 1394 is already supported by Microsoft Windows, and to some extent, Linux.

So why do we need a new standard?

Part of the problem is probably Intel. Initially they announced that they were a supporter of 1394, and that they would build support for it into all of their chipset. They did this with USB, and now it's hard to buy a PC without USB. But when push came to shove, for some reason they didn't do it. Apparently this is due to their work on "USB 2", which pushes the speed of USB into the same range as 1394, but unfortunately still has most or all of the limitations of USB.

Part of the problem is probably Apple. They made ridiculous royalty demands ($1/port), and scared many vendors away. They've since backed down to much more reasonable numbers, but some of the damage was done.

Part of the problem is probably the ATA committee itself. They may be experiencing "NIH" syndrome, preferring to invent a new standard rather than using an existing one, no matter how suited the existing one is.

Part of the problem is just the standard chicken-and-egg question. If computers don't have 1394 interfaces, why should disk manufacturers build 1394 into disk drives? If disk drives don't have 1394, why should computer manufacturers build 1394 into the computers? Of course, serial ATA may have the same problem, but it may be less pronounced. The very fact that serial ATA is less functional may make it an easier sell from a marketing point of view.

What should be done? IMNSHO, they should scrap the proprietary Serial-ATA interface, and adopt 1394 as the official Serial ATA standard.

Re:Big difference

As soon as I posted that comment, I knew I should have clarified: Of course 1394 can be used internally. USB can be used internally, too. Parallel ATA can be used externally. But that's not how they're intended to be used.

I should have remembered that the G4 has an internal 1394 port... To make up for my mistake, I spent the past 20 minutes searching for internal firewire devices, and I couldn't find a single one. If 1394 is really intended to replace ATA, then why does Apple still use ATA drives in their latest machines? (It's not like they're big on backwards compatibility these days...)

I stand by my original statement: 1394 and Serial ATA are not competing standards. They do different things conceptually, even though the technology is similar. It wouldn't be the first time that marketing has driven technology.

Re:Serial ATA: The Unnecessary Standard?

1. [...] Explain to me why Apple includes ATA drives inside the G4 when it has an internal Firewire port?

Because ATA drives are currently cheaper than 1394 drives. Not due to any inherent manufacturing or materials cost, but simply because Apple doesn't drive the hard disk market, PCs do. If PCs and Macs ubiquitously had support for 1394, the drives would appear and would ultimately be cheaper than both ATA and SCSI drives.

The same is true for Serial ATA drives. You can bet that the first Serial ATA drives will NOT be cheaper than equivalent Parallel ATA drives. That will only happen as they ramp the volume.

If there was customer demand, they could start ramping production of 1394 drives now. Why should we wait until mid 2001 for them to start ramping Serial ATA? (I addressed this question in my earlier post.)

2a. You do not need topology internal to the computer. So hubs and daisychaining is meaningless.

You need a non-trivial topology any time you want to hook up more devices than you have dedicated ports.

2b. Since four ports of Serial ATA could technically operate in parallel (the bandwidth is not shared), four devices have a combined available bandwidth of 600 MB/s, as opposed to firewire's 400 Mbps (~40 MB/s)

If manufactured in equivalent quantities, hardware support for four 1394 ports (at 3.2 Gbps by 2001) should not cost more than four Serial ATA ports.

2c. The max bandwidth of firewire, 3.2 Gbps, is ~320 MB/s, making it about the same speed as two Serial ATA ports running at the initial speed (which is not the maximum bandwidth)

Sure, but why compare two Serial ATA ports to only one 1394 port? With either interface the per-port cost will be about the same, but 1394 will offer 3.2 Gbps at about the same time that Serial ATA 1X will offer 1.5 Gbps. Probably sooner, in fact.

3. Serial ATA is a point to point protocol. There is no master.

Of course there's a master. The computer is the master; the disk drive is the slave. 1394 supports any mix of masters and slaves. If all you want to do is add one disk drive to a PC, you don't need that capability. But why restritct yourself if you don't have to?

4. Serial ATA is already supported under both windows and linux, past the extent of IEEE1394. This is because Serial ATA uses the same protocol as regular ATA, and thus works exactly the same on an OS level. Unless you look at the chipset, the OS probably will not know it isn't using parallel ATA.

If you want to realize the claimed 1.5 Gbps performance of Serial ATA 1X, you'll probably need a different hardware-level register interface, with different device driver software (at the lowest level; some higher-level driver code may stay the same).

It is true that 1394 hardware interfaces are not register-compatible with ATA ports, but 1394 drivers are already written.

5. ATA is a cheaper solution than Firewire. Currently all firewire drives are actually ATA drives with a ATAFirewire chip inside, so Serial ATA reduces their cost, as well as requires a smaller engineering cost to change the chips on the drive to be native Serial ATA.

Parallel ATA is cheaper today because it is manufactured in high volumes. Neither Serial ATA nor 1394 drives will be cheaper than Parallel ATA until and unless drives incorporating those interface start get manufactured in high volumes. Once the volumes are that high, Serial ATA or 1394 drives should be cheaper to manufacture than Parallel ATA, because they can use chips with lower pin counts, smaller board footprint, and fewer connector pins or edge fingers.

Since both Serial ATA or 1394 offer the potential for higher performance than Parallel ATA at lower cost, but both would cost about the same amount, why not prefer the one with greater flexibility and performance.

What makes you think that Serial ATA is a proprietary interface??

Bad choice of words on my part. What I really meant was "a new and incompatible interface". Why use one of those, instead of a superior existing interface?