Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
Hardware

Serial ATA 1.0 Draft Released 123

Posted by Hemos from the getting-closer dept.
Several readers submitted the news from CNET's story regarding the Serial ATA 1.0 Draft Released. Looks like the replacement for IDE is getting closer - a peak transfer rate of 150 MBs per second is nice to have under the hood.
This discussion has been archived. No new comments can be posted.

Serial ATA 1.0 Draft Released More

Serial ATA 1.0 Draft Released

Comments Filter:
  • What's wrong with SCSI? Can this do something that SCSI cannot?

    Nothing is wrong with SCSI. Actually, SCSI will always be superior to this standard. I guess, however, if you wanted to say that there is something wrong with SCSI, its that it is far too expensive for most people. The reason for this is that SCSI is a much more intelligent device and thus requires much more sophisticated electronics. That is one reason why IDE is so popular, is because it is far cheaper. But with this reduced cost, you are limited to the number of devices you can put on an IDE chain. IDE=2 per chain, SCSI=8 (including controller) per chain. There are some other technical details (disconnect, for example) which make SCSI better, but again.. price.

    --

  • Yes - and it is sooo difficult to simply cut your Ribbon cables w/ an Utility knife, twist one end around once or twice, zip it all up with a cable tye or two - and Voila! you have a 'non-air impeding' cable from MoBo to HD. I dont know anyone who clocked their box who hasnt done this... simple and effective. Have a look at the /. article here [slashdot.org]
  • RTFA -- or maybe C|Net skipped the important details, adn I'm just working off of what I know.

    There is a steady phasing out of the 5V line in PC power supplies in favor of 12V only. More bandwidth -- it is widely believed that ATA/100 was the very last drop that could be squeezed out of parallel ATA -- improved cooling and cabling options, because the 6 or 8 wire cable with be sinificantly less obstructive to airflow, and becuase the cable can be twice as long and more flexible, it can be routed out of the way more easily.

    You might be happy with your SCSI setup, but your wallet probably isn't. I went with an all SCSI rig a couple of years ago -- it performed *better*, but not enough to excuse the price. Wait until you see the $/GB divide of SCSI and IDE grow even wider. Wait until they stop making hard drives for your specific protocol and you're stuck with drives that once were big but are no longer as large as today's entry level drives...

    Shit, I'm late for work.
  • Serial ATA is just another serial connection like USB and FireWire. They're trying to make us use serial ATA hard drives which will slow down when we connect another one to the chain. Support will be decent, fortunately (which isn't the case with USB and FireWire; Intel shuns 1394 in favor of USB2, and USB still has yet to see some non-WinHardware). I don't care if I have to convert all my stuff to Ultra SCSI, I'm just not going to accept this proletarianizing of hardware.
  • Forget Scsi! The idea is smaller cabes that are less prone to block airflow! Cooling is where it's at. Anyone who's built a computer will see why this makes sense I tell ya
  • PCI HD controller
  • I feel sure the manufacturers of the drives will make whatever they think ppl will buy. They don't care so much about other ppls content as long as their drives sell.

    We will see non-compliant drives even if this standard is widely accepted (which it won't be) just like dvd makers now seem to "forget" to strictly apply region code checks.
  • Would someone explain to me how serial is faster than parallel? I mean parallel is just that, sending multiple bits in parallel, serial is sending single bits in series. I would think that sending less information in series would be much slower than sending more in parallel.

    What am I missing?

    -josh
  • Good questions to ask. The article says:
    Double the bandwidth

    Serial ATA offers about twice as much bandwidth as the current Parallel ATA standard, known as ATA 100. Serial ATA's first incarnation, dubbed Ultra Serial ATA 1500, will offer a peak bandwidth or transfer rate of 1.5 gigabits per second.

    "That basically equates to about 150MB of data per second," Ravencraft said. In comparison, Parallel ATA's ATA 100 offers a peak transfer rate of 100MB per second.

    Now Slashdot is linking to stories that are wildly inaccurate! 1.5 != 2, for crying out loud! It's important to note that they tried to make SATA sound twice as fast as PATA, because they know that anything less is doomed to failure.

    Remember floppies? The 1440 KB variety replaced 720 KB only because they were similarly priced, backward compatible at the hardware level, and doubled capacity. Then 2880 failed to gain acceptance because it failed to deliver the same advantages.

    Same thing with the two-sided CD format. The added cost of the hardware will ensure that it is never widely adopted. Unless you can flip the thing over and treat it like a second disk in older drives. Maybe then it will fly.

    As a demonstration technology, it will perhaps be interesting, but until the cost is about the same as PATA, and it gets closer to really being "double the bandwidth", I don't look for any better market penetration than, say, ESDI drives with MCA host adapters.

  • Just think, using this to network a couple computers together.
    Serial ATA Networking...SATAN.
  • Since ATA100 isn't really capable of doing 100 MB/s. =)

    'course, if they are naming this the way they name current ATA specs... then we might just get 100 MB/s out of it.

    --

  • > In addition, Serial ATA will let each drive communicate directly with the processor. Currently, the different drives must share a common connection.

    Finally, a good reason not to have to pay out the wazoo for SCSI, while hopefully doing away with IDE interrupts.

    http://www.mp3.com/subatomicacorn
  • That's almost useless.

    I have 4 IDE ports on the mobo and a PCI controller with another 4 ports. However due to the length limitations of the IDE cables I cannot use all of them. And I have plenty of space in my full tower case.
  • Simple, it's to keep up with those 1.4 GHz processors, which are effectively twice as fast as 1 GHz processors.

  • Encrypted filesystem? (Score:3)

    by TrumpetPower! ( 190615 ) <ben@trumpetpower.com> on Thursday December 21, 2000 @08:03AM (#545364) Homepage
    Just out of curiousity...what's to prevent using an encrypted filesystem on one of these (yet-to-be-created) drives?

    I mean, if all the drive ever sees is a stream of 3DES bits, how's it supposed to know that there's anything contraband going on to it?

    Or would this make encryption an illegal DCMA-circumvention device?

    I don't think we need to worry, just yet.

    b&
  • Which doesn't make much sense, since the whole m.o. of slashdot is that almost every "story" is "stolen" from another source. /. is a news clearinghouse, not a news provider.
  • "Serial ATA offers about twice as much bandwidth as the current Parallel ATA standard, known as ATA 100. Serial ATA's first incarnation, dubbed Ultra Serial ATA 1500, will offer a peak bandwidth or transfer rate of 1.5 gigabits per second.

    "That basically equates to about 150MB of data per second," Ravencraft said. In comparison, Parallel ATA's ATA 100 offers a peak transfer rate of 100MB per second."

    Yeah CNET, glad to see that in the new math system, double 100mbps is 150, not 200. And to think, my whole life I've been lied to. Maybe thats why my boss laughed at me when I asked him for double my pay from 10/hour to 15/hour...

    Go figure...

    -C

  • SCSI can go up to 320MB/s but that's shared among all of the drives on the chain. SerialATA gets rid of the whole chain idea, it's a point to point connection, so while it's only a 150MB/s link, it's not shared between drives.

    You also need to ask, what is the peak transfer speed of the fastest HDD around, right now it's around 60MB/s if I remember correctly, not even enough to max out ATA66, the only time you need ATA100 is if you're using the slave channel. So while 150MB/s seems slow compared to the fastest SCSI connection around, you never have to share it with another device.
    -Ted
  • there is no law _requiring_ copyright control on harddrives, or anywhere

    Currently. Willing to bet we are not going to see one in our lifetimes? After DMCA everything is possible.

    And it's not like the government has objections to outlawing hardware: e.g. scanners able to listen to cellphone frequencies.
    Kaa

  • really? (Score:1)

    by crayz ( 1056 )
    Man thanks for telling me that. I'll turn off my Blue G3(rev 1 even) right now and disconnect my Maxtor 13 gig slave. I'll also have to take the 60 gig maxtor slave drive out of my G4. What a bummer.
  • The only reason IDE is popular is because it is cheap. When you go all the way back to PIO transfers the IDE interface is among the simplest. DMA doesn't make it that much more complicated.

    Data transfer rates aside (the physical transfer
    rates from the media can't keep up, so why the push for faster bus rates?), SCSI still out performs IDE. Give me tagged command queing and Serial ATA MIGHT be competetive, but then the drives won't have the price point, and SCSI has the faster transfer rates, etc.

    Also one thing to note, IDE drives and SCSI drives seem to be using different storage technologies even when you don't consider the transfer mediums.

    In standard half height 3.5" drives SCSI is availible in up to 36 GB (Like my IBM UltraStar 10k rpm drive), but IDE is availible in up to 80 GM (Like the Maxtor I put in my TiVo). The SCSI drive is 10k rpm, the IDE is 5.4k rpm. The SCSI drive was twice the price.

    The controllers aren't the problem with SCSI, my ultra160 dual channel controller was $179, not really bad. The premium on the drive price is another story though.


  • why bother with a new interface that isn't as good as firewire, when firewire is already here, and runs at 400mbs instead of a measley 150mbs??

  • Here's to hoping Firewire wins out this one.

    There are already Firewire drives of comparable price to EIDE and of comparable size.

    Firewire cards, devices, and OS support already exists.

    Firewire is already working over version 2, with plans for version 3, as well as wireless!

    Geek dating! [bunnyhop.com]
  • It was called Firewire. Apple already makes provisions for in in all their current PCs. You can also buy them at prices and sizes comparable to current IDE drives, so the aren't nearly as expensive or high end as serial ATA. The SATA essentially provides for EIDE what SCSI has had for ages, and what Firewire was built with as well; command queueing, daisy chaining multiple devices, and processor decoupling for the data chain. It's just a simpler design, with less wires. Me, I hope Firewire wins this one.

    Geek dating! [bunnyhop.com]
  • I've always wondered why the hdparm manpage says: Using DMA does not necessarily provide any improvement in throughput or system performance, but many folks swear by it. Your mileage may vary.

    It seems to help a lot on my systems.

    The same manpage mentions: It is also a good idea to use the -X34 option in combination with -d1.

    This never fails to crash my systems ;-)
  • I've already made two reply posts to others in the thread.

    Firewire! Go Firewire!

    They are at spec1 in hard disks, notebooks, PCs, cameras, camcorders, and PS2s. Spec2 is already on the way, with designs for spec3, as well as wireless.

    It is comparably priced to EIDE at similar sizes, and already has every single benefit that is being touted for SATA, but now and cheaper ^^

    SCSI is SCSI, and won't be going away any time soon. But here's to hoping Firewire wins!

    Geek dating! [bunnyhop.com]
  • ieee-1394 is 50MB/sec peak transfer rate...

    it has some interesting features, though a monster transfer rate isn't one of them - no addressing, point-to-point connection, etc.

    so, with serial ide, if each device can communicate with the cpu directly is there going to be a limit, like scsi and parallel ide, of one data transfer operation at a time per bus? or will it really be serial (like SSA) and allow multiple simultaneous data transfers?

  • Bus speeds are up, interface speeds are up, memory bandwidth is up, and we're still suffering under meager disk I/O data rates. MAKE DISKS FASTER! Instead of a 40 GB, 3 platter/6 head model, I'll pay the same for a 4 GB, 10 platter,20 head model. Instead of a "stack of records" style, arrange the heads like in a roll of coins. Yes, its more $$ for head design and driver hardware. I want 100MB/sec from the media!
  • I tend to agree with those saying the market will mean non-compliant drives will remain available one way or another.

    For example, I live in Australia. Here, DVD players are supposed to only allow the playing of Zone 2 DVD's (or whatever our zone is). That's so movies can be released earlier in Zome 1 (the US) and not playable by the rest of us.

    1st generation ones were zone restricted according to the standards. Some still are. But, the fact is, the majority of DVD players on the market here now, I believe, have that function disabled before they leave the factory and will play any zone DVD's. The consumer won that one here.

    Sure, this doesn't PROVE much because it's a slightly different situation, but it seems to be indicative of what often happens.

    Darren
  • I just posted another point somewhere in the threads here (have fun looking) but according to another story I found, the Serial ATA Standard will scale to 600MB/s. SCSI Wasn't always 320MB/s, and IDE wasn't always 100 MB/s. You have to keep in mind this is the first bit of the technology.

    Anyone know what Fibre Channel/IEEE 1394 current hard drive standards run at?
  • There is (or was) support for serial SCSI in the specs (the target at the time was high-speed serial links such as fiber, but it would apply to links similar to serial-ata as well). Also SCSI allows for target-mode, which would allow this serial bus to be used as a very-high-speed very-local-network.
  • Why would there be nothing else in the shops?
    How do you thing shops decide what to stock?
    How do you think manufacturers know what to make?

    Answer: Whatever the customers want to buy because that's the only way they can make money.
  • They didn't say that serial ata would be more expensive that scsi, they just suggested that it would be more expensive than ata100. But you're right in a way - how will they get it implemented on high end workstations when SCSI is so much better? Who will they aim the inital tech uptake at? It certainly won't get cheap enough if no-one takes it up. But it's looking like no-one will have a good enough reason to take it up if it can't compete with scsi and raid. What's the deal with serial anyway? I thought parallel was better just on principle?
  • YOU might not want to buy one, but the Great Unwashed will demand nothing else since non-conforming drives won't support the applications that want to use it - RealPlayer10, WMP-2002 etc.

    With the major manufacturers shipping these drives in preconfigured PCs, support will soon be strong enough to permit the applications to be fussy.

    But it will take 2 years to get the drives into the market (not just for sale, but in a fair proportion of users machines). It will then take another year before a new download format has some credibility. Three years of keen cracking is a very long while. The protection has to be strong or they'll have to start all over again with CRAP^H^HMP-2

  • Guess What?

    8 * 100 MB/s = 800 Mb/s
    1500 Mb/s / 800 Mb/s = pretty damn close to 2.
  • Maybe I am picky but what is FireWire's fastest speed? 400 Mbps/50 MBps? The current FibreChannel spec supports much more. The last FibreChannel controller I worked on did around 200 MBps on one loop. FibreChannel can use a hub/switch topology. FC may be a bit storage centric but no more than SCSI is. The only thing I find compelling about FireWire is its price. That is why my main point was that companies should work on making the best technologies less expensive rather than cheaper but lower performing alternatives. In my opinion FireWire and USB are just watered down options. They are cheap but slow. IDE, Serial ports, classic ports, etc. are even cheaper but even slower. How does FireWire not fit into this?
    InfiniBand is looking out to be a great replacment for PCI. It gives you the ability to pull the bus out of the box and also uses a hub/switch network like topology. It supports various speeds and a memory model that lends itself to clustering. While InfiniBand is pretty new, the industry is backing it. Intel already has a Target Controllers ready. Host controllers are coming soon.
    Combined they give you a lot of flexibility. The Infiniband network for all data traffic and the FibreChannel network for all storage traffic.
  • Answer: Whatever the customers want to buy because that's the only way they can make money.

    BZZZZT! Wrong!

    Not when there is a monopoly, it aint. Then they ram whatever they want down your throat. And we are talking about some BIG monopolies here, for starters, Micros~1 and the RIAA and the MPAA etc.

    Look into the 'secure content stream' or whatever is is called in windows (there is a technote somewhere on msdn)

    This could easily be coupled to such a hardware scheme, requiring only 'secured' content to be stored on these devices. Windows may only boot from such a hard-drive, movies may only play on a computer with one etc etc etc.

    WAKE THE FUCK UP!!!

  • Where is a very convincing argument why not use Firewire. It sounds like "$1 per firewire port".
  • Errrr ,,, cos there's nothing else available in the shops ...

    Why should there be nothing else? Re. to my question, why should the consumer (or OEM for that matter) purchase drives with such a 'feature' that may even hinder legitimate backups? No discernible benefit. Besides, consumers have a *right* in most countries to back up software they have legitimately purchased.

    I work for .. errm .. a lickable, fruit-like computer company & have some responsibility for download of customer software bundles to new Ma^H^H computers. I can see a firmware-based mechanism like this causing plenty of technical headaches for systems manufacturers and VARs. So why bother?

  • If I have these right:
    USB maxes out at 1.5 meg a second.
    Firewire tops out at 40 meg a second.
    SerialATA goes to 150 meg a second.
    Parallel ATA (IDE) does 16.7, 33, 66 or 100 meg a second.
    SCSI runs 5, 10, 20, 40, 80 and 160 meg a second.

    Supposedly serial ATA will transparently work with OSs that support parallel ATA. So you can run MS-DOS 5 and Win3.1 on your new 4.3 GHz P5.

  • Abit have had one that supports 8 devices for ages
  • As a clocker with rounded cables, let me summarise for you:
    * cutting and layering stiffens the cable.
    * twisting the cable also shortens it.

    SATA will have no more than 8 wires -- ATA/66+ uses 80 wires, and it's pretty fucking stiff without being rounded (something I don't recommend as a DIY). SATA will have a max cable length of 30 in, which is another foot longer than the maximum for IDE, which also means that it is even easier to route the cable around the airflow.

    SATA is going to be awesome. The transition of technologies will be one of the best since the switch from FPM/EDO SIMM's to SDRAM DIMM's -- trust me.
  • What's wrong with SCSI?

    Price.

    Can this do something that SCSI cannot?

    Yes, provide (much) more bang for the buck.

    Kaa

  • They tried copy protection in the 80's (Score:1)

    by Anonymous Coward
    Both in PC software and in video formats.

    And it was an abject failure.
    The people who wanted to quickly found ways to circumvent the idiot copy protection (hmmm, wait, under DMCA, that's illegal) and the cud-chewing herd just lost their data when the media died.
    These greedy f*ckers don't want to accept that inevitably, information (state secrets, corporate skulldugery, music, movies, software, games) will be free. It's an erosive force, just like water. In the internet age, once one person knows how to break a lock, everyone can know it.

    If something like this somehow makes it to market, all these companies are going to do is make life hell for their less sophisticated customers. Anyone with enough brains to bang two rocks together will circumvent this a week after it hits the streets.

    What really is starting to worry me about the issues swirling around the DMCA, MPAA/RIAA, DeCess, et al is that these corporate interests are attempting to bend our legal system to their ends. In order to line their own pockets, they promote idiot laws, that anyone who's a non-idiot blatently disregards. Just like our ill-advised drug policy, this is another case where we are training a generation to ignore the law, because it is craven and stupid. In pushing such measures, the media industry is further hastening the decay of the rule of law in the American republic.
    Historians will look back to this as a golden age rife with opportunities, and wonder why we allowed things to get so f*cked up.

    Annonymous Coward -- "First they came for fair use, and I didn't speak up because I wasn't a pirate..."
  • The reason SCSI is so much more expensive than IDE is simply a question of target market.

    IDE is targeted to JethroBoBillyBubbaSixPack who is running HoggingDOS98. He is limited by his crappy OS, and a drive that can disconnect from the bus and call back later when it is done with the operation is not going to do him a damn bit of good. He wants big and cheap. Real speed isn't important to him (although he is impressed by high RPM numbers).

    SCSI is targeted at folks building servers, high-end workstations, video editing stations, etc. In other words, people who are spending other people's money (for the most part). They know that being able to do a disconnect and free the bus for something else, being able to to scatter/gather and elevator optimatizations, being able to buffer disk I/O properly, and being able to hang many drives off 1 card are more important.

    As a result, SCSI is more expensive because people will pay more.
  • From Cnet's article:

    The Serial ATA Working Group published the 1.0 draft of its specification, which will effectively double the bandwidth, or capacity for data, between disk drives

    ...and lower down in the same article...

    "That basically equates to about 150MB of data per second," Ravencraft said. In comparison, Parallel ATA's ATA 100 offers a peak transfer rate of 100MB per second.

    When did 150 become (2*100) ?
  • When was the last time you bought a MPAA hard drive?

    Read the story. This abomination is pushed by IBM, Toshiba, Matsushita, and Intel. They (maybe not Intel, but it makes the chips) actually do make drives.

    Besides, the way to market it will be: "Our new drives do all old drives could, *plus* they allow you to buy songs/videos/etc. off the Internet!"

    Kaa
  • I would think that sending less information in series would be much slower than sending more in parallel.

    You would think that, but it works differently in practise. Sending one bit at a time is much, much simpler, therefore you can do it a lot faster. That's the crux of it anyway. The less wires you have the less crosstalk you get, too, another reason you can run the signals faster.



    What I want to know is, what's wrong with FireWire? Is it just not fast enough? What I want is PC hardware that gets rid of all these stupid legacy buses and just gives me USB for keyboards, mice, printers, floppy disks, ZIP disks and syncing PDAs, plus FireWire for hard disks, CD-/DVD-ROM drives and video. Multiples of each bus if necessary. Wouldn't that be a lot simpler/cleaner?



    Cheers, Robert.

  • Why not make an internal serial SCSI spec with smaller cables?

    It exists. It's called "SSA" (Serial Storage Architecture) and it's by IBM and it's quite neat. It is used quite extensively in RS/6000 systems and is an "open" standard. Unfortunately, about a year ago when I stopped working with RS/6000 machines, it seemed that SSA was suffering from Not Invented Here syndrome and the industry was moving to F-CAL. Oh well.

    Cheers, Robert.

  • There is a place for IDE, and a place for SCSI. For Captain Jow-Blow-Win98 at home on his desktop PC, there really is no need for SCSI. IDE is cheap enough for him not to bitch, and it runs just fine.

    I work at a company that is doing unified messaging... We need massive amounts of storage here. If I look out the door of my office, I see some disk arrays that hold a total of about 5 Terabytes of space. You think were going to be running IDE on this? OF COURSE NOT. SCSI is, and probably will always be, 100 times more robust than IDE will ever be. Hot-Swap, 15 devices on one interface as opposed to 2, reliability, RAID 5.

    I mean, who needs that in their own home PC? Not Captain Jow-Blow-Win98 user.. Or even Joe-Blow-Linux user for that matter.

    -Rob
  • Good grief. If any of you had read the microprocessor report article on this, you'd be a lot better informed.

    1. Serial protocol more noise resistant, longer cables at higher freq without signal problems or FCC issues.

    2. Compatible protocol. Should be fairly simple to create drives/chipsets that support both serial and parallel ATA. Host or drive side dongles will also readily available.

    3. Chipsets are getting so integrated that they just can't fit all those pins on them any more. This is becoming problem for creating Tinma style systems.

  • impressed by high RPM numbers

    Yeah, because we all know IDE's highest RPM, 7200, is much higher than SCSI's, 15000... Wait...

    Sure, IDE cannot disconnect from the bus while seeking, but that's not important. Why? People that do IDE right don't put more than one drive on a bus, and you get two for most motherboards. So you buy more cards, or get a motherboard that has 4 (usually 2 are RAID.) And all this is STILL much cheaper than SCSI. And speed? Well, with UltraDMA66 and 100, you can no longer complain about lack of speed on the bus, nor can you complain about the CPU being bogged down controlling the hard drives. It doesn't happen anymore, unless you set the drive in PIO mode. So don't spread your FUD about IDE.

    Yes, SCSI has its place. That place is servers. Even most scanners don't really need SCSI these days, although I admit the better scanners at least support SCSI if they don't require it. It's also true that the fastest drives are only on SCSI. But even for hard core users, IDE is enough.
  • I've submitted this story twice - I can't believe it hasn't been on Slashdot's front page. It has come up on the Linux Kernel Mailing List, and Alan Cox (sort of a "second in command" to Linus, and in charge of the 2.2.x series kernels) has this to say about it:

    > Does anyone have any details on this? I presume that the drive
    > firmware is capable of identifying copy-protected data during
    > a write. I also presume that nobody on lkml would condone
    [Alan Cox:]
    It seems to be very similar to the DVD stuff, including ideas for play once
    only blocks and the like. Pay per read hard disk...

    > such a terrible idea. I imagine that this system is pretty
    > easy to defeat if you can modify the filesystem. Perhaps even

    Its probably very hard to defeat. It also in its current form means you can
    throw disk defragmenting tools out. Dead, gone. Welcome to the United Police
    State Of America.

    > The consequences of being able to corrupt other people's backups
    > by introducing "copy-protected" data are intriguing...

    I'm just waiting for a few class action law suits against drive manufacturers
    when people's backup tools cannot cope


    Serial ATA is old news. BTW, if you were wondering, it looks like Linux will have support for Serial ATA. Andre Hedrick, one of the "senior" kernel developers, is a member of the Serial ATA working group.

    Torrey Hoffman (Azog)
  • That's a fake CmndrTaco generated first post if I ever saw one!
  • These greedy f*ckers don't want to accept that inevitably, information (state secrets, corporate skulldugery, music, movies, software, games) will be free. It's an erosive force, just like water. In the internet age, once one person knows how to break a lock, everyone can know it.

    You know, its people like you who are usually the first to complain though when some corporation wants to sell your personal data or medical information or something.... ironic, isn't it?
  • Linux will be one of the first OS's to have this protocol support..
  • It's CmdrTaco and not CmndrTaco.
  • If I have these right: USB maxes out at 1.5 meg a second. Firewire tops out at 40 meg a second.

    Not quite. USB1 is 1.5MBps and Firewire (1394a) is 50MBps. Newly arrived USB2 can do 60MBps, and IEEE-1394b [computer.org] starts at 100MBps and will go up to 400MBps in a couple years.

    And of course, 1394 supports up to 63 devices per bus, hot-swappable, networkable, yadda yadda yadda. But my favorite Firewire trick (target disk mode [apple.com]) is Mac only. Let's see IDE match that.

  • It provides performance roughly similar to "SCSI-3"; perhaps not quite UW-SCSI, but certainly better than SCSI-2.

    And it takes only about 3-4 wires to use it, rather than the 50-odd wires you find in SCSI and IDE cables. THAT is a pretty big deal; that simplifies the building and layout of systems.

    The really cool part would be to stick a whole lot of Serial-ATA interfaces onto a system; while it would be ludicrous to try to connect 6 SCSI cards to a PCI bus, as the attendant cabling for the potential 90 SCSI drives would be be frightening, and result in a system shaped like the star required to get those cables to go in 6 different directions, the simpler cabling of Serial-ATA might allow such a design to be much cleaner. :-).

    It's certainly not as nice as the latest and fanciest SCSI standard, but if it makes plugging in an extra few drives a cheap and simple matter, that will suffice. SCSI is neither simple nor cheap.

    (Note: The only SCSI variation that I don't have drives for, at this point, is the latest 80MB thing that just doubled "Ultra-Wide" bandwidth again.)

  • What the hell do I care? Y'know how expensive this will be? For the home user it'll be quite some time before they'll be able to take advantage of this new spec.

    Besides, SCSI is still better...
  • Peak is 150? and what about average? 60? 80? Cost of changing hardware? how it will act in RAID? and what for ATA 100 if we already got SCSI? (UW etc...)
  • IEEE-1394b (note the 'b') may well be finalized at their meeting next month. The speeds it will offer are 800mbps, 1600mbps, with extensions to reach 3200mbps, giving you:

    100Meg a second, 200Meg a second, and 400Meg a second. Take *that*, Ultra320 SCSI! :^)

    Also note that unlike ATA, SCSI, and SerialATA, FireWire doesn't require a computer - you can hook devices up directly to each other (digital camera to firewire hd, etc.)

    And for the overclockers/modders - skinny cables with FireWire, too.

    And note to the person who mentioned Apple's FireWire ports on the Mac motherboards - seems Apple has discontinued the internal firewire ports on the latest Macs, from what I hear. Still have the external one(s), though.

  • Re:and supported on Linux :) (Score:4)

    by Ashran ( 107876 ) on Thursday December 21, 2000 @04:35AM (#545413) Homepage
    Umm, even Slashdot Authors should read the article before posting a First Post! ;p
    SATA is software compatible to ATA, which means Every OS that supports ATA has allready SATA support :p

  • Oh, for fucks sake ... WAKE UP!! (Score:5)

    by Vryl ( 31994 ) on Thursday December 21, 2000 @04:37AM (#545414) Journal
    Silly editors, you missed the real story. In "The Register", here [theregister.co.uk].

    Exclusive
    Hastening a rapid demise for the free copying of digital media, the next generation of hard disks is likely to come with copyright protection countermeasures built in.

    Technical committees of NCTIS, the ANSI-blessed standards body, have been discussing the incorporation of content protection currently used for removable media into industry-standard ATA drives, using proprietary technology originating from the 4C Entity. They're the people who brought you CSS2: IBM, Toshiba Intel and Matsushita.

  • Genuine questions. (Score:3)

    by MartinG ( 52587 ) on Thursday December 21, 2000 @04:39AM (#545415) Homepage Journal
    Why is a replacement for IDE being designed now? What's wrong with SCSI? Can this do something that SCSI cannot? I recently changed over to SCSI and I couldn't be happier.
  • 1)What are the current peak transfer rates for IDE and SCSI?

    2)What sort of transfer rate do other technologies like IBM solid state drive, Flash etc have?

    (answers these and I can post inteligently (ish) ;o) )
  • Did someone mention my name ?
  • While the website points out that SerialATA is an internal, in-box protocol and that USB-2 and 1394 (Firewire) are primarily for connection of external devices, they never really address the point that both of these other protocols are also available for use in-box. Apple, for example, has provided 1394 connectors on the MLB to allow F/W drives to be installed. Also, both of these protocols are (largely) already established in the marketplace. So what's the big deal ... ??
  • so there is only one data transer per device per bus... lame.
  • the next generation of hard disks is likely to come with copyright protection countermeasures built in.

    Scary in itself, but how can they provide copyright protection as part of the drive? Surely, that would the job of the file system which, incidentally, is under our control! 8-)
  • Why not make an internal serial SCSI spec with smaller cables?

  • hardware geeks and case modders... (Score:4)

    by complex ( 18458 ) <complex@nOsPAM.split.org> on Thursday December 21, 2000 @04:42AM (#545423) Homepage
    hardware geeks and case modders rejoice, as serial ata uses a skinny litle cable, much like the audio out cable from your cd-rom to your sound card. makes the case a lot neater (imagine hiding the cables by taping them to the sides of the case!) and increases airflow. check out yummy pictures at http://www.anandtech.com/showdoc.html?i=1174 [anandtech.com].

    complex

  • Spec (Score:1)

    by sxpert ( 139117 )
    I have downloaded the spec. (it can be found here) [esitcom.org]
    They change the connectors.
    There will be new style data (that was expected) but also power (not expected) connectors. Thus you'll have to change power supply.
    It seems though that they made the connectors so that devices will be capable of hot-plug (some connections are longer than other, just like for USB. There is also a standard positioning (so that there will be a standard type of disk drawer, building on the "hot plug" apparent capability)
  • Of course the editors wouldn't see that, there seems to be a kill filter on anything that originates from the register.
    Number 9, Number 9, Number 9, Number 9
  • correction the link is this one [esitcom.org]

  • Re:Questions (Score:3)

    by Stormie ( 708 ) on Thursday December 21, 2000 @05:24AM (#545427) Homepage

    Hmm as I suspected this is slower than SCSI and so, given that technology's increasing affordability, rather begs the question: Why are they developing it?

    Is SCSI really becoming increasingly affordable? I'm talking from a position of ignorance here, but I remember that when I shifted from Amiga to PC/Linux (and therefore from SCSI to IDE) about 3-4 years ago, there was only a small premium on the prices of SCSI drives. Now, however, I was just looking at prices the other day (got hold of a Umax scanner which does USB and SCSI, but is only usable under Linux via SCSI.. which made me consider SCSI once more..), and it was horrible!

    I recently bought a 30Gb IDE drive for £125 (an IBM deskstar, not some crappy cheap drive) - the cheapest SCSI drive (according to www.pcindex.co.uk, the UK equivalent of pricewatch.com) is £134 - for 4.5Gb!! To get 30Gb would cost me more than £500 - 4x the price of IDE.

    Now I loved SCSI all those years I used it on my Amiga, it's clearly a more powerful and flexible technology. But I am 100% positive that it will never, ever again be anywhere near price competitive with IDE. I therefore predict that Serial ATA drives will prove to be far cheaper than SCSI, and be an eminently suitable technology for non-server use.

    (addendum: I was just about to post this, and I thought I should check pricewatch.com to see how different the US situation was. 30Gb IDE drives go from around $100 for crappy bands to $130 for IBM. SCSI 36Gb starts at $305. So it's not nearly as bad as the UK, but you're still basically talking double the price)

  • Great! How many devices? (Score:1)

    by Anonymous Coward
    IDE's 2 per chain blows chunks. Macs, which now use IDE to reduce cost, don't even support slave devices. They are one device per controller. Gaawwwdd that's inefficient!
  • http://msdn.microsoft.com/library/psdk/wm_media/wm form/htm/thesecureaudiopathmodel.htm [microsoft.com]

    Have fun!

    Psst! Yes, they *really* are out to get you!

  • i wonder if these SATA kabels could be used as a cheap highspeed network for boxes in a cluster. eventuelly they could not be directly connected via a crossover kable but a small adapter in between that simulates a small harddriver to both (or more ends) could be used for fast internconnection. IPoverSATA anyone?
  • The current limitation of HD transfers is not the connection between the HD and the MB, but it's the speed of the HD read/write heads. Now, given, if there is some caching on the HD, you might see some speed improvement. But how much, really?

    I upgraded my computer last year and moved from ATA/33 to ATA/66. I guess maybe it's sorta faster. Maybe. Or that could be wishful thinking. I even tried some benchmarks, and they looked good. But on the day-to-day use of my computer, how different is it? Am I really booting much faster? Do my applications really start that much quicker? And if the hard drive can't read data any faster, would it do me any good to have a 500 TB/s transfer rate?


  • The increasing sizes of commodity hard drives, and the increasing use of hard drive space, will drive the demand for higher and higher bandwidth busses. Hard drives are doubling in density roughly every 12 months, and doubling in peak deliverable bandwidth every 24 months (since the density of a track increases sqrt(2) every 12 months, causing the head to cover twice as many sectors per revolution of the platter).

    For a good long while now, busses were capable of sustaining much more bandwidth than the drives were capable of delivering, so busses did not have much need to increase in speed. Lately, however, the peak bandwidth of individual drives has been approaching bus capacity, hence the sudden advances in bus technology (ATA/33, /66, /100, UltraSCSI, Ultra160, Ultra320). SCSI is a little ahead of the curve because it has become expected for a single SCSI bus to service multiple drives.

    Eventually I think IDE will have to follow in SCSI's footsteps and go to optical fiber. It's the only way to keep up the pace. SATA is (again, IMO) a transient technology, an interim solution between PATA and FATA, much as VLB was an interim solution between ISA and PCI.

    As to "why IDE?", the answers are pretty much what people have already said -- price (hardware manufacturers already have uber-integrated one-chip solutions for IDE devices and controllers, and the only difference between SCSI and IDE hardware is the logic), and established technology base (ie, everyone's commodity systems software has standardized on IDE). It's like the x86 instruction set, a legacy that we're more or less stuck with for the foreseeable future.

    -- Guges --

  • I guess, however, if you wanted to say that there is something wrong with SCSI, its that it is far too expensive for most people. The reason for this is that SCSI is a much more intelligent device and thus requires much more sophisticated electronics.

    That's true, to an extent, but how much more can it possible cost to stamp out circuit board A than board B? Surely not the $150 or so "SCSI tax" that you'll find on the pricetag.

    No, the real reason is even simpler: the people who need SCSI (people with high-load servers) can afford SCSI. If you can't pay the difference, then you're on a budget so small that you're probably not running a large-scale operation that would require the extra throughput.

    you are limited to the number of devices you can put on an IDE chain. IDE=2 per chain, SCSI=8 (including controller) per chain.

    Actually, make that 16 per chain with modern designs.

  • Seagate has already demonstrated SCSI at 320 meg/s.

    Parallel IDE only does 100 meg/s on a VERY, VERY good day. Most of the time, the difference between ATA66 and ATA100 is quite negligible.

    Supposedly serial ATA will transparently work with OSs that support parallel ATA. So you can run MS-DOS 5 and Win3.1 on your new 4.3 GHz P5.

    Now why in god's name would anyone want to do this? If this is the only other redeeming feature of serial ATA aside from the promised speeds, send it back.

    I'll thank these bastards not to control what I can store on my own machine.

    --

  • I bet that when these come out, most of the systems running this will not be running the correct device drivers.

    I've lost count of the number of UDMA capable machines I've seen running in PIO mode under both Linux and NT. Not only are data-transfer speeds lousy, but CPU utilisation during data-transfer is obscene. Then they wonder why their applications are running slowly.

    Clueless MCSEs appear to do this all the time!

  • My question from the beginning is how much more expensive is a HD SCSI controller than an IDE controller?

    Well, part of the problem with SCSI controllers is the undeserved worship that a few manufactures get. A Tekram DC-390F is at least as good as an Adaptec 2940, but the powers that be have made sure that everyone only buys the most expensive cards on the market.

    I still defer to my original explanation for why the average SCSI controller is more expensive than a similar IDE controller: if you need it, you can afford it.

  • Now all the motherboards need is support for more than 4 ide devices. Has anyone seen some of the nice gigabyte [giga-byte.com] boards that are being produced? Its nice to finally have support for more devices. Now if they made these boards to support the new standards, I would be in heaven.
  • Peak transfer rate for IDE I believe is 100 MBs (ATA 100). SCSI I think is 320 MBs (Ultra 320). With a RAID Array you can do even better. And if I'm wrong on the numbers, I'm sure the Slashdot fanatics will correct me :)
  • the scsi standard has been around forever. why are they trying to recreate the wheel. sure scsi is expensive, but this ata 1.0 stuff isn't going to be cheap. if as much money, effort, and time had been spent on getting people to use scsi disks as has been spent on ide crap, then the price for scsi devices would be compairable to that of ide.

    use LaTeX? want an online reference manager that
  • Traditional SCSI is not the right replacement for IDE. It has exactly the same problem: the use of multiple signal wires causes electrical problems (interference, synchronization). It turns out that you can do better by using a serial rather than parallel connection. Now you might say that we should go for serial SCSI, and that's basically what firewire is. Search for "serial SCSI" in Google.
  • No we are not. We are talking about drive manufacturers and retailers. They make and sell the drives that people buy.

    When was the last time you bought a MPAA hard drive?

    If you are worried about your operating system refusing to work with existing drives then dont use an operating system that doesn't give you the choice.

  • May not be as bad as that, check this out... (Score:4)

    by Fross ( 83754 ) on Thursday December 21, 2000 @05:34AM (#545452)
    Microsoft for one, as stated in the article, is seriously opposed to this. So should be any OS or large application vendor, for the reasons described therein. They do have some big clout. :)

    Not all the HD manufacturers are in this little committee, and I'm sure they'll do whatever will make them more money. If this means going against the committee and either persisting with ATA-100 or making their own version, if there is a market for this product, they will do it.

    Now this is _very_ important - New standards need new chipsets - the interface for ATA is held on the motherboard. Intel is on this committee, sure, but what about AMD? VIA? Now, they wouldn't want to do something differently from Intel, would they? ;) Without a chipset with these new ATA commands in it, the drives are useless. So here's a weak point.

    I'm sure there'll be workarounds built as quickly as DeCSS was. Cables with de-protection dongles built in? Nice one.

    There are more potential problem scenarios for this. If file X is "protected" in this manner, what about X.zip? :) What about downloading stuff off the Net? I'm sure programs will also be used to strip copyright information off a file as well, or some apps will be written to do filecopying by getting around this.

    (this is a tangent from this - the new ATA standard would be a hardware check on the HD, i know, but from the article i gather it would be effectively on request of the file/application)

    In short, I would keep a close eye on it, but I wouldn't worry. If it becomes a reality (as in, a major choice and in stores), tools will be all over Warez sites. Think of the volume of DeCSS out there, now multiply it by the number of Windows users there are - and this would be a hack for something REALLY annoying.

    It'll be tried and will fail - there is no law _requiring_ copyright control on harddrives, or anywhere, so the fight will be on, with people boycotting the new drives and sticking with ATA-100 or SCSI-UW, or hacking them. Until the next standard comes along, which won't have this control on it, and will sell like hotcakes.

    Get our yer mirror sites.

    Fross
  • Man, I so seriously hope that you are right. This bullshite stinks, and I think you are correct, we have to boycott the 'MPAA Hard-drives'.

    But ... my point was ... like you know, you buy computer with a well, basically, MPAA DVD drive. It refuses to work without a 'MPAA Hard-drive', so none of the HD manufactures make 'non compatible' drives, and we are screwed.

    It is a plausible scenario, and must be fought somehow.

    Hopefully, one day this era's fascination with attempting to own information (basically, long numbers) will be looked back on with derision, scorn and probably mirth.

  • Re:How is serial faster than parallel? (Score:4)

    by Azog ( 20907 ) on Thursday December 21, 2000 @09:10AM (#545456) Homepage
    Serial is faster than parallel because they can crank the clock speed way, way up on the bus. You can't do that with parallel because you end up having major problems keeping all the data pulses properly synchronized. Also, with more signal cables you need more grounding cables - that's why the Ultra 66 and 100 drives need 80-wire cables to work properly. And even then, the ATA-100 stuff only really goes that fast if everything is just right and the moon is in the right phase (or so I understand.)

    Serial cables are just much simpler electrically, even though the clock speed has to be 16 times higher for the same bandwidth.

    Disclaimer. I am not an electrical engineer. I just read stuff off web sites.
    Torrey Hoffman (Azog)

  • Re:Genuine questions. (Score:3)

    by Just Some Guy ( 3352 ) <kirk+slashdot@strauser.com> on Thursday December 21, 2000 @09:14AM (#545457) Homepage Journal

    If you have looked at a modern IBM drive lately, and sat a SCSI and IDE next to each other, the only differece is in what chips are used.

    I've heard (note: rumor!!!) that hard drive mechanisms go threw the same winnowing process as PentiumIIs used to. That is, the ones that pass a very high testing threshold are reserved for sale as SCSI units, while lower-performing samples become IDE. While one unit theoretically should be identical to its neighbors on the assembly line, small variations in the manufacturing process make for better and worse results.

    Therefore, you'll pay more for SCSI drives, partially because their mechanisms are of a higher quality to begin with.

    Can anyone provide evidence for or against this?

  • SCSI is overkill for home computers (Score:3)

    by RayChuang ( 10181 ) on Thursday December 21, 2000 @06:45PM (#545462)
    Folks,

    The reason why Serial ATA is being developed is simple: going to SCSI--especially Ultra2-Wide and Ultra 160--is a VERY expensive option.

    Have you seen the cost of Ultra 160 SCSI adapters? Or how much the cabling costs? Or how much Ultra 160-compatible SCSI drives cost? Pretty expensive, and no thanks.

    Serial ATA will of course initially cost more than UDMA/100, but it still would be much less expensive to implement than Ultra 160 SCSI. And because Serial ATA does not use those pesky flat ribbon cables, installation is also much less of a hassle, especially now you have much less interference with interior air flow of a system case, which will actually promote longer life of computer components. The best thing is that Serial ATA does not require a drastic change in the operating system to support it out of the box other than getting motherboard chipset drivers for the South Bridge chip that has the Serial ATA support, and given that most motherboard manufacturers include a CD-ROM disk with these drivers as standard....

  • Re: WAKE UP! (Score:3)

    by Splat ( 9175 ) on Thursday December 21, 2000 @04:52AM (#545464)
    From The Register's story:

    "Where were you when they copy-protected the hardware, Daddy?"

    "Well son, way back in the year 2000, we had these wonderful devices called CD Recorders. We could copy almost any CD! Except for evil CDs with defective sectors burn into them, they required a little more work my lad. And we had these great devices called IDE Hard Drives. You could store 80GB of anything you wanted on one of these disks! And copy it to any drive!"

    "80GB? You're old dad! How many gigabytes is a terrabyte again? Johnny down the street just got a new Fiber Channel 740TB Drive. Can I get one for the holidays?"

    "We'll see, son. I'm liking the looks of the new nanodrives. But anyhow, continuing my story. We had all sorts of magical hardware. We had Orb Disks, which held 2.2 GB! And these things called zip disks, which held 250 or 100MB!"

    "100MB?! Ha ha dad. I can't even fit a Word 2044 Document on that!"

    "You think 100MB is small? Wayyy back in the day we had things called FLOPPY DISKS. And they only held 1.44MB, 720K, 360K, or .. EVEN LESS! But they were wonderful devices, these floppy disks. We could make as many copies as we wanted as one, and store whatever we wanted! Oh those were the days, when the MP3s flowed free, and the DeCSS rebels weren't laughed at by the public. Yes. Those were the days."
  • Why can't these companies concentrate on making the good technologies out there cheaper? Enough of these watered down attempts like USB and FireWire. Enough with PCI and PCI-X. We know we have things that are better. I am so tired of good hardware costing so much more just because the vendors can get away with charging more. They claim you are paying for performance, development costs, etc. but I think you are paying for a lack of real competition. All these companies that make SCSI drives yet prices in this market didn't fall like the IDE market? Does it really cost that much more to build a scsi drive than an IDE one? What about a fibre channel controller and drives vs. similar SCSI controller and drives? It is unbelievable!
    By the by...although InfiniBand is pretty new it has great potential. It makes a great platform for both storage and network traffic. I am currently working on drivers for storage (fc) and network (GB ethernet) devices over InfiniBand. And our first target platform is...Linux!
  • Also, why should I buy one of these drives if there is no discernible benefit to the consumer & the non-knobbled drives are still available?

    Errrr ,,, cos there's nothing else available in the shops ...

Slashdot Top Deals

A morsel of genuine history is a thing so rare as to be always valuable. -- Thomas Jefferson

Close