Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

[P]
Serial ATA is here

By Work in Technology
Tue Aug 13, 2002 at 05:01:49 AM EST
Tags: Hardware (all tags)
Hardware

Everyone who has put together their own computer knows and curses what has been a staple of building one for over 20 years: the ribbon cable. Inventions such as the rounded IDE cable have helped, but are not a good solution.

Very soon now, these days will be no more.


While within the last couple of years some companies have made novelties such as rounded IDE cables a reality, however they are still thick, stiff and relatively difficult to work with. They also still have the 40 cm maximum length limit which make mounting drives in a tall case difficult.

Tom's hardware has a spectacular article on the upcoming serial ATA standard (current ATA is parallel).

Some highlights of the upcoming transformation are:

  • Much smaller cables and connectors. There are only 7 wires, the connectors are 8 mm wide.
  • Faster. The first generation motherboards will still probably use the PCI bus, leading to a max transfer rate of 133Mb/s, however serial ATA has a potential rate of 150Mb/s with future revisions possibly achieving up to 300-600Mb/s.
  • Compatible with current parallel IDE drives and motherboards via adapters.

    What I look forward the most to is the flexibility this will allow in case design. The trend toward small, quiet PC's has been marred by the difficulty of cooling them. Stiff wide ribbon cables murder good airflow through any case. The change to these small and simple cables will allow much more flexibility in drive placement, cooling options and overall case design.

  • Sponsors

    Voxel dot net
    o Managed Hosting
    o VoxCAST Content Delivery
    o Raw Infrastructure

    Login

    Related Links
    o spectacula r article
    o Also by Work


    Display: Sort:
    Serial ATA is here | 112 comments (82 topical, 30 editorial, 0 hidden)
    Jerry Pournelle (4.00 / 1) (#8)
    by wiredog on Mon Aug 12, 2002 at 02:56:24 PM EST

    In his latest Chaos Manor column he talks about the usefulness of Serial ATA for video production.

    Can't sleep. The clowns will get me.
    "faster" is red herring (4.00 / 4) (#10)
    by RelliK on Mon Aug 12, 2002 at 03:05:41 PM EST

    however serial ATA has a potential rate of 150Mb/s with future revisions possibly achieving up to 300-600Mb/s.

    And the point of that is? The current top of the line hard drives can barely achieve a sustained read trasfer rate 50 MB/s (write is slower). Only one device at a time can work on an IDE channel. That means that anything above ATA66 is nothing more than a marketing gimmick. It reminds me of this ridiculous article which was linked on slashdot a month ago where the reviewers compared ATA100 and ATA133 HDs and were surprised ATA133 was not any faster.

    Serial ATA does not make any progress towards concurrent operation of the devices on the same channel. On the contrary, serial ATA takes a step backwards by allowing only one device per channel, period. Your glorified 150MB/s are simply wasted because no HD in the world can transfer data that fast.

    The one advantage of serial ATA that you failed to mention is that it is, supposedly, hot pluggable. It remains to be seen how well that will work out, but if all goes well, serial ATA will become the new standard for external devices.

    One other advantage of serial ATA is that it's backwards compatible with parallel ATA.

    In short, serial ATA is generally a good thing, but it's overhyped beyond belief.
    ---
    Under capitalism man exploits man, under communism it's just the opposite.

    The point of faster. (4.00 / 2) (#11)
    by steveftoth on Mon Aug 12, 2002 at 03:11:37 PM EST

    Right now, there isn't much point to having a really fast single connection point. But don't forget about RAID, and the fact that speed is ever increasing. They are just saying that they are prepared for the next generation of speed increases and that this design will be able to handle it. My only real hope for Serial ATA is that the chips implementing it don't burden the CPU much. One HUGE problem with USB is that if you come close to maxing out the bus it will use a lot of your CPU to run the USB controller. Try burning a CD with a USB cdrom drive and you will see what I mean. SCSI has always been the best solution for those with limited main CPU because the controllers are smarter. Though they are also more expensive.

    [ Parent ]
    did you read what I wrote? (3.00 / 3) (#13)
    by RelliK on Mon Aug 12, 2002 at 03:36:52 PM EST

    But don't forget about RAID

    To have RAID you need multiple drives. Serial ATA supports only one drive per channel. What is the point of having 150, 300, or 600MB/s channel if the fastest HDs can barely sustain 50MB/s?

    One HUGE problem with USB is that if you come close to maxing out the bus it will use a lot of your CPU to run the USB controller.

    Of course. USB was designed that way. If your machine becomes slow while you're burning a CD, then a certain CPU manufacturer can convince you to buy a faster CPU. Why do you think said manufacturer was pushing USB so hard when it is plain obvious that FireWire is superior?
    ---
    Under capitalism man exploits man, under communism it's just the opposite.
    [ Parent ]

    RAID (5.00 / 3) (#16)
    by marx on Mon Aug 12, 2002 at 03:57:22 PM EST

    You can make a RAID device which looks like a single IDE device to the ATA interface, but contains several disks inside. You can then use striping (i.e. alternating reads and writes) to multiply the speed by at least a factor 2, possibly more.

    Join me in the War on Torture: help eradicate torture from the world by holding torturers accountable.
    [ Parent ]

    And how exactly do you do that? (1.00 / 4) (#32)
    by RelliK on Mon Aug 12, 2002 at 06:24:03 PM EST

    You can make a RAID device which looks like a single HD to the applications, but at some point you still need to connect multiple HDs to the board. Serial ATA supports only one device per channel, i.e. it is physically impossible to connect multiple devices to the same channel. To make a serial ATA RAID you have to connect several HDs to several different channels.

    Not only is your point moot, it is simply stupid.
    ---
    Under capitalism man exploits man, under communism it's just the opposite.
    [ Parent ]

    Hmm... (none / 0) (#40)
    by Danse on Mon Aug 12, 2002 at 07:18:08 PM EST

    Just a wild ass guess here, but couldn't you have a RAID controller with several Serial ATA channels (3 or more, including a channel between it and the motherboard) that would deal with the work of being the intermediary between the drive array and the CPU?






    An honest debate between Bush and Kerry
    [ Parent ]
    ATA RAID (5.00 / 3) (#43)
    by marx on Mon Aug 12, 2002 at 07:44:58 PM EST

    To make a serial ATA RAID you have to connect several HDs to several different channels.
    No, now you are mixing up the interfaces. For simplicity, let's say we have a RAID box to which you can connect parallel ATA drives. You then connect this box via serial ATA to the computer. You can then have 4 times the transfer speed of a single drive over the serial ATA interface.

    The same principle applies to SCSI RAID boxes which have IDE drives inside them. Check out this one for example.

    Another solution could be having multiple serial ATA interfaces inside the RAID box. I don't know if that would be cost effective though.

    Not only is your point moot, it is simply stupid.
    And your mom is a bitch.

    Join me in the War on Torture: help eradicate torture from the world by holding torturers accountable.
    [ Parent ]

    I guess you missed the point (1.80 / 5) (#48)
    by RelliK on Mon Aug 12, 2002 at 10:24:57 PM EST

    The point is though that ultimately you need to connect those drives to something inside that box. How do you do that? If you use serial IDE, you need to have a separate channel for each drive.(*) That the drives may then be (essentially) multiplexed onto the same channel is beside the point. The whole reason I posted this message is to dispel the "its soooo fast" myth. Ultimately serial ATA supports only one drive per channel (even inside your glorified box), so 150MB/s bandwidth is wasted considering that today's drives can barely sustain 50MB/s. For the same reason, ATA100 and ATA133 are nothing more than marketing gimmicks.

    (*) Realistically, even for standard IDE RAID, you need to have a separate channel for each drive, since only one device per channel can work at a time. This is what the folks at 3ware are doing.
    ---
    Under capitalism man exploits man, under communism it's just the opposite.
    [ Parent ]

    No, *you* are missing the point (5.00 / 1) (#58)
    by curien on Tue Aug 13, 2002 at 06:11:18 AM EST

    The point is though that ultimately you need to connect those drives to something inside that box. How do you do that? If you use serial IDE, you need to have a separate channel for each drive.

    So what? Who cares what the RAID box uses inside? The point is that you can now connect that box to your computer via IDE, and the interface won't be a bottleneck. This is just one more example of IDE encroaching on a SCSI-dominated niche.

    --
    Murder your babies. -- R Mutt
    [ Parent ]

    Problem with that. (4.00 / 1) (#60)
    by crankie on Tue Aug 13, 2002 at 06:18:18 AM EST

    S-ATA is specifically not designed for outside-the-box connections. It is not a replacement for USB or similar systems. The cables are designed to be up to 1m long, so connecting to a loud angry RAID array is not really going to happen.

    Consider this though:
    S-ATA RAID array, populated with cheap S-ATA disks and controllers, connected through a bridge chip, to fibre channel. Which can run over several kilometers. Nice.

    Also, will people please stop dissing PCI. If you're worried about bandwidth, you're probably running a server. PCI comes in 32bit/33MHz, 64bit/33Mhz, 64bit/66MHz. That last one can do 600 megs per second.

    ~~~
    "The great thing about hardcore socialists is the silence they emit once they start earning a decent wage." - tombuck
    [ Parent ]
    Real world (5.00 / 1) (#69)
    by Miniluv on Tue Aug 13, 2002 at 11:56:04 AM EST

    SCSI also isn't really intended for outside the box connections, however it gets used regularly. 1M is plenty of cable to connect from my 1U server to my 4U 1TB RAID enclosure. This means I can buy a SCSI raid enclosure and not waste $400+ on a dependendable SCSI host adapter to merely relay SCSI commands on to my RAID controllers in the enclosure.

    If you think the above isn't a real world scenario, then you need to do some research. I've implemented solutions just like the above, except that I did use SCSI between the host and the enclosure as serial ATA isn't really widely available yet, and certainly wasn't then.

    There's very little point in making single ultrafast drives, since the only people stupid enough to buy single ultralarge drives aren't going to pay the additional price burden of also making them ultrafast. Instead, the enterprise storage market will continue to buy the right ratio of performance to size to price to MTBF and just chain them together on multitasking bus setups.

    "Too much wasabi and you'll be crying like you did at the last ten minutes of The Terminator" - Alton Brown
    [ Parent ]

    Not really. (4.50 / 2) (#72)
    by crankie on Tue Aug 13, 2002 at 12:17:54 PM EST

    LVD SCSI can run at cable lengths of up to 35 feet. Section 6.3.6 of the S-ATA 1.0 spec stipulates 1m as the maximum cable length. So while yes, there are SCSI RAID enclosures available, and yes, they can be connected by 1m cables, they can also be connect by 35 foot cabling. Just because I have a 50cm HSSDC fibre-channel cable doesn't mean I cant have a 10km optical cable. This is not true of S-ATA, making it substansially less flexible for outside-of-the-box connections.

    So in summary, yes you could do an entire system using S-ATA, but I wouldn't want to. Yes, you could do the front-end using SCSI, for smaller systems because SCSI is not comparable with S-ATA. But for a wider area system, using FC to connect to a switch within an FC fabric setup is, to use a technical term, "da bomb".

    ~~~
    "The great thing about hardcore socialists is the silence they emit once they start earning a decent wage." - tombuck
    [ Parent ]
    Sure (4.00 / 2) (#76)
    by Miniluv on Tue Aug 13, 2002 at 02:55:09 PM EST

    But going with FC and a SAN switch is not nearly as easy as calling up one of dozens of vendors and buying a SCSI or ATA raid enclosure and then using a serial ATA cable to connect it.

    I've done a decent bit of storage work, and still never run into a situation where a SAN couldn't be conveniently avoided, nor have I run into a situation where it even remotely made sense to spend the tens of thousands of dollars necessary to get the equipment, cable, etc necessary to begin contemplating hooking up drives.

    SANs are great, however only for a small subsection of the market for large storage servers. This would be why the switches still cost a couple grand per port, and the volume of sales has never taken off.

    "Too much wasabi and you'll be crying like you did at the last ten minutes of The Terminator" - Alton Brown
    [ Parent ]

    Ethernet is to FC what SATA is to SCSI (5.00 / 1) (#83)
    by MfA on Tue Aug 13, 2002 at 04:41:43 PM EST

    FC and SCSI are just going to get marginalized more and more and their cost disadvantage will keep growing ... wether at some point the cascade this causes will stop remains to be seen.

    Apart from cable length, which is really not much of an issue, SCSI has no fundamental advantage over SATA which cant be solved by a smart controller. Try a 3ware controller for instance, with the present level of integration such smart controllers really arent a cost factor anymore ... the pin counts needed for an ATA channel were probably more of an issue as far as costs were concerned, and SATA will solve that.

    Yes due to marketing decisions SCSI drives have higher reliability and lower access times, but reliability can be compensated for by more redundancy and access time by simply using more drives ... and with the present price difference between SCSI and ATA drives thats not as bad a proposition as it seems.

    Gigabit ethernet is getting commoditized ... this will make life very hard for SAN. I think NAS will pretty much take over, and that that will pretty much be a SATA only affair.

    [ Parent ]

    Growing disk sizes (4.00 / 1) (#85)
    by sgp on Tue Aug 13, 2002 at 08:01:50 PM EST

    Reminds me of a situation a colleague came across a while ago - the customer insisted on striping their 10GB database across, IIRC, at least 6 disks. That's what they'd done on the old system, that's what they'd do with the new system.

    Since the smallest disks we could get were 9.1 GB, that's a 546% storage capacity waste. The fact that the new 9.1 GB disks were significantly faster than the old ones (although they didn't know the speed of the old disks, being around 1GB in size, they'd presumably be well under 7200RPM) didn't give the customer a problem.

    They ended up happily buying far more storage than their database would ever need, just so they could stripe them.

    There are 10 types of people in the world:
    Those who understand binary, and those who don't.

    [ Parent ]

    Cost effective. (5.00 / 2) (#59)
    by crankie on Tue Aug 13, 2002 at 06:12:35 AM EST

    "Another solution could be having multiple serial ATA interfaces inside the RAID box. I don't know if that would be cost effective though."

    Short answer: Very cost effective.

    At the moment, most raid systems use SCSI or Fibre Channel, or indeed, SCSI over Fibre Channel. FC disks are very expensive. S-ATA disks are aimed at being in the same price range as regular ATA drives and also hot-pluggable. This makes them an excellent candidate for raid systems.

    ~~~
    "The great thing about hardcore socialists is the silence they emit once they start earning a decent wage." - tombuck
    [ Parent ]
    FW is not always the better choice. (none / 0) (#23)
    by steveftoth on Mon Aug 12, 2002 at 05:02:22 PM EST

    FW is good, but it's not the best. Mostly because FW is more expensive. The interface chips for FW are much more expensive. This might not be the case if everyone used it but that's the way it is now. IT doesn't make much sense to put a $5 + chip in a $10 keyboard. USB makes sense when dealing with slow devices, the new USB is fast, but the design flaws in it will cause it to use too much cpu, and you still won't be able to do a device to device transfer of data.

    [ Parent ]
    Did I claim FW is best? (5.00 / 1) (#34)
    by RelliK on Mon Aug 12, 2002 at 06:29:03 PM EST

    I simply pointed out that it is more suitable than USB for transfering a lot of data. Also, design flaws in USB are not flaws per se. USB was intentionally designed to be a CPU hog so that, as I already pointed out, they can sell you a faster CPU.
    ---
    Under capitalism man exploits man, under communism it's just the opposite.
    [ Parent ]
    One advantage? (none / 0) (#12)
    by KnightStalker on Mon Aug 12, 2002 at 03:24:09 PM EST

    I'm just regurgitating this from Tom's article...

    Two other advantages are thin, flexible cables and 1-meter length restrictions instead of 40-cm. This is worth a lot.

    I don't know if current ATA specs support it, but Serial ATA also adds tagged command queuing.

    It would be really nice if there were support for more than two devices per cable, though.

    [ Parent ]

    one advantage you failed to mention (none / 0) (#14)
    by RelliK on Mon Aug 12, 2002 at 03:40:54 PM EST

    I pointed out the thing you failed to mention. What's the point of repeating what you already mentioned in the article?

    It would be really nice if there were support for more than two devices per cable, though.

    I don't know what you are talking about. Serial ATA supports only one device per channel.
    ---
    Under capitalism man exploits man, under communism it's just the opposite.
    [ Parent ]

    That's even worse (5.00 / 1) (#18)
    by KnightStalker on Mon Aug 12, 2002 at 04:24:03 PM EST

    I was interpreting this paragraph:

    RocketRAID 1520 is the name of the latest serial ATA controller from HighPoint. It offers two channels, each of which controls one or two hard disks.

    which is under a picture of a controller with only one plug per channel, to mean that you could have two devices per plug. You seem to know more about it than I do. Is there an adapter that allows two cables per connector on the card, or is Tom just wrong?

    [ Parent ]

    in the same article (none / 0) (#37)
    by RelliK on Mon Aug 12, 2002 at 06:56:52 PM EST

    ABIT sent us the IT7-MAX2, which is fitted with two serial ATA connectors.... Two Marvel bridges provide the serial ATA interface. These use two of the four IDE channels of the HighPoint HPT374 controller. The IT7-MAX2 can therefore handle eight conventional IDE devices, as well as two serial ATA devices.
    ---
    Under capitalism man exploits man, under communism it's just the opposite.
    [ Parent ]
    Ok... (5.00 / 1) (#42)
    by Danse on Mon Aug 12, 2002 at 07:36:27 PM EST

    After reading the paragraph that KnightStalker quoted, as well as the one that you quoted, it seems that there is conflicting, or at least very confused information in the article. The HighPoint controller offers 4 IDE channels. Each channel can support 2 regular IDE devices. But it also says that it can handle 8 conventional IDE devices, as well as 2 serial ATA devices. Since the 2 Marvel bridges are using 2 of the 4 IDE channels of the HighPoint controller, does this mean that a serial ATA device can co-exist with 2 regular IDE devices on the same channel?

    Then there's the fact that the paragraph that KnightStalker quoted seemingly directly contradicts the paragraph that you quoted. I'm not sure that's so, because the part that you quote doesn't explicitly say that Serial ATA only supports one device per channel. It just says that that particular board only supports 1 per channel. Since the other paragraph claims that 2 devices are supported by each channel, what's the real answer? Is it just a difference in the controllers?






    An honest debate between Bush and Kerry
    [ Parent ]
    re (5.00 / 1) (#47)
    by RelliK on Mon Aug 12, 2002 at 10:05:28 PM EST

    Since the 2 Marvel bridges are using 2 of the 4 IDE channels of the HighPoint controller, does this mean that a serial ATA device can co-exist with 2 regular IDE devices on the same channel?

    No. That board has 2 non-raid ATA channels in addition to the built in Highpoint controller which has 2 additional channels and 2 serial ATA channels.

    Since the other paragraph claims that 2 devices are supported by each channel, what's the real answer? Is it just a difference in the controllers?

    No. Serial ATA supports only one device per channel. That's not a big loss though, since on the standard IDE only one device per channel can operate at a time (the other device has to wait for its turn).
    ---
    Under capitalism man exploits man, under communism it's just the opposite.
    [ Parent ]

    Breaking 50Mbps (none / 0) (#66)
    by tjw on Tue Aug 13, 2002 at 10:35:49 AM EST

    Your point about current drives not being able to sustain reads at over 50Mbps is quite valid. Mechanical hard drives are slow, noisy, and prone to failure.

    I don't understand why there aren't more SSD (Solid State Drive) solutions out there. I'm not talking about DiskOnChip or other slow EEPROM solutions.

    What I want is an affordable solution that uses commodity SDRAM. What I want is a product similar to Cenatek's Rocket Drive, but instead of interfacing through the PCI bus using special operating system drivers, it should be hidden behind a IDE, SCSI, or Serial ATA controller.

    I know what you're thinking, SDRAM doesn't remember it's state during a power failure rendering it useless for this task. Although that's true, this limitation could be countered by including a battery backup.

    So here's your task Manufacturing Industry, take one of those Rocket Drives, attach an IDE controller and a rechargeable Lithium Ion battery and stick it into a 5.25" FH drive enclosure. You sell a million, I get 133Mbps transfer speeds.

    [ Parent ]
    FH vs HH (none / 0) (#75)
    by Mitheral on Tue Aug 13, 2002 at 01:07:38 PM EST

    If they could fit it in a half height drive it'd be even better, 5.25 drive bays are lacking on most modern cases.

    [ Parent ]
    RAM prices keep dropping... (none / 0) (#86)
    by sgp on Tue Aug 13, 2002 at 08:07:10 PM EST

    This has been suggested so many times, I've "had the inspiration" myself before I found that others have also been inspired...

    What I don't understand, is why this hasn't happened yet. It'd be more expensive than a hard disk, but fast enough to sell... a RAMdisk-in-a-box as a plug-in ultra-fast storage solution. Steve (all hyphened-out)

    There are 10 types of people in the world:
    Those who understand binary, and those who don't.

    [ Parent ]

    you're kidding, right? (none / 0) (#98)
    by RelliK on Wed Aug 14, 2002 at 05:37:35 PM EST

    First of all, even if such a solution were implemented, it would not be cost effective. You can buy an 80GB HD for the price of 0.5GB of RAM.

    Secondly, such a solution would be simply stupid. If you need faster access time on a file server, database, etc. you can just add more RAM and thus improve the cache hit ratio. Oh, and you can access the cached data directly over a 2100MB/s bus (even greater for DDR333, DDR400, Rambus, etc) instead of going through 133MB/s PCI bus to 150MB/s ATA channel to RAM.

    With a UPS you can make your own RAM disk, but it's really pointless. Nobody does that any more since it's more efficient to let the OS/DB handle the caching and permanent storage rather than try to reinvent the wheel, badly.
    ---
    Under capitalism man exploits man, under communism it's just the opposite.
    [ Parent ]

    Ram Drives (none / 0) (#99)
    by wnight on Wed Aug 14, 2002 at 09:03:17 PM EST

    It's actually a very niche product. With a decent OS you'll get better performance by using the ram in the drive as real RAM, instead of a drive. The OS can choose what to cache on the fly and will almost always do a better job.

    Ram drives are best suited to devices that have to store a lot of data in a burst and yet don't want to implement memory buffers, or single-tasking embedded systems that can't have moving parts.

    With a dumb OS (Dos 6.x for instance) it can make a huge difference. I used to have a script that made a ram disk on the fly, copied Doom2 into it, and launched it. When I quit it would update the zip with any config files that had changed. With a real OS you'd be better off simply pre-caching the game before loading it.

    One other application is if you require very low-latency response. You might want to put a 3d game on a ram drive because you're willing to sacrifice the performance of the system as a whole to make sure that your level load time is as low as possible.

    [ Parent ]

    Write is not slower (none / 0) (#104)
    by apm on Fri Aug 16, 2002 at 03:35:46 AM EST

    Actually, write is slightly faster than reads on modern drives because they cache the commands, freeing the drive to receive the next command while the first one is writing. Most drives will also reorder cached write commands in firmware to save time on seeks, rotations, etc. (Tag command queueing brings that same ability to reorder to read commands as well.)

    Also, while most drives transfer at 50 MB/s or less when reading from disk, they can return data substantially faster if the data is in the buffer cache, which happens pretty often if the OS reads the same blocks all the time (like, say, the FAT or volume bitmap on a Mac system).

    [ Parent ]

    Not quite (none / 0) (#111)
    by Salamander on Thu Sep 26, 2002 at 11:33:08 AM EST

    While what you say about writes being cached is true, it doesn't totally refute the other person's claim. Write caches are finite, so if a host tries to sustain writes at a rate higher than the disk or storage subsystem's internal bandwidth you get stalls and the extra external bandwidth really is wasted. Also, circumstances often dictate that write caching be turned off. Lastly, IDE's lack of overlapped operations and queuing and such makes it a lot harder to take advantage of raw bit rate.

    In short, there really are many situations - particularly for home users - where the extra speed really isn't useful and serves only as a marketing gimmick.



    [ Parent ]
    Serial ATA is here (2.00 / 3) (#27)
    by truth versus death on Mon Aug 12, 2002 at 05:10:22 PM EST

    And it is called FireWire!

    "any erection implies consent"-fae
    [ Trim your Bush ]
    no. (none / 0) (#97)
    by ph0rk on Wed Aug 14, 2002 at 04:09:22 PM EST

    as firewire inside isn't a reality (yet).
    [ f o r k . s c h i z o i d . c o m ]
    [ Parent ]
    Apple tax (none / 0) (#110)
    by Otterley on Thu Aug 29, 2002 at 03:37:58 AM EST

    Yeah, except few motherboard manufacturers want to raise their prices by $10-$25 through paying the Apple tax.

    [ Parent ]
    Um (none / 0) (#112)
    by truth versus death on Sun Sep 29, 2002 at 11:44:42 PM EST

    What the heck are you talking about?

    The proposed $1 fee never happened. It is 25 cents per a device no matter how many ports.



    "any erection implies consent"-fae
    [ Trim your Bush ]
    [ Parent ]
    ATA and IDE (5.00 / 7) (#30)
    by ghjm on Mon Aug 12, 2002 at 05:30:08 PM EST

    Several editorial comments have asked the author to define the acronyms "ATA" and "IDE" - here's my attempt.

    "ATA" = AT Attached
    "IDE" = Integrated Drive Electronics

    These terms are meaninful only from a historical perspective.

    The AT was the first IBM personal computer to include a 286 processor, and the second to include a hard drive as standard equipment (after the XT). At the time, hard drives used the ST-506 interface. The drive was a simple physical device, with control and data lines run out to the controller through ribbon cables. The control lines generally did things like: "If you apply +5V to this pin for 10ms then the head will step forward by one track." The controller card, which was inside the PC, actually took care of all the logic for moving the stepper motors, identifying sectors as they spun under the heads, and so forth. This eventually became a serious limitation in drive design. You couldn't just invent a new type of head actuator and ship it to the market, because it wouldn't work with existing controller cards.

    The first IDE drives took the functions of the controller card and put them onto the circuit board of the drive itself. Hence, IDE ("integrated drive electronics"). The standard bus architecture at the time was what we now call ISA ("Industry Standard Architecture"), which was then called the AT bus. In place of the old ST-506 interface, IDE drives presented a simplified AT bus connector through their ribbon cable. This made IDE controller cards very simple: they connected some of the pins from the AT bus out through the ribbon connector, with just a little bit of on-board logic for access control and address decoding. Thus, the name ATA ("AT Attached") because, unlike ST-506 drives, IDE drives were directly attached to the AT bus.

    Now where did I put my teeth...

    -Graham

    Advanced Technology Attachment (none / 0) (#38)
    by x3nophil3 on Mon Aug 12, 2002 at 07:02:57 PM EST

    Defining one third of an acronym is not quite good enough.

    [ Parent ]
    No, that's just wrong. (none / 0) (#78)
    by ghjm on Tue Aug 13, 2002 at 03:13:35 PM EST

    Nobody who was around at the time would ever decode the acronym that way. It just doesn't make any sense. And it's "attached," not "attachment."

    Also, if I recall correctly, I went to some trouble to define the terms "AT" and "AT-bus" earlier in the text.

    -Graham

    [ Parent ]

    Some of this is actually coming back to me. . . (none / 0) (#44)
    by IHCOYC on Mon Aug 12, 2002 at 08:50:12 PM EST

    . . . though not having paid attention to this sort of thing for some time, it's still a haze.

    One thing I distinctly remember was an oddity of certain ribbon cables that went into very old machines. I've still got some of them in a parts drawer somewhere. What made them odd was the fact that in the cable, somewhere, some portion of the wires were separated out, given a half twist, and then went on to the other end backwards.

    What were they smoking when they thought up this design?

    Heus, nunc, mihi cantate hanc æruginem.
    [ Parent ]

    ah yes (none / 0) (#46)
    by Work on Mon Aug 12, 2002 at 09:49:32 PM EST

    i remember those. I vaguely seem to remember them being attached to certain floppy drives?

    [ Parent ]
    those are floppy cables. (5.00 / 1) (#54)
    by cryosis on Tue Aug 13, 2002 at 01:06:56 AM EST

    I do believe that they have 34 wires and 2-5 connectors. You'd have your motherboard connector then probably a 3.5" B drive connector, a 5.25" B drive connector, a 5.25" A drive connector and a 3.5 A drive connector. That's how the one laying on my desk looks, anyway. I've seen them with out the 5.25" connectors, as well. Those 5.25" connectors are the most annoying things I've ever had in my computer. They're always in the way. The twist in the cable was so that if you had a machine that had two floppies, you wouldn't have to configure anything on them, just plug them in and the one connected to the A connector would be A: and the drive on the B connector would be B:. Simply a time-saver.

    Everything you want to know about floppy's

    [ Parent ]
    The twist was to show which connector ... (none / 0) (#71)
    by Rizzen on Tue Aug 13, 2002 at 12:17:30 PM EST

    ... was used to connect to the A: drive and which for the B: drive.  Why they ever decided to lock in the drives to the cables to specific letters, I'll never understand.
    The years of peak mental activity are undoubtedly those between the ages of 4 and 18. At age four, we know all the questions; at eighteen, all the answers.
    [ Parent ]
    MSFT? (4.00 / 1) (#84)
    by sgp on Tue Aug 13, 2002 at 07:42:03 PM EST

    I was always under the impression that it was DOS which specified this - without DOS, we'd be calling them fd0 and fd1, or somesuch.

    I'm not too sure of the history, though - it was DOS2 which supported hard disks (C:).

    For fortunecity, http://members.fortunecity.com/pcmuseum/dos.htm seems a pretty detailed history.

    In looking around, http://www.computerhope.com/dissues.htm is a pretty short list of known bugs for an OS which is old enough to drink.

    There are 10 types of people in the world:
    Those who understand binary, and those who don't.

    [ Parent ]

    Look further, young Jedi... (5.00 / 1) (#106)
    by ghjm on Mon Aug 19, 2002 at 01:27:09 PM EST

    DOS copied all this from CP/M. Here's my best recollection of what a CP/M command line session might look like:

    A> SAVE 1 X.COM
    A> DIR *.*
    X        COM
    A> PIP B:X.COM=A:X.COM
    A> B:
    B> DIR *.*
    X        COM
    B> ERA *.*
    B> DIR *.*
    B> A:
    A> DIR *.*
    X        COM
    A>

    Note that this session shows several characteristics normally associated with DOS:

    - Drive letters (A>, B>)
    - Eight-dot-three filenames
    - *.* as a global wildcard
    - COM as a file extension for executable images

    DOS 1.0 was written as a workalike for CP/M that ran on the 8086 processor. The one nice thing about DOS was that it had a "copy source destination" command rather than CP/M's weird "PIP destination=source" command. Other than that, DOS 1.0 had very few new ideas - it was all copied from CP/M.

    DOS 2.0 added support for subdirectories and hard disks, but this was not the first time they appeared in microcomputers - the UCSD p-System, for example, had those features (though not *nested* subdirectories) running on IBM PC hardware, well before DOS 2.0 came out.

    -Graham

    [ Parent ]

    cheap cable manufacturers (none / 0) (#94)
    by automaton on Wed Aug 14, 2002 at 01:51:55 PM EST

    Some manufacturers (still today, I think) are so cheap and not provide any indicators such as the red line or that notch (whatever) thing on the cable. Twisting a cable and looking for the error somewhere more 'high-level'. The memories...

    [ Parent ]
    ATA is not IDE (none / 0) (#77)
    by ggeens on Tue Aug 13, 2002 at 03:09:36 PM EST

    ATA is an extension to the IDE standard. It added support for "cable select" - a way for a device to automatically decide whether it's master or slave. I don't know if there are other differences.

    I suppose most motherboards support this, even if the connector is labeled as IDE.

    L'enfer, c'est les huîtres.


    [ Parent ]
    Please, for God's sake let's not get into it (5.00 / 2) (#82)
    by ghjm on Tue Aug 13, 2002 at 03:56:50 PM EST

    But for the record, the terms IDE and ATA are in fact completely interchangeable. Both terms refer to the standard now formally known as ANSI X3.221-1994. This standard specifies pin 28 as cable select, but it is optional to implement. No IDE or ATA standard has ever specified pin 28 as "reserved" or "unused" - it has always been cable select; but some manufacturers chose not to implement it. Whether a given drive of this era supported cable select had nothing to do with whether they called themselves IDE or ATA.

    The first revision to ATA/IDE was known as ATA-2 or EIDE, and its major new feature was DMA transfers (specifically, it added PIO modes 3 and 4, and DMA modes 1 and 2). It may have made implementing cable select mandatory, or perhaps the manufacturers just didn't want to let each other win a checkmark on the bullet list. For whatever reason, after ATA-2, all the drives I've ever seen have supported cable select.

    -Graham

    [ Parent ]

    Serial ATA is not here (yet) (4.50 / 2) (#33)
    by steveftoth on Mon Aug 12, 2002 at 06:24:34 PM EST

    This article at Tom's HW is evidence that it's just another slow news day here on the internet.  Goto Pricewatch.com or any other web site that sells Hard Drives and controller cards.  Guess what, you won't find one that sells Serial ATA cards and drives.  That's cause they are not on the market yet.  Now I know that he's not claming that you can buy any on the market, but the point is that Serial ATA is not here in the stores today, and won't be for some time yet.  Someday, they will use it to interconnect all your drives to your computer.  

    Personally I don't see why Serial ATA should take off, todays hard drives are very good, the cableing is not that bad and the speed is good for the price.  The only reason that I see to upgrade to Serial is to reduce the cableing mess that exists inside all computers today.  Though if we were to build a computer with all Firewire drives it would work today.  However all firewire drives are IDE drives wrapped in a Firewire interface ;)

    If you read the article... (none / 0) (#35)
    by Work on Mon Aug 12, 2002 at 06:31:01 PM EST

    you'd see that it mentions that the first drives and controllers will hit the market beginning in fall.

    The cabling reason is precisely why it will take off. Much easier to manufacture machines with that kind instead of ribbon cables.

    [ Parent ]

    Cabling not that bad? (4.00 / 1) (#57)
    by ajduk on Tue Aug 13, 2002 at 04:59:24 AM EST

    Sorry - I put PCs together for friends and family, so I get to mess around inside PCs on a regular basis.  Trust me, routing IDE cables really is one of the most fiddly tasks, and then making sure they don't hit one of the many fans you get in the average PC (you have to keep clear of the graphics chip fan, the Northbridge fan, and the CPU fan...

    And then the connector pulls out of the motherboard slightly, so you have to reach through the nest of cables to get at it..

    So this is a good, practical innovation, from my point of view.

    [ Parent ]

    Not really (none / 0) (#70)
    by rasilon on Tue Aug 13, 2002 at 12:08:50 PM EST

    I used to build custom PCs professionally,and I really didn't find it a problem. A reel of 40-pin IDC cable, a bag of connectors and a crimp tool are really quite cheap and making the cables yourself allows you to make the cables perfect -- no rats nest, no taught cables that pull out when you brush them, no fouling on fans... Add to that some pritt-pads, cable ties and and tie-plates and you have cable runs that look professional, are easy to work with and actually help airflow.

    [ Parent ]
    Smartass.... (none / 0) (#87)
    by sgp on Tue Aug 13, 2002 at 08:10:55 PM EST

    Cables I make have about a 35% chance of working... :(

    There are 10 types of people in the world:
    Those who understand binary, and those who don't.

    [ Parent ]

    It just takes practice (none / 0) (#90)
    by rasilon on Wed Aug 14, 2002 at 05:16:10 AM EST

    I used to make the most abysmal cables when I was younger. I was lucky, I grew up in a family where electronic construction was normal so I got plenty of practice when it didn't matter. Not that you need one for IDE cables, but I got my first soldering iron when I was 8, and learned why you don't solder in shorts when I was 9.

    [ Parent ]
    Of course... (none / 0) (#89)
    by ajduk on Wed Aug 14, 2002 at 04:09:01 AM EST

    a) You have more kit than me and (althougj it pains me greatly to say it) you're probably better at putting PCs together then me.

    b) Serial ATA means that you don't have to do all this - you'd just have to use a few cable ties.

    [ Parent ]

    It is honestly quite cheap (none / 0) (#91)
    by rasilon on Wed Aug 14, 2002 at 05:38:35 AM EST

    a. Crimp tools are very useful, even if you only use it a few times, IMHO the tenner it costs is worth the reduced aggravation. The cables themselves cost pennies. Money is a good motivator. b. I'm not arguing against SATA, I'm a fan of Fibre Channel for exactly that reason. I just wish it was vaguely affordable in which case I expect ATA would vanish. As a useful standard for home PCs, SATA may well be a winner.

    [ Parent ]
    As with others, I don't see the need (none / 0) (#107)
    by Quila on Tue Aug 20, 2002 at 11:06:31 AM EST

    Connect the logical dots:
    1. A Firewire drive already has 400Mbps, a USB 2 has 480Mbps.
    2. There's already a large market for portable Firewire and USB hard drives. They're making all of those drives as IDE drives with a bridge
    3. No drive standard will come out unless there's a large demand for production
    4. In #2 I showed this demand
    5. Plus by having these native, you remove the cost of the bridge
    Just forget serial ATA and go Firewire or USB 2 for internal and external.

    [ Parent ]
    HighPoint Serial ATA Controller... (4.50 / 2) (#36)
    by questionlp on Mon Aug 12, 2002 at 06:40:37 PM EST

    The HighPoint two channel Serial ATA controller that was featured in the article is really an ATA/133 controller that has two chips that perform the conversion from ATA/133 to Serial ATA (more info can be found here.)

    It would be really nice to see Serial ATA controllers to be integrated into the southbridges, thus hopefully removing the 132MByte/sec barrier of a 32-bit/33Mhz PCI bus. Unfortunately, outside of workstation and server class motherboards, 64-bit PCI slots are not very common (even the Tyan Tiger MPX motherboard isn't really aimed at the home user).

    I also hope that the controller, cable and drive manufacturers do resolve the problem with the cable connectors being a bit loose.

    -- http://closedsrc.org

    SCSI (3.33 / 3) (#41)
    by PhadeRunner on Mon Aug 12, 2002 at 07:20:20 PM EST

    IDE/ATA in any flavour is never going to be as good as SCSI.  Its another case of the "innovators" flogging the dead horse of a historically broken architecture simply to get more speed to the unwashed masses.

    Funnily enough, sounds like the x86/PC architecture there but thats a WHOLE other story.  

    BTW, IDE cables were flat for a reason.  Round cables are NOT a "performance" upgrade.  

    Price vs. Performance (3.00 / 3) (#50)
    by Torgos Pizza on Mon Aug 12, 2002 at 11:57:30 PM EST

    Yet, the price of SCSI prohibits SCSI drives from being marketed to the consumer. The price/performance ratio accompanied with the larger drives that will be available make Serial ATA a winner.

    It's the old BETA vs. VHS argument. SCSI may be a superior technology, but Serial ATA will come standard on all new motherboards come the fall and the newest drives will have it. SCSI doesn't stand a chance against it and Firewire.

    I intend to live forever, or die trying.
    [ Parent ]

    some other things (2.50 / 2) (#51)
    by Work on Tue Aug 13, 2002 at 12:19:07 AM EST

    scsi is a bitch to configure. is scsi even hot-swappable without special hardware?

    [ Parent ]
    SCSI is a protocol (5.00 / 2) (#64)
    by psyconaut on Tue Aug 13, 2002 at 07:52:34 AM EST

    SCSI's a protocol predominantly....hence the previous comment about Firewire being SCSI.

    Hardware standards such as SCA/SCA2 implement physical layer connection for hot-swapping. Hot swapping and SCSI is basically just an issue of safely physically disconnecting the drive rather than special protocol support.

    Many people comment that you can get ATA drives with similar performance to SCSI these days, so why bother? Quite frankly, ATA is a "dumb" protocol....SCSI is a more elegant and better engineered solution (things like the elevator algorithm for laying data onto the drive, drives with intelligent controllers, etc).

    -psyco

    [ Parent ]

    Standards (4.40 / 5) (#55)
    by truth versus death on Tue Aug 13, 2002 at 02:51:16 AM EST

    SCSI doesn't stand a chance against it and Firewire.

    Except that FireWire is SCSI. =)

    "any erection implies consent"-fae
    [ Trim your Bush ]
    [ Parent ]
    "the fall line" (5.00 / 6) (#63)
    by tps12 on Tue Aug 13, 2002 at 07:49:00 AM EST

    And it looks like "breezy" is the word when it comes to the new fall line soon to be introduced by component makers in Paris and New York later this month. Expect lighter, more playful cables anchoring down denser drive platters. Keeping cool is always a concern, so fans will be everywhere. Unlike in the past, though, don't expect designs to flaunt them. This season, it's all about stealth.

    [ Parent ]
    who claimed they were a performance upgrade? (4.00 / 1) (#52)
    by Work on Tue Aug 13, 2002 at 12:26:08 AM EST

    ive never seen that claim anywhere about rounded IDE cables. Personally I think they're gimmicky novelties simply because they cost a great deal more and for what? Slightly better airflow around them?

    The solution to airflow in a case is to remove most all obstructions. Most cases still have these big thick round IDE cables hanging down in the way anyway because they're too short and stiff to move elsewhere.

    The reason ribbon cables are flat, i believe, is because of the high frequencies used allows it to minimize interference. The rounded cables use lots of shielding and insulation between the wires to achieve the same line noise quality, but this in turn costs quite a bit extra.

    [ Parent ]

    What do rounded cables have to do with anything? (3.00 / 2) (#61)
    by crankie on Tue Aug 13, 2002 at 06:27:46 AM EST

    Serial ATA, a seven-wire high-speed-serial interface is being discussed, not different cabling on the existing parallel technology.

    ~~~
    "The great thing about hardcore socialists is the silence they emit once they start earning a decent wage." - tombuck
    [ Parent ]
    Oops. (none / 0) (#62)
    by crankie on Tue Aug 13, 2002 at 06:29:24 AM EST

    Being the author, you probably knew that.

    ~~~
    "The great thing about hardcore socialists is the silence they emit once they start earning a decent wage." - tombuck
    [ Parent ]
    rackmountables (none / 0) (#100)
    by thrig on Wed Aug 14, 2002 at 09:29:29 PM EST

    At work, we've been looking at a cluster for folks to run various computational biology software on.  We're looking at dual P4 or althons in 1U rackmounts, and the hardware vendors indicate that they are having all sorts of trouble with airflow on the beasts.  Including turning the RAM sideways, airflow routing tunnels to the CPU, etc.

    Most of the systems have the SCSI/IDE cables running flat across the top or the bottom of the case, though rounded cables may help in some circumstances.  Will have to ask the vendors again once the rounded cables come out how things work out.

    [ Parent ]

    2U half-depth (none / 0) (#109)
    by Otterley on Thu Aug 29, 2002 at 03:35:56 AM EST

    Try 2U half-depth rackmount systems -- they're half the depth of a normal system but you get the same density as using full-depth 1U systems. Plus you get to take advantage of P4 or Athlon solutions. I recommend the folks at Rackable Systems; they supply companies like Yahoo! and Hotmail.

    [ Parent ]
    I had to use rounded cables. (none / 0) (#103)
    by Ghost Shrew on Thu Aug 15, 2002 at 03:57:49 PM EST

    I couldn't move anything around, as I had to put the raid controller in one of the top two slots. Unfortunately, the mass of cables blocked air flow from the front of the case, no matter what I did. Rounded cables solved the problem.

    Free tabletop RPG!! Grey Lotus
    [ Parent ]

    fool (2.50 / 2) (#79)
    by trhurler on Tue Aug 13, 2002 at 03:15:02 PM EST

    Serial ATA is considerably faster than any existing SCSI that isn't a one-off hack(ie, than any standardized SCSI,) and with one device per channel, it doesn't suffer any of the ATA or SCSI contention problems. Granted, the SCSI protocol would have been a nice thing to add to the ATA spec(it is presently a strict subset of SCSI, which you might know if you weren't just repeating what you read somewhere.)What's going to make SCSI faster again? Serial SCSI, humorously enough.

    Incidentally, you shouldn't be so disparaging of economic issues. If it weren't for commodity hardware, the one thing that is almost certain is that neither you nor I would have access to any computers at all for any purpose other than possible commercial activity, and odds are the market for such skill would be so small and the opportunity to learn it so rare that neither of us would be in this industry.

    Finally, properly twisted and constructed, there's nothing wrong with round IDE cables; they can easily exceed the transfer rate of existing IDE devices without problems. The gain in airflow can be substantial, and when you've got a hot video card, a hot processor(or two,) and so on in your machine, and everything in there that is expensive is also heat sensitive, you might just prefer to shell out for expensive round cables properly made rather than see your hundreds or thousands of dollars investment in hardware burn itself to a crisp.

    --
    'God dammit, your posts make me hard.' --LilDebbie

    [ Parent ]
    You'll never see the speed (none / 0) (#92)
    by ScottW on Wed Aug 14, 2002 at 06:48:54 AM EST

    You're forgeting that the fastest hard drives currently top out at 33MB/s. Even though serial ATA promises speeds at up to 150 MB/s (600 MB/s in the future), you're not going to see that kind of speed since you can't hook up 2 or more drives to the same channel and let them share the avabile bandwith.

    And while you're arguing economic considerations, don't forget that a new piece of hardware allways comands a high price when it comes out. An example of this: my ATi AIW Radeon costed around $400 when it came out, most places were selling it for around $500. I piad a little under $150 when I got mine several months latter. Serial ATA is no exception even if they sell it at a loss. Meanwhile, SCSI has been around for a long time, and prices for newer SCSI devices have been dropping.

    Arguing "Serial ATA is cheaper" and "serial ATA is faster" is stupid and pointless; serial ATA's price and speed doesn't mean jack shit when you can't buy it yet! Last I checked (5 mins. ago), no one is selling serial ATA devices yet, at least no one who sells to the mainstream. SCSI devices have been avabile for a long time, and can allready offer speeds up to 160 MB/s, with 320MB/s around the corner and fast speeds in the future.

    Finally, there will always be applications where ide/serial ATA will not work, or simply will not yeald acceptable results. There's a reason why the majority of servers, RAID arrays, clusters, and serious workstations use SCSI, IDE/serial ATA's limitations make it unuseable for the type of applications those things are used for, and the type of configurations commonly used.

    [ Parent ]

    this keeps coming up... (none / 0) (#93)
    by Work on Wed Aug 14, 2002 at 12:47:17 PM EST

    all about how current harddrives can't get up to speed..who says what you have to hook up to serial ata has to be a mechanical harddrive?

    [ Parent ]
    Heh (5.00 / 3) (#95)
    by trhurler on Wed Aug 14, 2002 at 02:01:30 PM EST

    First of all, we're talking about the near term future and onward, not the present. All arguments consisting of "you can't buy it yet" are completely off topic.

    Second, we're talking about a standard that is supposed to last a couple of decades. Of course its high end performance isn't yet attainable; if it was, the standard would be out of date in a matter of months. Nevertheless, putting a single device on each channel has large performance wins. You eliminate a source of contention and the need for contention management overhead. Two drives of equal capability, one on a serial ATA interface and one on a SCSI controller with five other devices, will not perform equally. The SCSI device will lose. (Which leads into my last two points, but I'll get there...)

    Third, we're talking about commodity hardware, not high end flash in the pan soak the customer while you can stuff like video cards. Price structures will be different than you're envisioning. Sure, it may take a few weeks or months for prices to settle out, but so what? At their peak, they'll be cheaper than SCSI.

    Fourth, SCSI advertised speeds are complete hogwash. You will never, ever see those transfer rates in any real use - nor will you see ATA advertised rates. You might think that since SCSI offers such high peak transfer rates, its average rates would be higher too. This is because you do not actually know what you're talking about. A SCSI chain with more than one device on it won't even peak as advertised, much less sustain anywhere near what it "ought to." The latest greatest devices on the latest greatest controller will be lucky to pull 4MB/s if they're all on the same chain. Serial ATA is easily capable of that.

    Fifth, and most contentious I'm sure, I address your final paragraph. The entire thing is a bald assertion. You heard it elsewhere, and you are repeating it. In your head, you're comparing integrated IDE chipsets on motherboards to $200 SCSI card addons. That's not reality. Reality is, you compare your card to a $200 ATA controller, and the SCSI card comes up short. The SCSI card will have two or three interfaces, but will only be able to use at most two of them at a time. The ATA card will have six or eight interfaces, so you can put that many devices on it without any channel contention whatsoever. The ATA devices will cost half as much or less. The ATA controller will have support for as much as a gig of SDRAM cache; the SCSI card will have no cache beyond its read/write buffers. The ATA card will have a system-side design like a SCSI card, so it won't burden your CPU any more than a good SCSI setup would. In short, the SCSI card, with eight devices, will get its ass kicked in by the ATA card with eight similar devices. In addition, the ATA card probably has support for RAID in case you want it.

    The reason SCSI sells in high end gear is that for businesses, price isn't as big an issue, and SCSI has a performance image carefully sculpted by marketing departments across the industry, while ATA's image is carefully maintained by the same people as "consumer grade." Technical capability has little or nothing to do with this.

    You sound to me like the people who go around lamenting the replacement of Sun's SBUS with PCI. "Oh, but the SBUS has better performance for blah blah blah." Ends up that it doesn't, but they were told that so many times that they still believe it. In point of fact, modern PCI buses(the only kind Sun ever used,) are so ridiculously fast compared to the best SBUS arrangement that it isn't even funny, and for their really good stuff, they use one bus per device(like Serial ATA, amusingly enough,) getting even BETTER performance.

    PCI and ATA have led to a world where $1000 Sun machines of today outperform machines Sun sold for $100,000 or more only a few years ago. PCs of today are faster than the biggest fastest machines from a time I can easily remember. Your view of reality is shaped by marketing.

    Incidentally, tech review websites and magazines and so on are no source of accurate information. Why? Because they depend on freebies and handouts from the companies whose products they test. They're just another marketing expense, as far as the companies are concerned.

    Don't buy the hype. SCSI is what you use in the commercial world because it is politically acceptable - not because it is actually better suited to the task.

    --
    'God dammit, your posts make me hard.' --LilDebbie

    [ Parent ]
    Bait (none / 0) (#101)
    by Rhodes on Wed Aug 14, 2002 at 09:52:41 PM EST

    Do you have any numbers to back up your claims?

    [ Parent ]
    Tell you what (5.00 / 1) (#102)
    by trhurler on Thu Aug 15, 2002 at 11:36:24 AM EST

    You pay me to be the consumer reports of computing, and I'll spend the thousands of dollars I'd need to do reliable, meaningful tests that would provide real numbers on current hardware. Otherwise, just go back and keep feeding at the marketing trough if you don't want to look into it yourself; I lose nothing by this. I've given you decent reason to be suspicious and to look into it if you care; if you don't, that's your business.

    --
    'God dammit, your posts make me hard.' --LilDebbie

    [ Parent ]
    well (none / 0) (#105)
    by EriKZ on Sun Aug 18, 2002 at 06:17:18 PM EST

    He was probably hoping that you'd have some links to share.

    You seem to be very well informed, so assuming this wouldn't be out-of-hand.

    [ Parent ]

    Digital Rights Management? (4.50 / 2) (#73)
    by guido on Tue Aug 13, 2002 at 12:27:04 PM EST

    If I recall correctly, Serial ATA was supposed to introduce some sort of Digital Rights Management.
    Is that true or have I mixed something up?

    Guido

    Backwards compatible. (4.00 / 2) (#74)
    by crankie on Tue Aug 13, 2002 at 12:44:47 PM EST

    Serial ATA is backwards compatible with parallel ATA. The controller presents an identical interface to the host system. With the slight exception that instead of seeing a primary and a secondary on a single channel, you see two primaries on two separate channels. No additional drivers are necessary. Existing OS's should be able to just up and use it.

    Based on this, I doubt that there is any sort of DRM scheme in place. If there were, it would have to be some sort of "value-added extra" which was not backwards compatible, and so it would have to be possible to disable it.

    ~~~
    "The great thing about hardcore socialists is the silence they emit once they start earning a decent wage." - tombuck
    [ Parent ]
    Been following Serial ATA awhile now (4.00 / 1) (#80)
    by Fen on Tue Aug 13, 2002 at 03:44:13 PM EST

    Will not get another PC without it. This may have been an offshoot of USB, firewire, and infiniband (server tech I worked on). We see lots of things going serial now, and it is making things simpler and more elegant. I'm not sure why they couldn't just make a single standard (USB2 or firewire) for in the case and out of it, like SCSI, but it really came down to cost of mass-producing. By the way, that's MB (off by factor of eight).
    --Self.
    3GIO (5.00 / 1) (#96)
    by nitzmahone on Wed Aug 14, 2002 at 02:05:12 PM EST

    Funny you should mention a single interconnect for everything, in-case or out- that's exactly what 3GIO is all about...

    3GIO architecture page

    [ Parent ]

    This might link in well.. (none / 0) (#108)
    by tonyenkiducx on Wed Aug 21, 2002 at 06:11:03 AM EST

    Allthough its still a couple of years off, the companies involved are confident that it will eventually replace hard drives.. There was a much better article on BBC online, which also described the maximum estimated capacities(In the 100's of GB's), and the lower power usage, but I cant find that one.

    I post this because a lot of people are saying its pointless because HDD's only do 50MB/s, but its opinions like that, that are the reason were stuck with the current IDE problem that we have.

    None-Volatile memory

    Tony.
    I see a planet where love is foremost, where war is none existant. A planet of peace, and a planet of understanding. I see a planet called utopia. And I see us invading that planet, because they'd never expect it
    Serial ATA is here | 112 comments (82 topical, 30 editorial, 0 hidden)
    Display: Sort:

    kuro5hin.org

    [XML]
    All trademarks and copyrights on this page are owned by their respective companies. The Rest © 2000 - Present Kuro5hin.org Inc.
    See our legalese page for copyright policies. Please also read our Privacy Policy.
    Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
    Need some help? Email help@kuro5hin.org.
    My heart's the long stairs.

    Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!