Slashdot Mirror
Seagate Barracuda V Serial ATA Drive Reviewed
Mike Parsons writes "Andrew and Adam over at Explosive Labs have a nice review up on the Seagate Barracuda V, one of the first production Serial ATA drives. Keep in mind, Generation 1 of Serial ATA was not meant to be a 'incredible performance jump.' Rather, its intended purpose was to make the industry transition seamless to allow time to mature the future generations of SATA. Generation 2 and 3 of SATA show more promise for those interested in performance, as white papers behind them gives you the nice fuzzy feeling for speed!"
The problem with parallel data transfer is that, if you have lots of data channels running at the same time, you have to synchronise the arrival of all this data. It's very difficult to do this at high speeds (if you look at your motherboard you'll probably find a few places where the track wiggles back and forth to synchronise the arrival of data).
In recent years, it has become possible to run data connections at very high speeds - but only when you have only one data line. A USB 1.0 connection is comparable in speed to an ECP parallel connection, and there are far faster serial technologies nowadays.
Like car accidents, most hardware problems are due to driver error.
This is one reason why SCSI is so much faster/more expensive; all scsi controllers have this functionality so throughput is maintained even when the parallel data is clashing.
Whilst serial is theoretically slower than parallel, by removing these synching issues you can guarantee better performance in consumer-priced hardware. At the server end, SCSI will remain as price is less important than performance, and as I said, parallel is still more efficient if it has a decent sych algorithm *on board*.
theres another review at storage review
Gyrate Dot Org - "Where high-tech meets low-life"
Why is this serial stuff so much quicker than the older parallel connections?
Parallel bus interfaces are much harder to get to work properly, because usually there are very small differences in the lengths of the individual wires in the ribbon cable and so the signal delay varies from wire to wire; so you have to design your controllers to account for these delays (and that's why there was so much voodoo magic involved in configuring early SCSI equipment). The practical limit of synchronous transmission rates is much lower for parallel than it is for serial. That's why Ethernet and FireWire are serial interfaces.
Bush Lies Watch
I always thought parallel was faster
:-)
:-)
Yes, provided they both run on the same clock speed. In this particular case they're not.
When you ramp up the clock speed of the parallel bus you get all sorts of problems (synchronisation issues, multiple wires affecting each other's capacitance, inductivity and such). One way of avoiding those problems was UltraATA's 80 wire IDE cable. And that came with increased price tag, and didn't ultimately solve all the problems, it just postponed them for a generation or two.
The other way was to abandon parallel all the way and go serial. Since with serial (one pair of wires) you don't get any above mentioned problems you can ramp up the clock much higher, and thus get better thrhoughput, although you're transfering just one bit at a time.
At first it sounds counter-intuitive, but it just goes to how much intuition is worth.
Oh wait, too early in the morning. Was the USB comment a joke?
If you want a fast parallel protocol, think about trunking multiple gigabit ethernets. Instead of running bits in parallel, you run packets in parallel. You get more bandwidth, without having the timing issues of a bit level parallel cable.
Running multiple serial links in parallel is also a win for fault tolerance. If one cable is sliced, the connection is still up, just slower.
I don't expect to see multiple SATA cables to a single drive, but I wouldn't be surprised by multiple SATA cables to a RAID array.
Trust in google.
: www.explosivelabs.com/reviews/barracudav_sata/+%22 %2Bwww.explosivelabs.%2Bcom/reviews/barracudav_sat a/%22&hl=en&ie=UTF-8
http://216.239.39.100/search?q=cache:AXz0ph7JjFsJ
Daniel
Carpe Diem
All compliant adaptors and disks support hot swap
linux will support hot swap as any other removable drive
the only current system without full compliance (I don't know why they do this consistently) are Apple Powermacs. They do not have hotswappable SATA.
AFAIK, there's always one cable per SATA drive, all cables going back to the controller.
What's this Submit thingy do?
Why get rid of the old MOLEX? Since an adapter is included with the drive it doesn't seem that there are any new voltages required.
The new connectors provide 3.3V as well. The first generation of drives will not make use of it.
The Seagate drives may be a tad slower, but if you want to build a home multimedia server that will sit in your living room, they have the benefit of being unbelievably quiet.
At 24dB, you have to put your ear to a Barracuda drive to hear it, whereas the Western Digital drives put out a whopping 39dB!
If I remember things correctly this is to make it possible to run many different voltages (something like 3) to the drives, suitable for different sized drives.
The spec can be downloaded here (about 1 meg), if someone cares to verify my claims. It's all there.
.: Max Romantschuk
You can get multiple voltages from any difference in potential. (Voltage is just a term describing the difference between the charge density, or 'pressure' between two points)
For example, if I placed two 1 KOhm resistors in series between "GND" and "-12V", at the contacts between the two resistors, the voltage compared to GND is -6V, and the voltage compared to "-12V" is actually +6V.
However, due to resistor tolerances and Thevenin resistance, it's much more preferrable to have the power supply give a steady, regulated supply of -6V and +6V, if you need them.
What's this Submit thingy do?
So that you can do hot plugging. The current MOLEX connector cannot be used for this - first, it requires far too much force to connect or disconnect. Second, there is no guarantee of ground before any other pins connect. Third, there is no standard on where the power connector will be located in the drive bay or with respect to the data connector.
SATA fixes all of this.
Is this just another one of those PITA upgrades?
Frankly, I can't see how anyone would consider anything about SATA a PITA. Smaller, more flexible cable, no jumpers, no master/slave crap, and a standardized power connector. Where's the pain? (Ok, you'll pay maybe $20 more for the drive at first, but that pain will disappear shortly)
SSA had this years ago (up to 160 MBytes/s), as well as Fibre Channel. The only reason for not going serial was a large installed IDE base to be accomodated.
Alright so far most of the posts are misguided, so I'll answer a few questions. But first, this was mentioned in the slashback that is still on the front page. Please post any corrections.
1a. Yeah!, faster drives.
No. Find me a drive that can use PATA-100 to the max let alone PATA-133 and I'll be a very happy customer. Current drives do not use current capacity, the only time the bus becomes an issue is where you are bursting from the drive cache to the controller, which is not enough to really worry about except in certain situations (The same data is read continuously).
b. Yeah!, Faster drives.
No. Why a second point? The first point dealt with bandwidth. This one is for latency. Please remember that most SATA controllers on motherboards, etc (atm at least) are actually a bridging chip to a PATA controller. This incurs a slight latency delay. If you do a lot of small file accesses you will be effected.
2. Whats the point we already have enough speed?(ie I already know 1.)
a.The point is smaller cabling, making cases less cluttered, meaning better cooling, and easier to keep wiring neat and out of the way. Why no use rounded cables? You didn't think the cables where a ribbon shape for looks did you? The cables are meant to be ribbons to reduce the interference between each pair (limits it to the pair on each side). Rounding the cables causes all pairs to interfere with each other resulting in a much shorter maximum cable length before there are too many interference errors on the bus.
b. Point to point cabling, knock a cable loose, or have a misbehaving drive and you loose one drive. With PATA you can loose 2, or with SCSI you can loose up to 14 (Wide, not typically a problem on modern auto-terminating devices)
c. You can disconnect a drive from a powered controller without risking blowing the controller chip (Possible with PATA). Making removable hard drive cradles finally usefull on ATA systems.
d. Longer in-spec cable lengths. PATA cables (Sorry I forget the length off hand) can't reach the top 5 1/4" drive bay in a full tower case. SATA cables can. Why not use longer PATA cables? Cables longer than PATA spec tend to suffer badly from interference based errors, resulting in a lot of resends on the bus, sometimes causing bad data on drives.
3. The performance isn't what I hoped (or a WD JB is faster)
This drive isn't intended to be the fastest on the block, it is meant to be quiet. Seagate drives have the new fluid bearings, they haven't been the fastest on the block for a while now, what makes you think this one would be different?
I personally think this is a good drive to be first to SATA, as the people likely to appreciate the quiet drive would also desire the better air flow offered by smaller cables, meaning slower case fans, and a quiter PC.
4. Why don't they compare a PATA Barracuda V vs a SATA Barracuda V.
The PATA has a 2mb cache the SATA has an 8mb cache (and a slightly faster access time, by 0.6ms). They aren't directly comparable, the SATA version is obviously aimed to be the top of the line model.
5. The power connecters. The Barracuda V requires the same power voltages that current PATA drives do, so an adapter works fine. However it was intended to supply drives with multiple voltages (such as 3.3v, 5v, etc) so that the electronics can use a different voltage than the drive motors, reducing the power consumption of the electronics, and therefore the heat output. Some drives get very hot, and every little bit helps.
I think thats all.
High speed USB is 480 megaBITS/second
which translates to 60megaBYTES/second
Serial ATA is faster.
--
Faster transfer speeds (Firewire2 will leave SATA in the dust).
Firewire 2 = 800 Mbps = 100MBps
SATA = 150MBps
Firewire 2 faster? Don't think so. Sure, Firewire 2 will ramp up to twice that speed eventually, but so will SATA...
SATA is also a lot simpler to implement: chipset manufacturers can reuse most of their old, highly-optimized Parallel ATA controller core. Similarly, OS writers can reuse most of their old ATA drivers. SATA has less overhead than Firewire, it's designed for data storage and data storage alone, and it doesn't do daisy chaining.
Firewire's a nice technology, and it would work for hooking hard drives up internally, but it doesn't do the job as well as SATA does, it's over-complicated (and thus expensive), doesn't have the track record, and probably most importantly, has some serious political opposition (Intel anyone?). It's always going to be the Cinderella of the ball.
What are you talking about? USB is slower than SerialATA not faster... Much much slower in fact.
Here's a quick comparison
SerialATA 1.0 - 1.2Gbps (150Mb/sec)
USB 2.0 - 480 Mbps
USB 1.1 - 12 Mbps
Firewire (IEEE1394) - 400 Mbps
Parallel Port - 1 Mbps
Serial Port - 0.115 Mbps
Figures taken from the actual spec on serialata.org and from here.
Nick...
> And that's not even mentioning that WD
> probably has the most unreliable, loudest
> drives on the market.
I strongly disagree with this statement. Unlike Fujitsu and IBM, Western Digital do not have the reputation of making unreliable drives.
Their new drives feature Fluid Dynamic Bearings and make almost no noise whatsoever (I have 80Gig ones in my computer). You can just hear it spinning up if you put your ear to the case, but I promise you it's silent from then on - even while it's moving the heads.
Nick...
Basically this means that rather than treating the wire as a fixed capacitance, inductance, resistance, it must be treated as a distributed system. Each dx has a dc, dl, dr. The longer the wire the higher the impedence. Now you have to take into account the bundling the wires ans assuring that they are all equal length and impedence. This is why IDE went from a 40 pin connector to an 80 pin conector. The data pin count remained the same, but the grounds increased.
Next you have to take into account the drivers and receivers. Each has certain variations in their physical properties. Taken as an individual, you minimize the tollerances. BUT when you add many of them in parallel, the tollerances add as well. The end result is that the overall speed is limited due to the summation of tollerances in the wire AND in the silicon. This is physics and no real way around it.
You forgot the other big advantage.....nice thin wires that are easier work with in the case. I KNOW someone will mention the round ATA cables, but those don't bend so well. The S-ATA wires are more like the wires for the CD-Soundcard. They are much easier to work with. :)
All 4 'cables' in my fiber-optic patch-cord? Not all the world is copper.
Free Online Dark Fantasy RPG - http://www.blackmud.com