Seagate and AMD have demonstrated a 6Gbit/s link between SATA disks and a server, double the speed of the current SATA 2 spec and the same as the fastest SAS speed. The justification for the faster speed is to try and keep up with the accelerating growth in capacity - and cache capacity - of SATA disks, now at the 2TB level and …
Bits and bytes
Would 6Gbits/sec not be 768Mbytes/sec, rather than 600Mbytes/sec?
The healthier alternative?
"6Gb/its/s == 768Mbytes/s"
Virtually all of these new serial protocols use what's called 8b/10b encoding. In order to send a single byte, 10 bits are used. The extra bits are used for error correction, DC balancing and clock recovery.
(It's the same with PCI express, SAS and USB3.0)
600MB/s pah ... rubbish
2GB/s is where it's at...
(yes, yes, I know it's 24 SSDs rather than one drive but ... it's just a link... and it is rather good)
Re: 600MB/s pah ... rubbish
OMG geeks. Oh wait they used Windows.
maybe we can start using all the 3Gbps first...?
Not to be the dude to say that progress isn't good or anything but.... Harddrive speed hasn't change from EIDE to SATA... the connection has, the bandwidth has, but the speed hasn't. Though with this faster speed would it mean TWO drives on a single SATA Connector or would that be too much to ask for?
While the bandwidth is nice, there will be no speed difference between the 3Gbps and the 6Gbps at all, because these drives are still spinning at 7.2k rpm and still seeking around 4.0ms, unless THAT is sped up there's little difference in speed...
Though I would like to see a study how the high capacity affects seek-time. IE: Does the head moving over the drive on a 2 TB drive seek more data than a 500GB drive in the same distance?
Save your pr0nz
In record time!
Obviously a drive with the same physical disc size and double the capacity will be able to read data off the disc faster, up to double the speed. It might be because there are two physical discs inside when previously there was one, with two heads reading in parallel; or it might be that the storage density has doubled so that twice as many bits pass under the head in the same time for the same rotational speed. That's not to say that any of these things are actually done in real drives, or that the transfer speed really is higher in real drives, but you can see that at least in principle a speed increase is *possible* without increasing the rotation speed.
@Dale & Jordan
@ Dale - Technically, doubling the areal density will only increase the transfer speed by 41% (ideally), since the density increase will be evenly spread along both the x and y axis - 41% more bits for a given length of track, but also 41% more tracks for a given width of disc - and the head only reads/writes one track at a time.
@ Jordan - Hard drive speeds HAVE increased since EIDE - there are now conventional rotational hard drives that would be ill-served by SATA-I-1.5Gbs. And don't forget SSDs! The Intel X-25E can achieve 250 MBs, which using 8b/10b is 2.5Gbs - that's close enough to the limit of SATA-II-3Gbs to make it justifiable starting work on the next generation.
OCZ Vertex2 - 550MB/s
i am still waiting for the new OCZ Vertex2 drives ... 550MB/s read spead and i guess a lot of other SSD vendors will come out with similar products soon. Building RAIDS directly into SSD sounds like an easy to do thing and will make the new SATA3 standard already look old:
Finally someone with some sense comments on one of these things. As soon as I started reading I knew someone would be bound to pop up with something like Jordan's comment.
In the end you absolutely need to develop and deploy a higher bandwidth interface before you make drives that run over it. If you do it the other way around then you end up with disks which have a bottleneck at the interface, instead of at the r/w heads, which makes it pointless to develop faster drives. And it does take a couple years for a new interface to become wide-spread, which is what a storage maker would be looking for before investing too much in drives which require the speed.
And people assume that working on faster interfaces some how means people aren't working on faster drives. Which is wrong for many, many reasons. I don't think I'll ever understand why people complain about progress.
Poundland joins HP, Dell, IBM in server biz
I was thinking of SSD style RAID, but done cheaper, gaining speed and redundancy from sheer numbers of very small cheap devices.
A few hundred poundland memory sticks, plus a few hundred cascaded poundland USB hubs. Add software RAID 5 in Windows, and Robert's your cousin's father!
What could possibly go wrong?
re: Suburban Inmate
I know you were not being serious, and not to belittle you but firstly iirc (and usb2 may be different) usb is limited to a chain of 127 devices per port on the device controller. Secondly 480mbit is crap next to SATA, thirdly the bandwidth is shared between everything on the chin and finally there is a cpu hit for usb. and i know cpu's are fast and all but its going to be a big hit with hundreds of things on the line.