One question on serial ATA

[phynicle]

i was just wonder will serial ata drives match up to it's supposed bandwidth of initially 150mb/sec????
since it'll basically feature the same technonlogies as the current spinning drives...how much difference will we see then?

 

[Saem]

I'm not sure, perhaps they can pulls some tricks, by hiding behind the serial latency of request completetion. At the same time, I'm not sure what changes have been made at the drive implementation level, if serial ATA calls for a more SCSI like design where the drives are smarter and can fulfill parallel requests we might see some sustained throughput speed ups. It all depends on what serial ATA brings to the table, besides the new I/O interface.

From my understanding serial ATA blows chunks -it's nothing special- and serial ATA II is a low cost SCSI killer.

 

[Althornin]

From my understanding serial ATA blows chunks -it's nothing special- and serial ATA II is a low cost SCSI killer.

CAre to elaborate further?
Links? I'm dying for osme more info here

 

[Saem]

This is all I could muster up at the moment...

http://www.anandtech.com/showdoc.html?i=1593&p=2

I'll find some more about it later.

I took a stab in the dark and typed www.serialata.org in my browser and there was the homepage. Anyways, here is a pdf on serial ATA II, it's better than the SCSI we have now and it won't break the bank. It'll just take some time to get down to the desktop. >=(

http://www.serialata.org/news/S2why.pdf

 

[Dave Baumann]

Anyways, here is a pdf on serial ATA II, it's better than the SCSI we have now and it won't break the bank.

Of course, by which point SCSI320 will have proliferated and they'll be looking at SCSI640!

 

[phynicle]

so we'll only be seeing little improvement on the side of hd speeds
sustained throughtput of about how much mb/sec???
and what is thes best hd on consumer markte and it's mb/sec

 

[Rob M]

Faster Drives

It may be that HDD companies decide to use 10K IDE drives to push Serial ATA. In which case there would be an indirect speed benefit.

 

[phynicle]

but still only spinning drives
what's the point of more bandwidth when you can't even fill it up?

 

[Althornin]

To decrease acces times, instead of only increasing rotational speed, why not add another set of heads onto the drive? This could also increase read and write times if the drive was designed properly...or is this to difficult of an engineering problem?

 

[Saem]

Does IDE support multiread and writes?

 

[phynicle]

i seem to remember that current drives depending on how many plates they have have multiheads already...but you don't see them adding any more efficiecny that just adding a couple mb/sec to the sustained read/write

anyway it just cost alot more too

 

[pcchen]

It is quite nature to think that more r/w heads bring more bandwidth. However, the unfortunate reality is, drives with more r/w heads are usually slower at seek time.

The main reason behind this is, that these extra r/w heads all share the same actuator. That is, they all move together. Therefore, unless you can align all platters perfectly, only one head can become active at one time.

After all, it is much cheaper to just buy another HD to double the bandwidth.

 

[MfA]

Dunno what multi-read/writes are supposed to be for sure, but Ill make a slightly different guess than the previous two posters (since their replies are irrelevant to the actual interface). In theory ATA allows you to que multiple reads and writes, with data being routed through bus-mastering of course, in practice the spec is a little fucked up in that respect ... few drives support what little there is of it, fewer controllers and even less drivers.

Of course theres good reason for manufacturers to keep it broken ... seek time alone wont sell SCSI drives.

 

[phynicle]

thanx for cleaering some things up pc chen and mfa
but still i would like to see other developments apart from the current spinning drives

anyone know of any other developments being done to replace these drives apart from solid stat hard drives?

 

[Saem]

Well the problem is that spinning media is the way to go for now. Or atleast so is my impression.

A while back Nec or Hitachi came up with plastics that can have data density many order of magnitudes greater than what's possible today. This means if you can keep rotational speeds steady, throughput of drives will sky rocket. Big if. Other than that, holographic, plastic (RAM on potato chip bags and whatever else are all possiblities, I suppose.

 

[Joe DeFuria]

Eh....screw bandwith and performance improvements...I just like the cleaner cable design of Serial ATA. I will be dancing in the streets the day I can rid my computer case of ribbon cables.

 

[Saem]

Eh....screw bandwith and performance improvements...I just like the cleaner cable design of Serial ATA. I will be dancing in the streets the day I can rid my computer case of ribbon cables.

Yeah, a new place to hide the dirty magazines.

 

[Dave Baumann]

I just like the cleaner cable design of Serial ATA. I will be dancing in the streets the day I can rid my computer case of ribbon cables.

I saw a link the other day (sorry, can't remember it now) to a Far East site (so I couldn't read the text) but they had a picture of what appeared to be a standard IDE to serial ATA adapter. it was just a small board that was just about th hight of a drive that plugged directly into the back of the drive and the serial ATA connector plugged into that; it looked like it took power from a fan connector.

 

[Gubbi]

Drive bandwidth today, is limited primarily by the sensing amps and the IMR heads.

State of the art bandwidth today is 49.3MB/s, this is for the outer rim on 100+GB ATA drives. Inner rim throughput is <30MB/s. So serial ATA (basically FireWire technology) with it's 50MB/s bandwidth should be adequate. By the time harddisk vendors develops next generation platters and heads, we will be looking at a serial ATA upgrade, just like we are seing with FireWire-II.

We are not going to see multiple heads (per platter) in one drive package, this is a cost issue.

People who really need better access time, will go for more spindles like they have always done. The SCSI disk market reflects this; volume is in small to medium sized drives.

Cheers
Gubbi

 

[Saem]

If you do a quick search on Anandtech for their IDF report, they show adapters for serial to parallel and parallel to serial. So one could technically get a nice and thing serial ATA cable, draw that to a parallel ATA drive and then just hook up the adapter, so that should clean things up in the case.