SPCR

http://www.hardwarezone.com/articles/vi ... =1147&pg=3

400gigs of goodness

And four spinning platters of loudness, unfortunately.

Not to mention eight seeking heads of loudness.

Erm, doesn't 133GB / platter = 3 platters for 400GB (6 heads)?

Granted that 3 platters will be noisier than 1 or 2 platters, but it's still quieter than 3 1-platter drives.

And while on the subject of multiple heads, why don't HDD manufacturers use the heads in parallel (similar to RAID 0)? I believe they don't because there's no appreciable difference in transfer rate between 1 platter and multiple platter drives.

Wow wtf I didn't even imagine they could do that...

Why are they only selling one drive with 100GB/platter (ST3200823AS, 200GB), when they can already do 133GB/platter?!? Where're the 100GB and 50GB models? Where're the 66GB, 133GB models?!?

*Scratches head*

Still, 3 platters/6 heads is bound to run louder and hotter than any of the current models.

-Ed

EDIT: Btw the reason is that one servo/head would be almost impossible to fit and prohibitively expensive. This way it's one servo for all heads.

Edward Ng wrote:EDIT: Btw the reason is that one servo/head would be almost impossible to fit and prohibitively expensive. This way it's one servo for all heads.

They don't need to move the heads independently. You just need to read data from all the heads in parallel instead of individually. But you're probably right about the expense part. Hard drives are a commodity these days and every cent in savings help the bottom line.

What's the likelihood of multiple pieces of needed data being on the same track of different surfaces needing access at the same time? That's why it would take multiple servos; each head needs to seek to a different portion of its duty surface anyway.

-Ed

---edit--- erased

http://faq.storagereview.com/tiki-index ... ernalRAID0

says it much better

"What's the likelihood of multiple pieces of needed data being on the same track of different surfaces needing access at the same time? That's why it would take multiple servos; each head needs to seek to a different portion of its duty surface anyway."

Think how (not-real) raid0 works, it'd write 1 bit on each drive at a time... so overall you're reading/writing 2 bits at a time instead of 1. You have twice your transfer rates.

Think of your normal 2 platter harddrive (4 sides used), when one head writes a bit, the other 3 heads are idle, what if all 4 read/write in parallel.. instant raid0. THey wouldn't need to seek to different spots, data would be written together, and would thus be able to be read together.

Thats the idea anwyas, still doesnt' work for other reasons.

Yes, but what you're saying assumes that once a file is written in place, it does not change from that place. Remember that files are frequently modified, so let's say that one file on a two-platter drive takes up four blocks (optimal condition); the original file would write to both platters at once, both sides of each platter at the same time, in the same spot; suddenly down the road you make a change to only one block in the file; the problem is suddenly the block has become two blocks worth of data (old data from one old block replaced by two blocks worth of data); that old data is taken out from the file table but the two new blocks are written elsewhere. Next time the drive goes to try and read this, it still has to seek at least twice.

With a drive that has completely separate servos, this wouldn't be a problem, just as true RAID with multiple drives, but in a drive with one servo for multiple heads, it remains a problem; as soon as a file is modifed or portions of it are moved out of track, the single-servo-multiple-heads simply won't serve. As they said, multiple servos just isn't economical.

Of course, activating TCQ or NCQ in such a drive would benefit greatly, but still, it won't entirely solve the problem like multiservo head assembly would be (imagine a multiservo head assembly in a drive WITH NCQ, hooked up to an NCQ-enabled controller and optimized drivers!!!).

It's a shame, but some things in life just aren't that simple.

-Ed

EDIT: Typo.

They could have done this internal RAID, I guess. That is, they could store simultaneously an exact copy of the data onto 2 platters. However, this would not make much difference because most of the time the hard disk fails as a whole, I think. This would help against some bad blocks but would not protect you from the heads failure, servo failure or electronics failure. For a RAID, it would be much more reliable to use separate drives or even separate controllers.

What about acoustic management? We've all heard about the legal problems that Seagate have faced, but what are they going to do in the future? AAM basically helped define Seagate's products. They really need to resolve their legal problems or come up with something else.

Does anyone here have any further info/rumours?

"emember that files are frequently modified, so let's say that one file on a two-platter drive takes up four blocks (optimal condition);"

Its not a block per platter side, its a 1/4 of the block per side... you're drive is using 4KB block sizes... you go to write one, in the end each platter ends up with 1KB of the block.

Its not just allowing the heads to work together when it comes in handy on a current HD, the heads would always work at a unit. Go to write one sector and the data ends up on 4 sides.... maybe easier(?) would be writing one sector per platter side... not sure who would want <2KB block sizes anyways.

In that case, Becks, I really don't know if there's any real way to predict the improvement in performance from it; the way it's optimized in the logic would be extremely critical.

Remember that standard RAID-0 works in stripe sizes, and does not simply split every single block across all drives. My understanding is that we're talking about a single drive performing RAID 0 via multiple servo'd heads; it would split each stripe across multiple surfaces. What you're talking about is completely different all together.

-Ed

EDIT: Hmm...somebody must've posted a reply and then deleted it...

Anyway, 20 bels for the single platter 120 GB (Why not 133 GB ???) Barracuda 7200.8 is better than 25 bels for the single platter 80 GB Barracuda 7200.7.

Wouldn't it be great if those numbers turn up true?

Tom's Hardware Gimp...I mean Guide elaborates on the new Seagate releases for Q304.

-Ed

Hello:

Storage Review has the announcment, too:

http://www.storagereview.com/

It'll have command queing, and native SATA 2 (is that the 300MB/Sec version?).

Should be.

Should be.

Yep there's an echo in here for sure.

Also, the 7200.8 series has an 8.0 ms average seek time (down from 8.5 in the 7200.7) and has 16MB buffers available. I'm getting the one-platter 120GB version!