20k RPM Raptors?

If there not quick with the release timeframe then it could be a wasted development. SSD's will hit 256GB this year with transfer speeds 200MB+ and seeks of >1ms. Prices of course are a major barrier, but these will drop at a fast rate also.

IMO consumers looking for storage will continue to buy the cheapest SATA drives in 500//750/1000+ GB ranges. Those looking for speed will buy smaller (32/64/128GB) SSDs for there OS/Games and use "storage" drives for everything else
 
The spindle speed has always been a sore point. If you compare harddisk development within the last 10 years or so, we have:

Capacity 10GB -> 1TB: 100x
Sustained data rate: 15MB/s -> 150MB/s: 10x
Spindle speed: 10kRPM -> 15kRPM: 1.5x

I have no values for full stroke, but I think that's about 2-3x as fast nowadays.

Especially the first two parameters mean that it takes us 10x longer to read out a harddisk compared to 10 years ago.
 
No way this is going to happen. Considering that Seagate and other harddrive manufacturers are jumping on the SSD bandwagon, I'm sure the next iteration of the "velociraptor" will be solid state based.
 
Can't you just plug both the SATA power and the molex power plugs on existing 10k Raptors to achieve 20k spindle speeds?








:D

 
The spindle speed has always been a sore point. If you compare harddisk development within the last 10 years or so, we have:

Capacity 10GB -> 1TB: 100x
Sustained data rate: 15MB/s -> 150MB/s: 10x
Spindle speed: 10kRPM -> 15kRPM: 1.5x

I have no values for full stroke, but I think that's about 2-3x as fast nowadays.

Especially the first two parameters mean that it takes us 10x longer to read out a harddisk compared to 10 years ago.

Spindle speed is purely a mechanical limitation though, where as capacity and data transfer are electrical.
 
Spindle speed is purely a mechanical limitation though, where as capacity and data transfer are electrical.

Nope, they're physical: they're both a function of how close together the bits are packed on the platter.
 
Nope, they're physical: they're both a function of how close together the bits are packed on the platter.

Really? Are the bits on the drive little LEGO blocks that the heads switch ON or OFF to make up the data?
 
It seems like if they took the new 300gb Velociraptor with the same density platters and dropped the capacity to 150gb and put a faster motor in it they might do it. Trading density for speed seems the way to do it.

I'm encouraged to hear 20k instead of 15k rpm though!
 
Really? Are the bits on the drive little LEGO blocks that the heads switch ON or OFF to make up the data?

Not quite. They're magnetic domains that are oriented one way or the other. But the way they're manufactured makes a correlation to a physical group of atoms and a magnetic domain. So the reason for platter densities being what they are is finding ways to make smaller groups of atoms that are willing to line up the same way. In other words, a physical phenomenon.

If you want to argue that the magnetic domains are the important part and not the metal platter that they're induced in, be my guest. It's sort of a philosophical question at this level.
 
Jump on teh SSD bandwagon please WD

I see the current 64GB SSD's are around $1000, so the samsung 256gb 200mb/s read will be at least $3000-4000.

128gb SSD for around the 500 bucks would be the sweet spot for most people, though lord knows when that kind of $/GB will come round. That kind of size is enough for OS, apps, and a shed load of games.
 
When the time comes that I can buy a 1TB SSD for less than $200 I will jump on faster than a fat kid in a candy store. Until then, I will continue using my 500GB HDD.
 
agreed .. & fat kids will be a few pounds trimmer until that fateful day :)
or until McDee's has doublecheeseburgers for $.99 - oh nvm they do... until 1/4lb's are $.45 .... or some other similar & group negotiated price :p
 
IMO if they want a fast raptor, they should make a Hybrid, then you get the best of both worlds.


They should take like a 640AAKS and toss in 32gb or so of SSD... you get your quick access times and you get your large storage capacities.
 
IMO if they want a fast raptor, they should make a Hybrid, then you get the best of both worlds.


They should take like a 640AAKS and toss in 32gb or so of SSD... you get your quick access times and you get your large storage capacities.

for some reason those dont sit well with me. could be the over complex and the multiple points of failure thing. Simple Stupid is a time proven formula that works.
 
for some reason those dont sit well with me. could be the over complex and the multiple points of failure thing. Simple Stupid is a time proven formula that works.

I guess when you want to mimick the real thing, it's probably better just going for the real thing :)
 
IMO if they want a fast raptor, they should make a Hybrid, then you get the best of both worlds.


They should take like a 640AAKS and toss in 32gb or so of SSD... you get your quick access times and you get your large storage capacities.

Only quick if it's accessing the data that is on the 32gb SSD portion though.
 
I never understood why they made the new raptors smaller. Smaller platters means the outside edge is traveling slower right? doesnt that mean slower transfer rates and cutting off the fastest part of the disk? Ive been boggled every since I saw the new smaller raptors. Heck if they did get them to 20k RPM the outside edge probably wouldnt be all that much faster than a full size 15k drive right? (or I may be horribly scewing up that guesstimation math).
 
IMO if they want a fast raptor, they should make a Hybrid, then you get the best of both worlds.


They should take like a 640AAKS and toss in 32gb or so of SSD... you get your quick access times and you get your large storage capacities.

The problem with hybrid disks as a concept is they're dumb. There isn't enough room in the budget to put enough logic to make it perform well. The best solution is an operating system that treats flash like flash and disk like disk, since the processor has plenty of time to go around, and it's updateable with new logic.
 
I never understood why they made the new raptors smaller. Smaller platters means the outside edge is traveling slower right? doesnt that mean slower transfer rates and cutting off the fastest part of the disk? Ive been boggled every since I saw the new smaller raptors. Heck if they did get them to 20k RPM the outside edge probably wouldnt be all that much faster than a full size 15k drive right? (or I may be horribly scewing up that guesstimation math).

A "full size" 15k drive often only has smaller platters in it. The speed of the edge of the disk isn't as big a performance indicator as it once was: with higher densities, transfer rates go up regardless of what the spindle rate is. These days the seek times matter more than ever, and having a smaller physical platter means smaller maximum distance to travel and more IOPS per cubic meter.

A 20k drive would be moving exactly 4/3 as fast as a 15k drive. Is that "much" faster? I don't know, it depends what you're doing.

If WD is indeed making a 20k drive, it strikes me as a desperation move. Flash is coming down in price as drastically as ever, and one of these days WD will be just plain obsolete.
 
But having more disk on the outside part wouldnt hurt the seek performance of the inside part. There would just be a slower seeking, faster transfer extra portion of the disk. Making it 500gb or so? Maybe its based off 640? I think these old raptor 74gig drives we have are actually 80gig drives with the inner most power disabled to increase benchmark read / write scores.
 
But having more disk on the outside part wouldnt hurt the seek performance of the inside part. There would just be a slower seeking, faster transfer extra portion of the disk. Making it 500gb or so? Maybe its based off 640? I think these old raptor 74gig drives we have are actually 80gig drives with the inner most power disabled to increase benchmark read / write scores.

But if the disk is physically bigger, the random seek rate goes down for the whole disk, provided you don't short-stroke the drive and defeat the purpose of making it bigger. And bigger platters tend to wobble more, which makes them harder to read from. The heads need to be very close to the disk surface, so wobbling is really bad for them.
 
WDC is about as likely to make a real 20kRPM drive as I am to do an about face and openly embrace Sun's new business guess^Wmodel.

WDC has always been dead last in R&D, contrary to popular ignorance - the 10k Raptors were old Enterprise parts with new PCBs, period. The new stuff is just now catching up to FC and SCSI drives that have been around for over 2 years.

In other words, this is pure garbage, and there is absolutely no chance of WDC producing a 20kRPM spindle.
 
Well, several years ago, some people thought that the 15,000RPM IBM disk would be a joke. It took an awful lot of time until they hit the market with a little less speed than originally advertized, but they made it! BTW: Mine is dated 23-DEC-01, unfortunately, it crashed too soon.

Now 15kRPM disks are available in a 2.5" 15mm form factor.

I say: They'll do it. Maybe with slightly smaller platters (1.8"-2.0"), but that won't matter, thanks to higher densities.
 
For one thing, 15k RPM is approaching the physical limits of drive technology. When trying to go above 15k RPM, the speed of the platters introduce minute vibrations that the read heads can not handle.
 
Well, that's the reason why they reduced the diameter of the platters.

The latency is a *serious* problem with todays harddisks. Even the VR shows "poor" random access, simply because one full revolution takes 6ms - in words: SIX MILLISECONDS. That's more time than a full stroke! No matter how fast the actuator and how short the track/head skew, the 10,000RPM add 3ms average. At 15k it's 4ms max, 2ms average, at 20k it's 3ms max, 1.5ms average. That would be a *huge* improvement.

There are power tools - according to the online dictionary "router" - where the interchangeable tool spins at 32,0000-36,000RPM. Should be no problem for two or three fixed platters of 2" diameter.

Let's wait and see:D
 
I don't think you guys understand what kind of an engineeiring obstacle spinning a platter faster than 15K really is.
The enterprise market drives all of this and they cannot break the 15K RPM barrier. There are HUGE basic mechanical hurdles to get platters spinning that fast. If they could, I'd already have 20K drives in my servers. Period. We have been "stuck" at 15K for many, many years now.
 
Back
Top