Has HD innovation stagnated?

[Sideswipe001]

I was just looking at StoarageReview.com and was thinking about a few things about hard drives. Hasn't it seemed like latey everything's stuck, speed wise?

The Raptor is the fastest home HD. It has been for months. No one has released anything to challenge it. Sure, there have been some 16 MB Cache drives, but from what I understand, they didn't help speed that much. What's the hold up?

In the SCSI world, the same holds true. How long have 15K drives been around now? And the fastest SCSI drive, the Fujitsu MAS series, has been out since April of 2003. Over a year and a half, and no newer SCSI drives have come out to challege the speed throne there.

Have the HD companies run out of ideas? Even SATA hasn't really sped anything up on the desktop front. Sure, more motherboard support it, but where are faster drives that actually use the extra bandwith SATA offers? Even RAID isn't really helping people at home, as Anand proved. So again, what is holding everything back?

 

[Jeff7]

HD technology itself is very old - it's still just a spinning disc with a head moving over it to read data. It's a few decades old now. CD drives are basically record players - instead of threads on the surface, there are pits, and the needle is now a laser. So yeah, that's stagnation in a sense.😛
They just keep upping the rotational speeds, areal densities, and the speed and accuracy of the R/W heads, while reducing the size. It's just constant refinements of the process.
There are some tricks that can be implemented in the drive's circuitry, like command queueing, and larger buffers, but it's still nothing major like a jump to holographic storage would be. And that's a long way off.

 

[AyashiKaibutsu]

Progress on harddrives will be stuck at the rate of growth it's at until they find a better way then a spinning magnetic disk. You can only spin a disk so fast. Also, there isn't too much need for rapidly increasing HDD speeds since they only really affect loading times and certain hdd intensive applications.

 

[Matthias99]

Originally posted by: AyashiKaibutsu
Progress on harddrives will be stuck at the rate of growth it's at until they find a better way then a spinning magnetic disk. You can only spin a disk so fast. Also, there isn't too much need for rapidly increasing HDD speeds since they only really affect loading times and certain hdd intensive applications.

Well, it's not that there's no need for faster storage (maybe not for casual users, but when you get up into workstation/server/enterprise storage systems, there is), but that it's proven far easier and cheaper in recent years to enhance performance through solutions like RAID than to wait for hard drive manufacturers to come up with bigger and/or faster drives.

If you have an application that's bound by STR or IOPS, RAID solutions can increase your drives' effective speed many times over in these disciplines. If you're limited by response time, generally the solution is more RAM and more intelligent caching (including bigger disk caches) -- even if you halved current disk response times (which would take 30KRPM disks!), you're still going to be thousands of times slower than system RAM, and millions of times slower than CPU cache.

 

[Sideswipe001]

Originally posted by: Matthias99
If you're limited by response time, generally the solution is more RAM and more intelligent caching (including bigger disk caches) -- even if you halved current disk response times (which would take 30KRPM disks!), you're still going to be thousands of times slower than system RAM, and millions of times slower than CPU cache.

Which is generally my point. We have a windows 2000 server, here at work. It takes 5 minutes to reboot. That is an extremely long time, to me, for sitting around and waiting for the server to come back up again. Granted, we don't reboot it much - it needs to be running! - but it's still wait time.

Could you imagine exactly how much more responsive even a "casual" computer user's computer would be, if you could run the entire OS, programs, etc. at the speed of RAM?

HDs are by far, the slowest of the "core" computer parts. Everyone would notice a difference if they were suddenly twice as fast as before. Hence, my impatience to see the technology go somewhere.

Yes, there is a speed difference between HDs of today and HDs of 5 years ago. But 5 years ago, UDMA was still increasing in speed too- UDMA66 (3), to DMA100 (4), to DMA133 (5). Though there was little need for the jump from 4 to 5, there was a need from DMA33 to 66, and 66 to 100. Now, we have a pipe for the HD that's bigger than any HD in the world can fill by itself.

 

[mrweirdo]

Sooner or later HD technology will have to die as we will need something beter.
I'm currious to see if some sort of flash memory will ever be able to replace the standard spining disk drive we have now.

 

[Matthias99]

Originally posted by: Sideswipe001

Originally posted by: Matthias99
If you're limited by response time, generally the solution is more RAM and more intelligent caching (including bigger disk caches) -- even if you halved current disk response times (which would take 30KRPM disks!), you're still going to be thousands of times slower than system RAM, and millions of times slower than CPU cache.

Which is generally my point. We have a windows 2000 server, here at work. It takes 5 minutes to reboot. That is an extremely long time, to me, for sitting around and waiting for the server to come back up again. Granted, we don't reboot it much - it needs to be running! - but it's still wait time.

That's a LONG time for a bootup; my system gets back to the WinXP login screen in about 60-70 seconds from a cold start. Unless you're including time to load applications, etc. after the system has gotten into Windows in your 5 minutes.

Could you imagine exactly how much more responsive even a "casual" computer user's computer would be, if you could run the entire OS, programs, etc. at the speed of RAM?

Well, if you have enough RAM, then everything WILL run at the speed of RAM once it's been loaded the first time -- and unless someone finally makes some inroads with fast, cheap static RAM (I know IBM has been doing research on fast magnetic-core RAM), you have to load it from *somewhere*. A faster disk will make things load faster, but once an application is loaded, it generally stays in RAM. If you're using something like Photoshop or Premiere, and you spend a lot of time waiting while loading/saving large files, then a faster disk would help you -- but this is somewhere that RAID can speed things up tremendously.

HDs are by far, the slowest of the "core" computer parts. Everyone would notice a difference if they were suddenly twice as fast as before. Hence, my impatience to see the technology go somewhere.

Well, you *can* get expansion cards that will emulate a hard drive using RAM (they load a specially-configured disk partition into RAM on boot, like RAMdrive software does, except in hardware), and solid-state hard disks of fairly decent size. They're just very expensive relative to standard hard drives. We just currently can't make cheap, fast (that is, close to RAM speeds) static bulk storage. The drive for the last 10 years or so has been to develop either cheap, slow bulk storage (hence 200+GB hard drives at < $.50/GB), and fast, expensive bulk storage (solid-state drives, SAN and NAS storage arrays). I'm sure the average user would *like* to have faster storage, but given the choice between faster storage and more storage, most people would choose more storage.

Yes, there is a speed difference between HDs of today and HDs of 5 years ago. But 5 years ago, UDMA was still increasing in speed too- UDMA66 (3), to DMA100 (4), to DMA133 (5). Though there was little need for the jump from 4 to 5, there was a need from DMA33 to 66, and 66 to 100. Now, we have a pipe for the HD that's bigger than any HD in the world can fill by itself.

Well, bigger data pipes aren't there because there isn't anything to fill them with! With SATA2 and SAS, though, you'll have much more bandwidth for attached drives.

 

[FishTankX]

Honestly, I think you all are expecting too much of HDD technology. HDD is not like solid state transistor technology. Important technical breakthroughs need to be made for you to be able to have higher data desnity. Eventually it might get to the point where room temperature would be enough to flip extremley dense bits of data around. We're lucky we haven't hit that point yet, but we're going to someday.

It's not like Lasers where you can use a smaller laser and get smaller pits. And boom, obvious improvement. I think HDD's are doing their job admirably. Storing large amounts of data permenantly for loading into RAM when it's needed. If you had enough RAM, after everything was loaded the speed of the harddrive would be largely irrelevant. I bet, in most games, you play about a couple of hundred times longer than you load. What's the rush?

The real emphasis should be made on making some sort of extremley fast static RAM.

Theoretically if you really wanted to reduce access times, you could build an SATA controller to string about 20 512MB compact flash Sandisk Ultra III cards together in RAID0. That would give you response time faster than even the fastest 15K RPM harddrive and give you throughput of over 200MB per second. It would cost you about a grand and a half and give you 10GB of storage. Wouldn't be much larger than a normal harddrive. Cache the most commonly used system files on there, along with the most commonly used program files and give the system a good chunk of RAM, and you're golden.

Solid state is where it's at. It's harder to get platters to spin much faster without having them explode.

 

[Elcs]

I think that we are nearing the end of what HDD's can do. Granted we have seen recent upgrades by way of the Raptor increasing rpm's to achieve speed and so on.

The 5 minutes bootup time struck me hard. Reminds me of my Athlon 700. It seems to take 2-3 minutes to boot when my Barton system will be up and running in under 1 minute (approx. just guessing).

Perhaps hard drives need to be more application or should I say user specific. I read some information on 16mb caches on HDD's and saw people arguing that it would just slow things down and such. Perhaps certain people will buy 16mb cache or 32mb cache HDD's for mass storage where response times arent that important and raw throughput is. Perhaps certain people may buy a specific HDD for booting up or transferring MP3's or gaming.

That scenario is unlikely to happen. It'd be like AMD... having 4 or 5 different 3200+ chips running at different speeds with different caches and different extra bits and bobs.

Hard Drives have been around for as far as I can remember. 20mb HDD in my dads Amiga1200 was phenomenal for its day. Now we are seeing 200-300 Gb HDD's.... way more than 1000 times more capacity. In 10-15 years, we've advanced that far. Looking at the speed differences over the years, theres a definate tangible difference between my dads 20mb HDD and my slightly aged WD SE 80Gb. Id be grateful that HDD's have taken us so far, so quickly.

All good things must come to an end. Hard Drives have lasted and will continue to last in PC's worldwide for a long time to come. The thing is, it looks as though for sheer storage, HDD's are the only way to go for a few years.

 

[piasabird]

Instead of buying 4 hard drives to make a raid array, it seems like they could make a hard drive with multiple disks do the same thing with firmware.

If they can make a processor look like 2 virtual cpu's why not a hard drive?

 

[Matthias99]

Originally posted by: piasabird
Instead of buying 4 hard drives to make a raid array, it seems like they could make a hard drive with multiple disks do the same thing with firmware.

If they can make a processor look like 2 virtual cpu's why not a hard drive?

You could, but it probably wouldn't fit in the standard 3.5" or 5.25" form factor spaces for drives. You'd need multiple actuators/heads so you could do multiple reads or writes in parallel, and that takes more space (and uses more power, and generates more heat, plus there are technical issues with keeping them in sync, etc.)

At least thus far, it's been easier in practice to mass-produce common drive designs and get external hardware/software to strap them together when such a configuration is needed.

 

[SUOrangeman]

I think the innovation sought is more evident on the capacity side. The first 400GB 3.5" drives are now available, and 250GB is becoming commonplace, and very price-competitve. If you're not gonna buy a Raptor anytime soon, would you really consider anything smaller than 250GB (closeouts, mis-pricings aside)?

-SUO

 

[Sideswipe001]

The only problem I have with size, is that some people just plain don't need that much size. I built my parents a computer with a 120 GB hard drive. They use less than 10 GB of it, and I really don't see them using any more than that in the near future. They care more about how long it takes the computer to turn on or install something than how much space it has, because it already has more space than they can possibly use.

I would rather have a 25 GB HD that could boot in 10 seconds than a 400 GB one that boots in 40. I just plain don't need that space.

If there was a way to install Windows onto a RAM drive, I would try it, just to see how fast it went, even knowing that it would disappear if I shut the computer down.

 

[FishTankX]

What I see happening soon is MRAM caches. Maybe in a PCI-E 1x slot or something, even that would be good enough. It would just cache the most commonly used files. this would be almost as fast as RAM but it would still be there when you turned off the computer. It could also mirror the contense of the RAM when you hibrenated. Can you say near instant boot? Just copy the MRAM into the real RAM. Shouldn't take more than half a sec.

 

[LifeStealer]

Eventually the disks will spin so fast they go back in time and you find yourself playing more and more escape from monkey island.

 

[thegimp03]

lol...cheap hard drives are good...I'm sure as speeds get higher heat dissipation increases significantly. This may be a problem.

 

[imported_jediknight]

Originally posted by: Sideswipe001
The only problem I have with size, is that some people just plain don't need that much size. I built my parents a computer with a 120 GB hard drive. They use less than 10 GB of it, and I really don't see them using any more than that in the near future. They care more about how long it takes the computer to turn on or install something than how much space it has, because it already has more space than they can possibly use.

I would rather have a 25 GB HD that could boot in 10 seconds than a 400 GB one that boots in 40. I just plain don't need that space.

If there was a way to install Windows onto a RAM drive, I would try it, just to see how fast it went, even knowing that it would disappear if I shut the computer down.

Why not use a Raptor or a small SCSI drive?
(Question: where can you get inexpensive SCSI controllers.. and what cost is reasonable?)

 

[FishTankX]

Any cost under $150 is reasonable but the universally accepted budget controller is the $33 LSI LSIU160

 

[gsellis]

Originally posted by: LifeStealer
Eventually the disks will spin so fast they go back in time and you find yourself playing more and more escape from monkey island.

Well, close. Our theory is that resultant force from the torque during a head crash will turn the PC into a projectile. Head crashes, PC flies across the room causing serious damage. 😀

 

[dmw16]

Just as a correction, we are a while off from platters exploding do to rotational speeds. We did some calculations in my advanced mechanics materials class and found that CDs could explode at anything beyond ~30k RPM, but hard drive platters can spin a bit faster than that before they go boom.

Beyond that, I think that the issues with hard drive speeds in a lack of innovation is something that the entire industry will have to deal w/ at some point. CPUs are getting hotter and the .09nm processes are pushing the limit of a minimum size. There just needs to be some revolution. Maybe in the next 10 years...

 

[BespinReactorShaft]

Originally posted by: LifeStealer
Eventually the disks will spin so fast they go back in time and you find yourself playing more and more escape from monkey island.

Or you realise in horror that the speed and capacity drops to the level of a 360KB 5.25" floppy, your screen goes CGA and FarCry turns into Frogger.

 

[Jeff7]

Solid state is where it's at. It's harder to get platters to spin much faster without having them explode.

SCSI drives spin as high as 15,000 rpm. IDE drives only just graduated to 10,000rpm with the Raptors.
According to this page, a CD at 52x spins at 27,500rpm. Granted, some will shatter at that speed, but CD's aren't made nearly as precisely as hard drive platters. They have tiny irregularities (look at the edge of a disc, it's all wavy) and they have printing. Plus they're larger than hard drive platters. So I'd expect a hard drive platter to be able to withstand maybe 30,000rpm without blowing itself to pieces. Problem there I'd guess would be accurately positioning the head over such a rapidly spinning disc, and maybe vibration.
But still, SCSI drives are 15K right now - so where are the 15K IDE drives?

 

[sharkeeper]

Speaking of CD's exploding...

Putting a stick on label near the outside of the disk and inserting into a modern "ultraspeed" drive is guaranteed to get funny results.

Kind of reminds me of a pair of duckpin balls in a top loading washer. At least the washer did have a safety switch for excessive tub wobble...except ours was broken.

Cheers!

 

[Pariah]

Originally posted by: Jeff7

Solid state is where it's at. It's harder to get platters to spin much faster without having them explode.

SCSI drives spin as high as 15,000 rpm. IDE drives only just graduated to 10,000rpm with the Raptors.
According to this page, a CD at 52x spins at 27,500rpm. Granted, some will shatter at that speed, but CD's aren't made nearly as precisely as hard drive platters. They have tiny irregularities (look at the edge of a disc, it's all wavy) and they have printing. Plus they're larger than hard drive platters. So I'd expect a hard drive platter to be able to withstand maybe 30,000rpm without blowing itself to pieces. Problem there I'd guess would be accurately positioning the head over such a rapidly spinning disc, and maybe vibration.
But still, SCSI drives are 15K right now - so where are the 15K IDE drives?

HD platters can physically spin a whole lot faster but it's not nearly as easy as making an optical disc spin at the same rate. HD platters weigh a lot more than an optical disc and obviously when you add multiple platters the increase is multiplied a few times over. This is one reason among multiple that as the RPM's go up, the size of the platters have gone down. The other part is the extreme data areal density compared to optical discs. The CDROM is about 25 years old now and stores 650MB on a disc significantly larger than a harddrive platter which is squeezing upwards of 50GB on one side of a 3.5" platter. It takes far superior mechanical/electronic designs to read platters spinning at "just" 15k than it does a CDROM at 56x.

The last part, is that we have reached a point of diminishing returns for spindle speed increase. The only aspect of performance that higher spindle speed improves directly is average latency. With a pretty standard 50% increase of today's 15k drives to 22.5k, average latency drops from 2ms to 1.33ms. That's a whole lot of engineering for 2/3's of a ms decrease in average access time which isn't going to impact performance much. Shrinking the platter size and increasing areal density will improve performance more than continuing to increase spindle speed.

There are new SCSI drives on the horizon. Actually, if their press releases were to be believed, they're months late which is par for the course when it comes to hard drive makers. SR has numbers up for the Maxtor Atlas 10K V and it posts some very impressive numbers and set the bar pretty high for what will be coming from the next generation of 15k drives.

 

[BaumerX]

Drive innovation is moving in a different direction. As brought out in previous posts...capacity is no longer an issue. Now HD companies are moving more toward small drives to fit in cell phones, PDAs, automobiles, mp3 and video players.

The bottleneck for HD speed is the ability to read the platters. Increasing the motor's speed is one way but it reduces a drives capacity because it is much harder for the data to be read and it makes drives more expensive.

Seagate is now working to make SATA drives faster by taking advantage of native command queuing technology. The thought is they will soon be able to rival 10,000 RPM hard drives at a slower spin speed so that the platters can still hold a maximum amount of data. There is some test data out there that shows a 7200 rpm drive can outperform a 10,000 rpm drive using NCQ. Platter capacities are also going to increase so that by the end of next year or into 2006 1TB drives should be fairly common with NCQ speeds. What will this mean for smaller drives? Time will tell.