Seagate does not have a News Thread

[Mercutio]

Seagate announced the other day that they're discontinuing 2.5" 7200rpm drives in favor of high capacity 5400rpm models and hybrid drives for performance.
Seagate also announced some 3.5" hybrid drives.

I've had a few hybrid drives and, other than a modest subjective improvement in boot/resume time, they appear to offer no real advantages over traditional 7200rpm models. So there's no doubt at all that this is just a matter of staking out a higher margin niche. Somehow I doubt that 16GB of cache is going to help a 2TB drive all that much, either.

Nothing too exciting or surprising, but hybrid drive still seem to me more like a product in search of a market rather than the other way around.

 

[LunarMist]

8GB flash is crap. Give me 64GB and a way to access it separately from the HD if needed.
I wonder if WD will continue to make 7200RPM mobile drives.

 

[jtr1962]

I agree that hybrid drives are a product in search of a market. I do however see less and less point to 7200 RPM drives in both 2.5" and 3.5" form factors. Real world performance increase over 5400 RPM is marginal at best. Also, at this stage SSDs are cheap enough to use for boot drives, with spinning disks being relegated to bulk storage where speed really isn't all that important. Capacity, reliability, and power consumption are more important for bulk storage, and in all of these categories 5400 RPM (or even slower) trumps 7200 RPM. I suspect one big reason why today's large capacity drives seem to fail a lot more often is precisely because high density and rotation speed are incompatible, yet nobody is willing to just dump 7200 RPM for marketing reasons, at least in the 3.5" form factor. I think here at least Seagate is doing the sensible thing, although sadly only for mobile drives.

 

[LunarMist]

I agree that hybrid drives are a product in search of a market. I do however see less and less point to 7200 RPM drives in both 2.5" and 3.5" form factors. Real world performance increase over 5400 RPM is marginal at best. Also, at this stage SSDs are cheap enough to use for boot drives, with spinning disks being relegated to bulk storage where speed really isn't all that important. Capacity, reliability, and power consumption are more important for bulk storage, and in all of these categories 5400 RPM (or even slower) trumps 7200 RPM. I suspect one big reason why today's large capacity drives seem to fail a lot more often is precisely because high density and rotation speed are incompatible, yet nobody is willing to just dump 7200 RPM for marketing reasons, at least in the 3.5" form factor. I think here at least Seagate is doing the sensible thing, although sadly only for mobile drives.

I agree that 7200 RPM mobile drives are not much faster than the corresponding 5400 RPM drives, but the difference between 7200 and 5400 RPM is more dramatic with 3.5" drives. IIRC Seagate has killed off the low RPM desktop drives.

 

[LunarMist]

Seagate News

 

[jtr1962]

I agree that 7200 RPM mobile drives are not much faster than the corresponding 5400 RPM drives, but the difference between 7200 and 5400 RPM is more dramatic with 3.5" drives. IIRC Seagate has killed off the low RPM desktop drives.

My point is that outside of possibly <500GB models, most people are using 3.5" drives solely for bulk storage where speed isn't necessary. SSDs changed the entire storage paradigm, to the point where anybody interested in a fast drive will just get an SSD, not a mechanical disk with a higher RPM. Seagate killed the low RPM desktop drives probably for marketing reasons, because there are still enough buyers who remember when 5400 RPM = slow and 7200 RPM = fast.

 

[LunarMist]

My point is that outside of possibly <500GB models, most people are using 3.5" drives solely for bulk storage where speed isn't necessary. SSDs changed the entire storage paradigm, to the point where anybody interested in a fast drive will just get an SSD, not a mechanical disk with a higher RPM. Seagate killed the low RPM desktop drives probably for marketing reasons, because there are still enough buyers who remember when 5400 RPM = slow and 7200 RPM = fast.

The low-pricked desktop systems mostly have a single 1-2TB 7200 RPM drive, and they are sold in large volumes. I suspect that >50% of desktop machines still do not have an SSD.

 

[ddrueding]

The low-pricked desktop systems mostly have a single 1-2TB 7200 RPM drive, and they are sold in large volumes. I suspect that >50% of desktop machines still do not have an SSD.

Indeed. Though it is probably closer to 80% without SSDs.

 

[Mercutio]

I agree that 7200 RPM mobile drives are not much faster than the corresponding 5400 RPM drives, but the difference between 7200 and 5400 RPM is more dramatic with 3.5" drives. IIRC Seagate has killed off the low RPM desktop drives.

The biggest difference between 5400 and 7200rpm drive is whether or not we can get the damned things to work properly in disk arrays. :cursin:

 

[LunarMist]

The 1TB is here. The 64MB cache probably helps as much as anything.

 

[Adcadet]

Except Seagate's new4 tb drive is 5900. For mass storage and ZFS arrays, should I go with their 3rb 7200 rpm drives or is the heat/power/noise savings on the 5900 drive really worth it?

how did we get 5400,7200, 10000, and 15000, anyway. And do they actually spin exactly at those speeds?

 

[jtr1962]

I personally feel for reliability reasons the lower the speed the better. A sore point with me is that 5400 RPM drives aren't made any more in higher capacities. I think reliability across product lines is suffering as a result. I also think from a logical standpoint it makes zero sense. Larger drives are almost always used for added bulk storage. That being the case, access time isn't terribly important. Nowadays anyone concerned with performance is going to be using an SSD as their boot drive, not a 7200 RPM mechanical drive. That will only be true even more in the future as price per GB of SSDs continues to drop. I'd really like to see the larger size drives made with as low an RPM as is feasible for them to work. If that turns out to be 3600 RPM, so much the better. Idle power seems to scale with the square of RPM as near as I can tell. A 3600 RPM drive will use only 1/4 of the idle power of a 7200 RPM drive. Typically, 7200 RPM drives idle at 6 to 7 watts, so that's 1.5 watts, give or take, for 3600 RPM. As you lower RPM, you can pack in more bits per inch because they're not going by as fast. I'm not seeing why a 6 GB 3600 RPM drive wouldn't be possible right now. As a result of the higher bit density, sequential read speed won't suffer much, if at all. Latency on average will increase 4.17 ms but the controller should be able to seek to a given track faster since the tracks are going by slower. End result-we might only have access times 1 or 2 ms more than now, although as I said access times really don't matter much considering the way most large drives are used. That's even more true with low end systems now having 4 or 8 GB of RAM. Once you open a program, it'll likely stay in RAM cache until you reboot the system. I'll also guess 3600 RPM drives would cost less per GB to make.

Why isn't this done when it's probably the most logical thing we can do to keep the mechanical disk market alive a little longer? As I said a few posts back, probably for marketing. There are people who are still living in the early 21st century who got used to buying 7200 RPM because it was faster, and now won't touch anything slower. It's short of like people buying cars based on engine horsepower. Someone might go "meh, this car only has 100 HP" Of course, if it weighs 1/3 as much as a 300 HP car, it'll accelerate just as fast.

 

[Chewy509]

how did we get 5400,7200, 10000, and 15000, anyway. And do they actually spin exactly at those speeds?

Back in the ye'olde days, the heads could only determine state changes (between a 1 and a 0) at certain speeds. As processes improved the frequency at which the head could determine a state change, meant they could use faster spindle speeds and thus improve performance without improving bit densities of the platter. So starting from fairly low spindle speeds of appoximately 1-2K rpm when 8" and 5.25" HDDs were in use, by the time we had the ATA interface in IBM XTs and ATs, the spindle speed was IIRC a lowly 4500rpm and top performance SCSI drives where 5400rpm.

As time marched on, we had 7200rpm, 10K and 15K drives all due to improvements in both the head technology, the drive mechanics (the head to platter distance) , platter technology (its per sq in) and the reliability of motors to maintain a set speed within 1% deviation. IIRC, CDs won't go above 52x due to physical limitations of the actual disc (they tend to explode at 60x - the rotations forces at this speed literally rips the disc apart), but with HDDS the companies had a choice of materials to make the physical disk from. Some choose metal alloy, IBM at one time used pure silicon (in the form of glass), and some choose plastics and various polymers.

And IIRC it was either Seagate or IBM/Hitachi were experimenting with 22K drives, both found that mechanically it was very tough to make a metal spinning disc not explode at this speed, the forces of the disc affected the placement of the head and the head technology was not up to reading at these speeds either. In fact they got better performance improvement by not speeding up the disc, but increasing the density of the bits per sq inch. (Modern 5400rpm drives may have higher STRs than older 7200 and 10K drives simply due to the bit density). Because of this change from simply using higher rpms to improve performance, they switched to an alternate method to get better performance, hence why we don't see anything higher than 15K drives. Also note, the higher rpms, will also reduce latency and seek times, as the heads can be placed over the desired area on disk quicker, hence why 10K and 15K drives are favoured for server type roles.

Then SSDs came along, well, mass produced SSDs that the average consumer could afford...

But to answer your question, yes, the actual spindle does spin at those speeds.

 

[jtr1962]

And IIRC it was either Seagate or IBM/Hitachi were experimenting with 22K drives, both found that mechanically it was very tough to make a metal spinning disc not explode at this speed, the forces of the disc affected the placement of the head and the head technology was not up to reading at these speeds either.

At 22K RPM the edge of a 3.5" platter is going 229 mph. Putting aside the enormous forces, it takes a huge amount of power to keep a disc spinning at that speed. Now getting to the forces, at the edge of a 3.5" platter @ 22K RPM, you have a force of 24,000 times the force of gravity. An penny would exert a force equal to the weight of an adult human under those conditions. Not surprising then that metal won't hold up at such speeds. In any case, the reason for spinning hard disks ever faster vanished once SSDs came along.

 

[LunarMist]

Except Seagate's new4 tb drive is 5900. For mass storage and ZFS arrays, should I go with their 3rb 7200 rpm drives or is the heat/power/noise savings on the 5900 drive really worth it?

I'm also interested in this 5900 RPM drive. Perhaps I can rearrange my Hitachi 4TBs to make the slower, cooler Seagate drives backups. But are the Seagoates reliable?

 

[ddrueding]

I remember a story about IBM back in the day having (by far) the fastest drive. Once a competitor got close they just added a second read head on the other side of the platter and held everyone off quite a bit longer.

It would be interesting to have a CD/DVD drive that simply sucked the entire disk to internal RAM cache and served the disk from there. Optical is much faster STR than seek, and could probably dump the entire drive very quickly.

Sorry for the OT. My thoughts echo that of JTRs on this matter. Though I believe Merc will step in soon and mention how none of these drives behave well in arrays.

 

[Chewy509]

At 22K RPM the edge of a 3.5" platter is going 229 mph. Putting aside the enormous forces, it takes a huge amount of power to keep a disc spinning at that speed. Now getting to the forces, at the edge of a 3.5" platter @ 22K RPM, you have a force of 24,000 times the force of gravity. An penny would exert a force equal to the weight of an adult human under those conditions. Not surprising then that metal won't hold up at such speeds. In any case, the reason for spinning hard disks ever faster vanished once SSDs came along.

Thank for doing the math, it certainly brings things into perspective... (The old saying, of flying/spinning rust is certainly true) As a side note, aren't most modern 10K/15K discs only 1.8-2.5" internally?

 

[LunarMist]

I remember a story about IBM back in the day having (by far) the fastest drive. Once a competitor got close they just added a second read head on the other side of the platter and held everyone off quite a bit longer.

Seagate was the last to have parallel heads with their Barracuda 2HP. That was in the era of a servo platter and ~100MB/surface.

 

[Chewy509]

It would be interesting to have a CD/DVD drive that simply sucked the entire disk to internal RAM cache and served the disk from there.

IIRC, some of the more expensive SCSI based CD-stackers/changers had very large RAM caches (512MB), and some even featured internal HDDs for caching as well... (This was before DVDs came out).

 

[LunarMist]

As a side note, aren't most modern 10K/15K discs only 1.8-2.5" internally?

They always used small radius platters that were/are limited by outer track velocity. In 2000 when the first 15K drives appeared, the outer casing was 3.5" for form factor compatibility.

 

[Mercutio]

Though I believe Merc will step in soon and mention how none of these drives behave well in arrays.

Outside of Western Digitalness, nothing about about modern hard drives pisses me off more than than that particular form of deliberate crippling. It was one thing when RAID was a deeply exotic technology only widely embraced in datacenters, but it's something else entirely when a $30 ITX motherboard with two SATA ports still has RAID support.

Anyway, yes, that.

 

[LunarMist]

There is something odd about the 1TB Hybridio drive. The read or write speed trace drops from the beginning to nearly the end as usual, but then jumps again to the max for the last few percent. It's as if the flashmob is at the end of the disc?

I also notice that in a PDSD the drive is clicking rather audibly when it has nothing to do, yet quiet otherwise. I'm not sure how normal that is for this model.

 

[Mercutio]

Other Momentus XT drives I've had have been very clicky.

 

[LunarMist]

This drive is very strange. It is almost pointless.

 

[P5-133XL]

I just pulled a Seagate drive that started clicking constantly on a server. All drives passed their diagnostics (I ID'ed the drive by pulling each drive individually till the clicking stopped) but it was causing constant controller resets and that was drastically slowing down the transfer rate on all six drives on the computer. It took over 48 hours (each drive was flagged by the diagnostic as taking too long) just to run the long-generic diagnostic on all 6 drives in parallel.

It was returned under warranty and the Seagate replacement is an upgraded the model that is much faster. It was a joy getting it replaced compared to most manufacturer warranty returns.

 

[LunarMist]

You were using Monumental drives in a server?

 

[P5-133XL]

No, just a normal 3 1/2" drive that developed a clicking problem.