xsamitt
Distinguished
- Mar 28, 2007
- 268
- 0
- 18,780
Interesting piece of gear.
With todays systems the storage is great...but the cost of the thing as well as the heat they mention kinda bothers me.As if our computers weren't hot enough.
Yes I admit it...I use air cooling.So when I hear words like hotter I tend to shy away.
I guess if you can't take the heat get out of the kitchen. 8)
With todays systems the storage is great...but the cost of the thing as well as the heat they mention kinda bothers me.As if our computers weren't hot enough.
Yes I admit it...I use air cooling.So when I hear words like hotter I tend to shy away.
I guess if you can't take the heat get out of the kitchen. 8)
It's about time, my 500GB drive is getting full (movies, music, work stuff).
I employ a 500GB drive in my work computer and then have a "server" with another 500GB drive that I backup to from my work computer via the network. Then in the server I have a removable 500GB drive that I do another backup to once a month and store off-site in case of theft or fire.
So my point is the cost to gain more storage is more than just one drive. For me to upgrade to 1TB is 3x the cost for no additional storage. And I get the feeling others are in the same boat as me.
We are getting bigger drives but no good way to easily back them up other than purchasing additional mirror drives.
I employ a 500GB drive in my work computer and then have a "server" with another 500GB drive that I backup to from my work computer via the network. Then in the server I have a removable 500GB drive that I do another backup to once a month and store off-site in case of theft or fire.
So my point is the cost to gain more storage is more than just one drive. For me to upgrade to 1TB is 3x the cost for no additional storage. And I get the feeling others are in the same boat as me.
We are getting bigger drives but no good way to easily back them up other than purchasing additional mirror drives.
2 Questions:
1) With the storage controller technology improving since ICH5R would it be a problem to redo the HDD Charts using a system with 2x Woodcrest Xeon or a 2x Opteron revision F? (I believe the associated mobos would have onboard controllers of the latest technology which may have performance advantages [compared to add-on cards/or ICH5R onboard])
2) The article mentions to have raptors as system drive and yet the test system does not. Would anyone mind comment on that?
(Afterall there is an article to switch from FX-60 to C2D in VGA charts, so I wonder if that helps)
1) With the storage controller technology improving since ICH5R would it be a problem to redo the HDD Charts using a system with 2x Woodcrest Xeon or a 2x Opteron revision F? (I believe the associated mobos would have onboard controllers of the latest technology which may have performance advantages [compared to add-on cards/or ICH5R onboard])
2) The article mentions to have raptors as system drive and yet the test system does not. Would anyone mind comment on that?
(Afterall there is an article to switch from FX-60 to C2D in VGA charts, so I wonder if that helps)
drummerdude
Distinguished
- May 9, 2006
- 369
- 0
- 18,780
Anandtech ran a similar article about a month ago, it's worth taking a look:
http://www.anandtech.com/storage/showdoc.aspx?i=2949
They did a number of non-synthetic benchmarks, and it turns out that at the majority of real-world tasks this thing is right up there with the Raptor, at considerably lower $/Gb.
Of course, the Raptor is still at the top. I've been noticing HEFTY price cuts on Raptors, they had 150GB models at Microcenter for 169 bucks. Hopefully WD will roll out something that can compete capacity-wise. As this Hitatchi drive shows very clearly higher data density= higher performance for most applications.
http://www.anandtech.com/storage/showdoc.aspx?i=2949
They did a number of non-synthetic benchmarks, and it turns out that at the majority of real-world tasks this thing is right up there with the Raptor, at considerably lower $/Gb.
Of course, the Raptor is still at the top. I've been noticing HEFTY price cuts on Raptors, they had 150GB models at Microcenter for 169 bucks. Hopefully WD will roll out something that can compete capacity-wise. As this Hitatchi drive shows very clearly higher data density= higher performance for most applications.
I think it leaves the Raptors right where they've been, but with a slightly thinning performance margin. The nice thing is that the new Hitachis will help force the price down further on the Raptors.
I see the Hitachi coming in with slightly lower STR figures in both read and write, and MUCH longer access times.
For anything seek-intensive, unless the Hitachi units have some awfully clever cache/seek optimization, the Raptors should still be the top performer, but the Raptor's advantage should decline with big file I/O.
As a reminder to beware of synthetic benchmarks versus raw benchmarks, I'd like to point out the File Write Performance figures shown with PCMark05 versus the h2bench Write Transfer Rates. In h2bench the T166 has a slightly higher peak write than the 7K1000 but a much more rapid falloff toward the end of the platter than does the 7K1000, yet in PCMark05 the T166 bests the 7K1000 by more than 14MB/S
It's impossible to draw sensible performance conclusions without being very knowledgeable of the inner workings of PCMark05.
That the 7K1000 was the hottest of the drives portrayed is not a huge surprise, but Mr. Schmidt did point out that "Two drives create more heat, which may result in increased cooling requirements." So on a BTU or Watt per GB basis there lays a huge payoff versus the 500GB units.
-Brad
sandmanwn
Distinguished
- Dec 1, 2006
- 915
- 0
- 18,990
Yes I remember reading that article. Thats why I mentioned adding a noise test as this new Hitachi was an extremely quiet drive. This drive would work wonderfully in an HTPC or small studio environment where noise is as big of an issue as storage.
I thought the concluding recommendation was interesting: "1 - Raptor for OS and 1 - Storage drive for data." I was planning 2 - storage drives (7200 rpm - 16mb cache) in Raid 0 for all. I would then back specific data onto an 2nd PC on my home network. Not sure what is the best performance / $ / for 150GB's after reading all of these articles.
choirbass
Distinguished
- Dec 14, 2005
- 1,586
- 0
- 19,780
as far as platter densities for raptors... the only way theyre going to get past the current 74GB barrier, is by implementing a newer technology such as PMR... ...but if they dont increase the density, they could go the other route of increasing the RPMs, at a possible decrease in capacity though, yet again (or they could do both even, a 15k raptor with PMR, so maybe no loss in capacity then)... a 15k raptor would run reallllly hot though, and most likely be noticably louder too, but, will also succeed at putting the raptor noticably faster again... till then though, raptors will most likely stay around where they are now for performance, and eventually be surpassed, if things dont change much
I've been thinking of ways companies can increase performance, and one that I thought about was going to 5 1/4". Remember Maxtor's BigFoot drives way back when? Most cases have plenty of 5" slots not being taken up by anything, so why not produce an enthusiast drive that takes advantage of the additional space to provide better heat managment and noise deadening?
In terms of RPM, I think data density is the preferred route. Higher RPMs run into a whole range of issues, whereas data density can be increased without compromising reliability or performance. Seek times will still be an issue due to rotational latency. Imagine a drive that combined the 10K RPM of a Raptor with the data density of this Hitatchi, or even the less dense .10 models from Seagate. That would be a killer drive.
In terms of RPM, I think data density is the preferred route. Higher RPMs run into a whole range of issues, whereas data density can be increased without compromising reliability or performance. Seek times will still be an issue due to rotational latency. Imagine a drive that combined the 10K RPM of a Raptor with the data density of this Hitatchi, or even the less dense .10 models from Seagate. That would be a killer drive.
choirbass
Distinguished
- Dec 14, 2005
- 1,586
- 0
- 19,780
im sure the reason they wont go with wider patters is because they would need to reduce the rpms more... you would get greater capacity by going with 5 1/4" platters, maybe even 1TB platters, but maybe only 2400RPMs or so as a performance compromise... raptors are limited to their current capacity because a smaller 2.5" platters, and older recording technology
but, 3.5" 7200rpm PMR is the current best balance though for consumers who want large capacity, and something that is acceptably fast, and quiet (and fairly inexpensive even)... but, i definetly agree that PMR should be used for a 10k hdd, as opposed to increasing rpms, and shrinking platter sizes yet again... you would still have 2.5" platters in use, but maybe they would offer up to 200GB or more per platter, which would be really cool, and would put the current raptors performance to shame even, lol
but, 3.5" 7200rpm PMR is the current best balance though for consumers who want large capacity, and something that is acceptably fast, and quiet (and fairly inexpensive even)... but, i definetly agree that PMR should be used for a 10k hdd, as opposed to increasing rpms, and shrinking platter sizes yet again... you would still have 2.5" platters in use, but maybe they would offer up to 200GB or more per platter, which would be really cool, and would put the current raptors performance to shame even, lol
Careful! Lengthy nerd stuff ahead ;-) .
This explanation of superparamagnetism strongly leads into a wrong direction. I ll elaborate and propose something better.
Superparamagnetism can arise in any ferromagnetic material (including ferrimagnetic materials for example). In ferromagnetic materials there is a collective magnetism, where the magnetic moments inside the material interact with each other. In the pure ferromagnetic case this interaction is positive and the magnetic moments want to align each other into the same direction. This leads to domains where moments are all aligned into the same direction. Applying an outer field enforces alignment of the magnetic moments with the field. That is what happens when a sector is written. The problem now is that this sector should hold this magnetization into a certain direction in order to hold the data. Thermal randomization works against this effect, but normally is not strong enough to break the interaction between the magnetic moments or switch the magnetization of the entire sector at once. The smaller you want to make the sector in order to increase the data density the fewer magnetic moments interact inside this sector. Therefore the overall magnetic interaction weakens and at some point the thermal energy will be enough to randomize the magnetic moments and destroy the magnetization. This is called superparamagnetism because the material is now paramagnetic and therefore not permanently magnetic anymore, but the paramagnetism is very strong and therefore super. The problem gets worse when magnetic fields (read/write head) or sectors with a different magnetization start to influence our sector, which is occuring more these days because they need to put closer in order to allow for higher data densities.
In order cut it short and propose something better:
- the superparamagnetic limit is strongly material dependent and can occur on scales significantly greater than 10 nm.
- it does not need the presence of an external magnetic field to occur
- it does not cause magnetically aligned sectors to influence each other, this is done by the interaction of the magnetic moments that is present in the originally ferromagnetic material all the time
- yet external fields and magnetically aligned sectors close by can enhance the problem
Therefore I think it would sound more correct like this:
However, increasing storage densities also increase the risk of the superparamagnetic effect, which makes magnetically aligned elements in the nanometer scale lose a lot of their ability to hold a magnetical alignment. Other magnetically aligned sectors close by or external magnets such as the read/write heads worsen this effect. Superparamagnetism hence can cause magnetically aligned elements to lose that alignment, which would jeopardize data integrity on a hard drive.
It s basically like with the fact that you can take 5 pens into your hands and brake them at once while you can t do that with 50 for example. Making sectors smaller takes more and more pens away. Therefore people search for stronger pens btw., see:
http://www.wit.edu/Academics/em/courses/khabari/ELEC645_files%5CIBM.pdf
(The perpendicular magnetic recording circumnavigates the superparamagnetic effect in a way that you extend the number of magnetic moments that interact into the depth of the material and read parallel to the magnetization.)
This explanation of superparamagnetism strongly leads into a wrong direction. I ll elaborate and propose something better.
Superparamagnetism can arise in any ferromagnetic material (including ferrimagnetic materials for example). In ferromagnetic materials there is a collective magnetism, where the magnetic moments inside the material interact with each other. In the pure ferromagnetic case this interaction is positive and the magnetic moments want to align each other into the same direction. This leads to domains where moments are all aligned into the same direction. Applying an outer field enforces alignment of the magnetic moments with the field. That is what happens when a sector is written. The problem now is that this sector should hold this magnetization into a certain direction in order to hold the data. Thermal randomization works against this effect, but normally is not strong enough to break the interaction between the magnetic moments or switch the magnetization of the entire sector at once. The smaller you want to make the sector in order to increase the data density the fewer magnetic moments interact inside this sector. Therefore the overall magnetic interaction weakens and at some point the thermal energy will be enough to randomize the magnetic moments and destroy the magnetization. This is called superparamagnetism because the material is now paramagnetic and therefore not permanently magnetic anymore, but the paramagnetism is very strong and therefore super. The problem gets worse when magnetic fields (read/write head) or sectors with a different magnetization start to influence our sector, which is occuring more these days because they need to put closer in order to allow for higher data densities.
In order cut it short and propose something better:
- the superparamagnetic limit is strongly material dependent and can occur on scales significantly greater than 10 nm.
- it does not need the presence of an external magnetic field to occur
- it does not cause magnetically aligned sectors to influence each other, this is done by the interaction of the magnetic moments that is present in the originally ferromagnetic material all the time
- yet external fields and magnetically aligned sectors close by can enhance the problem
Therefore I think it would sound more correct like this:
However, increasing storage densities also increase the risk of the superparamagnetic effect, which makes magnetically aligned elements in the nanometer scale lose a lot of their ability to hold a magnetical alignment. Other magnetically aligned sectors close by or external magnets such as the read/write heads worsen this effect. Superparamagnetism hence can cause magnetically aligned elements to lose that alignment, which would jeopardize data integrity on a hard drive.
It s basically like with the fact that you can take 5 pens into your hands and brake them at once while you can t do that with 50 for example. Making sectors smaller takes more and more pens away. Therefore people search for stronger pens btw., see:
http://www.wit.edu/Academics/em/courses/khabari/ELEC645_files%5CIBM.pdf
(The perpendicular magnetic recording circumnavigates the superparamagnetic effect in a way that you extend the number of magnetic moments that interact into the depth of the material and read parallel to the magnetization.)
There is something really wrong with Patrick Schmid. Why hasn't he been fired. Honestly, this is the most fubar crap I've ever seen on THG, including the forumz.
Well, make up your mind you retard, is a kilobyte 1024 or 210 bytes? It can't be both, so pick one. If it's 1024 then you've just assigned it a unitless dimension, congratulations. If it's 210 bytes then it only is in your mind, nobody else has ever heard that before.
Just for those who saw that in the article and were thinking WTF?!?! here's what is actually meant:
HD manufacturers use the same principle as the metric system. Kilo = 1000, thus kilobyte = 1000 bytes.
Now, back in the day the difference caused between 1024 and 1000 was small, because drives were small, say 2GB or less. The difference in "missing size" was negligible and people readily passed it off as "overhead."
Then, as drive sizes increased the differences in actual GB got to be large -- at the 1TB level it looks like you're missing some 70GB, certainly that's not overhead. So, then it became popular to use kilobyte in HD manufacturing to be 1000 bytes, though it long has been. In computing, a kilobyte has for a long time meant 1024 (2^10) bytes, since that's the closest base 2 can get to 1000 with integer powers.
Now, it is common to hear the 1024 byte "kilobyte" referred to as a kibibyte (KiB).
Also, Windows uses the kibibyte definition, thus the GiB in place of GB. Now, you'll often find that (new) versions of Linux will measure it according to HD manufacturers definition. Therefore, using this drive in a Linux system should show 1.0TB capacity.
In the history of computing has 210 never meant anything to anyone at anytime anywhere. This is a fabrication by Patrick Schmid in its entirety, and shame on Achim Roos for not removing Patrick from the project at the mere sight of such a lamentable egregious error.
Why would Patrick Schmid include a comparison as to the pro/cons of a 1TB vs 2x500GB and then NOT include a comparison of the performance?? Is it possibly because the 1TB RAID0 array would have a minimum transfer speed of approximately that of the maximum transfer speed of the 1TB, and at more than 25% off the price?? Now who understands what's up, captain genius.
Honestly, fire this guy and get an crack addicted lounge chair.
Luscious
Distinguished
- Apr 5, 2006
- 525
- 0
- 18,980
I can imagine putting 12 of these drives into one of these:
Norco DS-1220
That should give the HTPC enthusiast a whopping 11.2TB of storage - that's over 2,000 DVD's!!! Or more than 200 full-length 1080p hi-def movies.
Norco DS-1220
That should give the HTPC enthusiast a whopping 11.2TB of storage - that's over 2,000 DVD's!!! Or more than 200 full-length 1080p hi-def movies.
Yep, 12TB in a laptop plus a mobile quad-core cpu = HD Dream Machine.
Synthetickiller
Distinguished
- Aug 25, 2006
- 340
- 0
- 18,780
Considering the price vs performance, 2 of these might be the ticket for decent boot times AND tons of space. If you can sacrifice back up (no raid 5 etc) and use raid 0, you have a decent set up on your hands.
Considering the price of a Raptor (and the noise levels) I'd rather go SAS scsi w/ a 15k drive Seagate Cheetah. It costs more, but is much faster, more stable and probably has a much longer running life. I guess it depends on what you believe the "price of admission" is worth to get into the SAS world.
Add one or two of these terabyte drives and you have yourself a pretty kick ass set up.
Considering the price of a Raptor (and the noise levels) I'd rather go SAS scsi w/ a 15k drive Seagate Cheetah. It costs more, but is much faster, more stable and probably has a much longer running life. I guess it depends on what you believe the "price of admission" is worth to get into the SAS world.
Add one or two of these terabyte drives and you have yourself a pretty kick ass set up.
IF you have an Intel ICH7R or ICH8R soutbridge, then you can compromise a bit on that RAID0 setup you guys are all lusting for:
Get 2x1TB. W00T. Now,
- Make a matrix RAID array:
- 300GB RAID1
- 1.4 TB RAID0
Somehow I think 1.7TB is enough storage. All your pretty much meaningless files can be on the RAID0 - game installs, programs, just general non-critical storage and then you can put the OS, important documents and stuff on the RAID1. 300GB seems fair for keeping important stuff...
Get 2x1TB. W00T. Now,
- Make a matrix RAID array:
- 300GB RAID1
- 1.4 TB RAID0
Somehow I think 1.7TB is enough storage. All your pretty much meaningless files can be on the RAID0 - game installs, programs, just general non-critical storage and then you can put the OS, important documents and stuff on the RAID1. 300GB seems fair for keeping important stuff...
Since you mention noise...
The 15K Cheetah makes LOTS more noise, and in a more annoying spectrum too, than even four Raptors which just hum along quietly. In fact, even the seek noise of four simultaneously seeking Raptors isn't as bad as that Cheetah.
-Brad
xChaoSx
Distinguished
- Mar 5, 2006
- 117
- 0
- 18,680
Since you mention noise...
The 15K Cheetah makes LOTS more noise, and in a more annoying spectrum too, than even four Raptors which just hum along quietly. In fact, even the seek noise of four simultaneously seeking Raptors isn't as bad as that Cheetah.
-Brad
Just put a pillow in your case to reduce noise!
We're at Terabyte drives! Now if they could crank up the RPM to 10,000...
A Norco DS-1220 with those drives... Nice thought. I certainly wouldn't mind having uber storage in a itty bitty living space. Anyone disagree? Not me at least.
But having 5 platters, I think I'd personally wait for 4 or less. As was mentioned, lots of heat. I know, I know... there's always going to be heat and I should use cooling fans, but it would be good to worry less about the heat issues.
3 year warranty? I'd opt for the 5 year if and when it becomes available. It's nice to know you've got a longer safety net to get a replacement if you need it.
Makes you think how it took ~ a decade to go from 1 GB to 1 TB in drive space, when will we hit the next mark? Will it still be in hard drive, or perhaps solid state by then? It's certainly unknown, but curious nonetheless. Sure there's lasers and other technology, but I wish I had a crystal ball to find out. All I've got is my 8 Ball with the floaty triangle thingy. It keeps saying to ask again. :-(
By the way, great article!
A Norco DS-1220 with those drives... Nice thought. I certainly wouldn't mind having uber storage in a itty bitty living space. Anyone disagree? Not me at least.
But having 5 platters, I think I'd personally wait for 4 or less. As was mentioned, lots of heat. I know, I know... there's always going to be heat and I should use cooling fans, but it would be good to worry less about the heat issues.
3 year warranty? I'd opt for the 5 year if and when it becomes available. It's nice to know you've got a longer safety net to get a replacement if you need it.
Makes you think how it took ~ a decade to go from 1 GB to 1 TB in drive space, when will we hit the next mark? Will it still be in hard drive, or perhaps solid state by then? It's certainly unknown, but curious nonetheless. Sure there's lasers and other technology, but I wish I had a crystal ball to find out. All I've got is my 8 Ball with the floaty triangle thingy. It keeps saying to ask again. :-(
By the way, great article!
"Performance users should still go for a WD Raptor drive at 10,000 RPM to host the operating system, and add a second 7,200 RPM for storage"
Depends on your application. In the database world, once the OS is loaded it's rarely needed again and tends to stay in memory as it's relatively small. On a database server, dumping the OS on the smaller, slower device would be a better bet, with the data and logs on the faster device(s). Great bit of kit, especially for a file server, but if you're after maximum data access and transfer speeds in top end enterprise systems, a faster drive would be more appropriate. Sure we'll see 8ms drives at this storage point soon though
Depends on your application. In the database world, once the OS is loaded it's rarely needed again and tends to stay in memory as it's relatively small. On a database server, dumping the OS on the smaller, slower device would be a better bet, with the data and logs on the faster device(s). Great bit of kit, especially for a file server, but if you're after maximum data access and transfer speeds in top end enterprise systems, a faster drive would be more appropriate. Sure we'll see 8ms drives at this storage point soon though
TRENDING THREADS
-
News Windows 11 will reportedly display a watermark if your PC does not support AI requirements- Started by Admin
- Replies: 19
-
-
-
Discussion What's your favourite video game you've been playing?- Started by amdfangirl
- Replies: 3K
Facebook
Twitter
Instagram