I'm planning on building a NSLU2 system with an external enclosure and a 0.5TB drive. With noise and power consumption being important factors, spinning down the drive is essential, but I've realised the problems caused by using a USB-SATA interface, which might not be able to instruct the drive to kick back. The NSLU2 wiki states that a couple of the Raidsonic enclosures are capable of translating the sdparm spindown command properly, but none of the enclosures I've been looking at are mentioned.
Browsing these fora I've read about a feature with some drives that use an internal idle timer to spin down when the drive hasn't seen any traffic for a predefined time period (presumably hdparm -S), independently of the operating system. It's not entirely clear to me if this setting is preserved after the drive has lost power - one post claimed it's stored in nvram.
So to the question; I'm looking to buy a WD5000KS or HD501LJ - do these drives support the aforementioned independant spindown feature and the keep_features_over_reset flag?
In that case I'll just hook it up to my Ubuntu box via SATA and perform the required settings before mounting in the USB enclosure.
WD5000KS/HD501LJ internal spindown timer?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
WD5000KS/HD501LJ internal spindown timer?
Last edited by Elias on Wed May 09, 2007 6:23 am, edited 1 time in total.
I believe you can take from this thread that only Hitachi drives support APM:
viewtopic.php?t=39685
viewtopic.php?t=39685
They should remember the value after reset / power cycle. I don't know what flags you may have to tamper with if using hdparm, but HDDScan (running on WinXP) saves changes automatically to non-volatile memory. Both APM and spindown counter can be saved as non-volatile. If using Hitachi FTOOL (for APM) or HDDScan (for regular PM) they are non-volatile as default.
I'm using following drives in USB enclosures (and their corresponding PM & APM settings):
Maxtor MaXLine +II 250GB (PM enabled (don't remember the value), APM not supported)
Maxtor DiamondMax 16 300GB (PM enabled (3 hours?), APM not supported)
Hitachi Deskstar 7K250 (PM disabled, APM enabled (2min unload, 10min lowrpm)
Hitachi Deskstar 7K400 (PM disabled, APM enabled (2min unload, 10min lowrpm)
Western Digital WD3200JB (PM enabled (3 hours), APM not supported)
If I don't remember incorrectly, I have once tried enabling both PM and APM on 7K400 and it worked like it should. 7K250 review here at SPCR: "NOTE: There appeared to be some odd interaction between Windows power management and the Hitachi APM. We'd suggest you use one of the other but not both."
I'm not 100% sure but that "odd interaction" may be like following:
full rpm idle (7200rpm) --(internal APM timer expires)--> unload head (7200rpm) --(internal APM timer expires)--> lowrpm (4500rpm) --(Windows PM expires)--> spinup to full rpm ----> immediate spindown to 0 rpm.
With regular PM, it goes as it should:
full rpm idle (7200rpm) --(internal APM timer expires)--> unload head (7200rpm) --(internal APM timer expires)--> lowrpm (4500rpm) --(internal regular PM timer expires)--> spindown to 0 rpm.
I have made some tests connected to motherboard, to IDE-to-USB bridge, even without any data cable. I have no experience on HDDs in PVRs nor do I have experience with SATA-to-USB bridges, but as long as HDD controller firmware complies with the requirement of allowing 30 seconds time to repond (in order to allow spin-up or error recovery), the controllers shouldn't make the wrong assumption the HDD has "died". Same applies to USB enclosures. The enclosures shouldn't "crash" nor should Windows consider USB mass storage device "dead" even if it takes a while to spin up. (7K400 ~20 seconds, most other HDDs ~10 seconds or less.)
____________
A completely different topic: what HDD to choose for USB enclosure? HD501LJ may vibrate too much to be quiet. I have tried a single-platter Samsung PL40 in OneTouch (which has rubber "decoupling" pieces around screws) and that HDD sound louder than 3-platter Maxtor that originally was mounted in that enclosure. And that Maxtor wasn't among the least vibrating I have (merely average).
I think current low-vibration contestants are WD5000AAKS and T7K500. AAKS also happens to be the only non-Hitachi 3.5" drive to have unload technology but AFAIK it doesn't support APM. APM is technically possible only for drives with unload technology. Conventional contact start/stop drives cannot reduce rpm without causing excessive wear.
I'm using following drives in USB enclosures (and their corresponding PM & APM settings):
Maxtor MaXLine +II 250GB (PM enabled (don't remember the value), APM not supported)
Maxtor DiamondMax 16 300GB (PM enabled (3 hours?), APM not supported)
Hitachi Deskstar 7K250 (PM disabled, APM enabled (2min unload, 10min lowrpm)
Hitachi Deskstar 7K400 (PM disabled, APM enabled (2min unload, 10min lowrpm)
Western Digital WD3200JB (PM enabled (3 hours), APM not supported)
If I don't remember incorrectly, I have once tried enabling both PM and APM on 7K400 and it worked like it should. 7K250 review here at SPCR: "NOTE: There appeared to be some odd interaction between Windows power management and the Hitachi APM. We'd suggest you use one of the other but not both."
I'm not 100% sure but that "odd interaction" may be like following:
full rpm idle (7200rpm) --(internal APM timer expires)--> unload head (7200rpm) --(internal APM timer expires)--> lowrpm (4500rpm) --(Windows PM expires)--> spinup to full rpm ----> immediate spindown to 0 rpm.
With regular PM, it goes as it should:
full rpm idle (7200rpm) --(internal APM timer expires)--> unload head (7200rpm) --(internal APM timer expires)--> lowrpm (4500rpm) --(internal regular PM timer expires)--> spindown to 0 rpm.
I have made some tests connected to motherboard, to IDE-to-USB bridge, even without any data cable. I have no experience on HDDs in PVRs nor do I have experience with SATA-to-USB bridges, but as long as HDD controller firmware complies with the requirement of allowing 30 seconds time to repond (in order to allow spin-up or error recovery), the controllers shouldn't make the wrong assumption the HDD has "died". Same applies to USB enclosures. The enclosures shouldn't "crash" nor should Windows consider USB mass storage device "dead" even if it takes a while to spin up. (7K400 ~20 seconds, most other HDDs ~10 seconds or less.)
____________
A completely different topic: what HDD to choose for USB enclosure? HD501LJ may vibrate too much to be quiet. I have tried a single-platter Samsung PL40 in OneTouch (which has rubber "decoupling" pieces around screws) and that HDD sound louder than 3-platter Maxtor that originally was mounted in that enclosure. And that Maxtor wasn't among the least vibrating I have (merely average).
I think current low-vibration contestants are WD5000AAKS and T7K500. AAKS also happens to be the only non-Hitachi 3.5" drive to have unload technology but AFAIK it doesn't support APM. APM is technically possible only for drives with unload technology. Conventional contact start/stop drives cannot reduce rpm without causing excessive wear.
The tool used for configuring the drive is unimportant as it's just that - a tool for configuring the drive. It's up to the drive to honour this timeout, and it's my understanding that not all drives do this (though I assume all modern drives do, but unless someone can confirm this to hold true, it's a potentially painful €100 assumption).
The same goes for whether the drives hold this setting in non-volatile memory - something I've understood even fewer drives do (according to the hdparm manual), as I've found no source that claims it's a standard feature of modern 3.5" drives.
I obviously cannot rely on the USB-SATA bridge drive in the enclosure to relay any power management messages to the drive, even though in true Linux mannerism there are several ways of issuing such commands. My safest bet is an autonomous drive that knows when to take a break, and does not ever forget..
Regarding spin up times - I believe the timeout is 60 seconds. Both from googling and, well, suffice to say I jumped on the Deathstar bandwagon some years ago..
Anyway, spin up time is not a problem.
The reason for my interest in WD5000KS and HD501LJ is obviously noise. Apparently SPCR is the only site that reports on the matter seriously, and from what they've said about the drives and from testimonials on this forum I gather these are the most silent 0.5TB drives available, with the Samsung being a notch 'better'. Better as in I'd rather combat vibrations (Mechanical decoupling? Rubber, man!) than a turbine like noise produced by the bearings of Hitachi 'Rolls-Royce Trent' Deskstar (Noise insulation? Foam! Thermal insulation? Foam..)
The same goes for whether the drives hold this setting in non-volatile memory - something I've understood even fewer drives do (according to the hdparm manual), as I've found no source that claims it's a standard feature of modern 3.5" drives.
I obviously cannot rely on the USB-SATA bridge drive in the enclosure to relay any power management messages to the drive, even though in true Linux mannerism there are several ways of issuing such commands. My safest bet is an autonomous drive that knows when to take a break, and does not ever forget..
Regarding spin up times - I believe the timeout is 60 seconds. Both from googling and, well, suffice to say I jumped on the Deathstar bandwagon some years ago..
Anyway, spin up time is not a problem.
The reason for my interest in WD5000KS and HD501LJ is obviously noise. Apparently SPCR is the only site that reports on the matter seriously, and from what they've said about the drives and from testimonials on this forum I gather these are the most silent 0.5TB drives available, with the Samsung being a notch 'better'. Better as in I'd rather combat vibrations (Mechanical decoupling? Rubber, man!) than a turbine like noise produced by the bearings of Hitachi 'Rolls-Royce Trent' Deskstar (Noise insulation? Foam! Thermal insulation? Foam..)
"The reason for my interest in WD5000KS and HD501LJ is obviously noise. Apparently SPCR is the only site that reports on the matter seriously, and from what they've said about the drives and from testimonials on this forum I gather these are the most silent 0.5TB drives available"
The most silent 0.5TB drive reviewed by SPCR. Well, technically SPCR has only reviewed 0.4TB Samsung T133 but editors speculate similar noise characteristics for T166, probably due to numerous forum posts.
In addition to WD5000KS there's WD5000AAKS and T7K500.
AAKS has received both positive and negative comments... averaging close to KS. Seeks either more audible or slightly quieter (may depend on AAM). Idle noise character equal or worse. Vibration is lower than with KS.
T7K500's seeks are supposedly quieter than T7K250 and a lot quiter than 7K250 (not to mention 7K400 or 7K500). At least several users and some reviews (iXBT review for example) say so. Idle character and noise level could be tolerable but not as low as Samsung's.
"...with the Samsung being a notch 'better'."
SPCR reviews and recommendations always assume you can decouple the HDD properly.
"Better as in I'd rather combat vibrations (Mechanical decoupling? Rubber, man!)"
Do you call small rubber pieces decoupling? It may reduce transmitted noise at certain frequencies (for example I have noticed OneTouch enclosures with rubber "decoupling" keep whining ball-bearing drives surprizingly quiet) but they certainly cannot eliminate low-frequency vibration. Elastic suspensions have to be very soft. There's no way to implement that inside a USB enclosure.
"...than a turbine like noise produced by the bearings of Hitachi 'Rolls-Royce Trent' Deskstar"
I agree on the similarity of airplane turbine engine noise and noise of older Hitachis during spin-up procedure, but it is only during spin-up that noise resembles a jet engine. After full rpm is achieved it's more like wind noise. I think the "turbine" noise caused by combination of electrical whine that increases in pitch as rpm rises, and turbulent wind noise inside the HDD that increases in volume as rpm rises. When platters are at target angular velocity it gets more quiet... like a jet plane makes less noise when it reaches target altitude, tilts the nose down toward horizon and reduces amount of fuel injected to turbines to prevent velocity from rising too high (which would cause air resistance to make flying uneconomical as little extra speed is gained by burning several times more fuel).
Back from jet engines to Hitachi... anyway, like I already mentioned, I don't think bearings have anything to do with it. I typically categorize noise sources as:
- bearings (grind, coarse whine)
- electrical (sinusoidal whine, high frequency pure tone)
- turbulent (white or whitish noise)
- platter imbalance (humming, low frequency pure tone)
- noise caused by actuator movement (seeks + offline scans)
Bearing construction may affect how "platter imbalance" originated noise transfer from spindle to HDD casing and then to computer case but it isn't created by bearings. FDBs may be "worse" than old noisy BBs in this way. FDBs have a lot less free play than BBs. Free play allows platters to vibrate slightly without HDD vibrating but this is a bad thing if we want high data densities, high performance and high reliability. Vibration has to be countered where it is born and not by allowing any free play in bearings. Counterweighting imbalance is safe: use different sized bolts on different sides of spindle hub cap to compensate platter imbalance.
The most silent 0.5TB drive reviewed by SPCR. Well, technically SPCR has only reviewed 0.4TB Samsung T133 but editors speculate similar noise characteristics for T166, probably due to numerous forum posts.
In addition to WD5000KS there's WD5000AAKS and T7K500.
AAKS has received both positive and negative comments... averaging close to KS. Seeks either more audible or slightly quieter (may depend on AAM). Idle noise character equal or worse. Vibration is lower than with KS.
T7K500's seeks are supposedly quieter than T7K250 and a lot quiter than 7K250 (not to mention 7K400 or 7K500). At least several users and some reviews (iXBT review for example) say so. Idle character and noise level could be tolerable but not as low as Samsung's.
"...with the Samsung being a notch 'better'."
SPCR reviews and recommendations always assume you can decouple the HDD properly.
"Better as in I'd rather combat vibrations (Mechanical decoupling? Rubber, man!)"
Do you call small rubber pieces decoupling? It may reduce transmitted noise at certain frequencies (for example I have noticed OneTouch enclosures with rubber "decoupling" keep whining ball-bearing drives surprizingly quiet) but they certainly cannot eliminate low-frequency vibration. Elastic suspensions have to be very soft. There's no way to implement that inside a USB enclosure.
"...than a turbine like noise produced by the bearings of Hitachi 'Rolls-Royce Trent' Deskstar"
I agree on the similarity of airplane turbine engine noise and noise of older Hitachis during spin-up procedure, but it is only during spin-up that noise resembles a jet engine. After full rpm is achieved it's more like wind noise. I think the "turbine" noise caused by combination of electrical whine that increases in pitch as rpm rises, and turbulent wind noise inside the HDD that increases in volume as rpm rises. When platters are at target angular velocity it gets more quiet... like a jet plane makes less noise when it reaches target altitude, tilts the nose down toward horizon and reduces amount of fuel injected to turbines to prevent velocity from rising too high (which would cause air resistance to make flying uneconomical as little extra speed is gained by burning several times more fuel).
Back from jet engines to Hitachi... anyway, like I already mentioned, I don't think bearings have anything to do with it. I typically categorize noise sources as:
- bearings (grind, coarse whine)
- electrical (sinusoidal whine, high frequency pure tone)
- turbulent (white or whitish noise)
- platter imbalance (humming, low frequency pure tone)
- noise caused by actuator movement (seeks + offline scans)
Bearing construction may affect how "platter imbalance" originated noise transfer from spindle to HDD casing and then to computer case but it isn't created by bearings. FDBs may be "worse" than old noisy BBs in this way. FDBs have a lot less free play than BBs. Free play allows platters to vibrate slightly without HDD vibrating but this is a bad thing if we want high data densities, high performance and high reliability. Vibration has to be countered where it is born and not by allowing any free play in bearings. Counterweighting imbalance is safe: use different sized bolts on different sides of spindle hub cap to compensate platter imbalance.