Beware of aggressive APM on Windows 10 Build 1809

[R-T-B]

Parking has no correlation to bad sectors

It does have a coorelation to head crashes, which tend to cause bad sectors. There are a limited number of parks that can be performed.

 

[wyxchari]

Parking has no correlation to bad sectors, sectors go bad with age through magnetic weakness, not whether or not the actuator is active or not.

True, but parking the head more than 300,000 times, damages it and causes the data to not be recorded correctly on the disc. The head loses its precision and does not point correctly to the designated sector. The specifications of the manufacturers of hdd clearly indicate which is the maximum parking of the head for which a correct engraving of the data is guaranteed.

 

[Sora]

It does have a coorelation to head crashes,

any movement can see to that, the parking itself has no correlation to head crashes or bad sectors any more than the constant movement and flicking of the actuator, and you certainly cannot put a number on head movements over all which would exceed the parking count by an unfathomable factor.

you'll find that the number there is just the amount of times they have tested it to and not a value that coincides with actuator reliability in real world.

The head can fail well after and well before reaching it.

 

[R-T-B]

any movement can see to that, the parking itself has no correlation to head crashes or bad sectors any more than the constant movement and flicking of the actuator, and you certainly cannot put a number on head movements over all which would exceed the parking count by an unfathomable factor.

You disagreeing with manufacturers that rate a fixed number of head parks on their drive datasheets?

 

[Sora]

You disagreeing with manufacturers that rate a fixed number of head parks on their drive datasheets?

Manufacturer ratings are only the number tested to and averaged against before failure could occur, not failure does occur.

I'm not disagreeing with the testing, im disagreeing with your understanding of MTBF.

If it happened that MTBF actually meant "your thing will fail at this number" there wouldn't be thousands of 24/7 HDD's operating with smart thresholds of 0 for power on time and no read/write or badsectors logged.

Heres a bit of a spanner in the works,

Many of WD's drives within the same plattercount but different sku/spindle speed are manufactered with the same components and then some are tested to 300,000 and others to 600,000 - this is despite the actuator and headcount being the same across sku's where that cycle count changes and only the sticker on the drive changes.

 

[R-T-B]

Manufacturer ratings are only the number tested to and averaged against before failure could occur, not failure does occur.

I'm not disagreeing with the testing, im disagreeing with your understanding of MTBF.

If it happened that MTBF actually meant "your thing will fail at this number" there wouldn't be thousands of 24/7 HDD's operating with smart thresholds of 0 for power on time and no read/write or badsectors logged.

All I'm going to say is a lot of users here had WD green drives commit suicide from excessive head parks.

I know perfectly well what MTBF is.

 

[Sora]

All I'm going to say is a lot of users here had WD green drives commit suicide from excessive head parks.

if you'd care to look, most of those were well below the tested cycle count and actually a result of poor ramping of the actuator from a stopped state rather than the amount of times it had stopped and started, i believe most of them were also from the maylasia facility.

as Wizzard said about it years ago

these drives have a special power saving mechanism that moves the head in the park position if no accesses for 8 seconds. according to wd the mechanism is estimated to be good for 300k cycles. of course this is just an estimate, yours could fail after 1 million or 10 million cyles or just right about now ^^

And here we have 1,300,000 of them

http://forums.whirlpool.net.au/forum-replies.cfm?t=1351539&r=22139577#r22139577

 

[rparmar]

All the troubles depicted in this thread can be mitigated within the Windows power management settings themselves.
[...]
How this setting has escaped everyone's attention is beyond me, but it really is that simple.

Unfortunately, you are wrong. For some drives, this doesn't help at all.

Here we are in November 2021 and there's still no way to get a hard drive to stop clicking every few seconds. It's simply stupid.

 

[Superzuber]

Unfortunately, you are wrong. For some drives, this doesn't help at all.

Here we are in November 2021 and there's still no way to get a hard drive to stop clicking every few seconds. It's simply stupid.

Some enterprise drives have a mechanism that constantly controls sectors on the drive while idle (making them often even more noisy when idle) and workarround is a script to write 0-byte file on the drive in small time intervals.

 

[rparmar]

Otherwise you would know that the drive in question is the Seagate BarraCuda 10 TB (ST10000DM0004-1ZC101), so not an Enterprise drive. I deliberately bought this model to be quieter than the NAS counterparts... oh silly, ignorant me!

I have tried the workaround you suggest, along with many others (registry hacks, hard drive control apps, etc.). All to no avail. Neither does the drive respond to low-level APM calls. As stupid consumers, we are simply stuck with this problem, like it or not.

 

[wupoes]

[regarding AMD machine]

Guys, so can I or can I not (or should I or should I not) set the HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\amdsbs\Settings\CAM EnableHDDParking value to 0, because this thread became madness and I'd like my Barracudas to park less. It's better to spin constantly than to keep moving the head around. I cannot overwrite the drives' APM settings using ordinary tools.

Prime X470 Pro, SATA AHCI, Ryzen 5 2600, Windows 10, Barracuda drives that keep parking heads every 4-10 seconds. All the power management settings set to no idle, active, max performance, never sleep. No result.

---

Alternatively can the Intel people set their relevant registry values to 0 to resolve the issue?

---

In short... Is changing of any settings of the AHCI or RAID drivers okay enough on a Windows 10 machine, or should any other steps be performed to resolve the issue with the excessive head parking? What are the side effects of that? Anyone tested such solutions?

 

[lexluthermiester]

Guys, so can I or can I not (or should I or should I not) set the HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\amdsbs\Settings\CAM EnableHDDParking value to 0

Thank You for that. Been looking for that setting.

BTW, this setting applies to all systems, not just those that are AMD based.

 

[R-T-B]

BTW, this setting applies to all systems, not just those that are AMD based.

Are you sure? It's tweaking a "amdsbs" service...

 

[basco]

@wupoes

did ya try hard disk sentinel ? it worked for me on x99 intel with the sanddisk ssd and a 1tb toshiba hdd

 

[lexluthermiester]

Are you sure? It's tweaking a "amdsbs" service...

Yup. Intel based systems, including my Intel systems, have that registry entry and it is enabled by default. Disabling it will prevent drives from "sleeping" before APM triggers a sleep command.

 

[trparky]

OK, but why would one want to do this?

 

[lexluthermiester]

OK, but why would one want to do this?

To keep your HDDs from sleeping(parking it's heads and spinning down the platters)? That one explains itself..

 

[trparky]

To keep your HDDs from sleeping(parking it's heads and spinning down the platters)? That one explains itself..

But wouldn't making the motor that spins the platter work all the damn time be bad for the overall longevity of the drive?

 

[Regeneration]

It's healthier for the drive to be rather fully on, or fully off. Constant head parking is just killing the drive faster.

Look at those old 80GB HDDs without APM. Most still work even after 20 years, unlike the modern HDDs.

 

[lexluthermiester]

But wouldn't making the motor that spins the platter work all the damn time be bad for the overall longevity of the drive?

No. Maintaining the spinning platters does little wear & tear. Constantly spinning up & down and parking the heads puts tons of wear and tear on a drive motor and the heads.

 

[trparky]

But doesn't that do something?

 

[Regeneration]

View attachment 234204
But doesn't that do something?

No. That's another power saving feature. Windows completely shuts down non-OS drives after X time of idle.

 

[lexluthermiester]

View attachment 234204
But doesn't that do something?

Yes, but it's part of the AHCI power management that runs through the motherboard. The setting referred to above is a drive specific setting. It allows the drive to stay in a hybrid power mode, parking the heads and stopping the platter, but with the main drive controller still powered on. With the setting you are showing, when in effect, Windows puts the drive to "sleep", which still involves parking the heads and stopping the platters, but also flushing the drive cache and powering down the drive controller.

When setting the reg key above, the HDD(s) will not park the heads or spin down in hybrid mode, but will still fully "sleep" when Windows issues the command.

 

[Sora]

BTW, this setting applies to all systems, not just those that are AMD based.

no it doesn't.

Yup. Intel based systems, including my Intel systems, have that registry entry and it is enabled by default. Disabling it will prevent drives from "sleeping" before APM triggers a sleep command.

A driver that is not running on a device because of the absence of supported hardware or different vendor driver being loaded, cannot apply its parameters to the sub devices of an unaffiliated device.

SC.EXE Query amdsbs

SERVICE_NAME: amdsbs
TYPE : 1 KERNEL_DRIVER
STATE : 1 STOPPED
WIN32_EXIT_CODE : 1077 (0x435)
SERVICE_EXIT_CODE : 0 (0x0)
CHECKPOINT : 0x0
WAIT_HINT : 0x0

Parameters set in the registry correspond to functions in the driver, no intel driver supports "EnableHDDParking" so no intel device can benefit from the "EnableHDDParking" parameter

Service names respective to Intel include iaStorA, iaStorAV, iaStorAVC, iaStorV

redundant service iaStor used to respect the following parameters

"Port0AAM"=dword:000000fe
"Port0APM"=dword:000000ff
"Port1APM"=dword:000000ff
"Port1AAM"=dword:000000fe

and iaStorA
"EnableAPM"=dword:00000000

There is no confirmation that this parameter works for current versions of the Intel AHCI/RAID driver (AV, AVC) but it may.

 

[lexluthermiester]

no it doesn't.

Yes, it does.

no it doesn't.



A driver that is not running on a device because of the absence of supported hardware or different vendor driver being loaded, cannot apply its parameters to the sub devices of an unaffiliated device.

SC.EXE Query amdsbs

SERVICE_NAME: amdsbs
TYPE : 1 KERNEL_DRIVER
STATE : 1 STOPPED
WIN32_EXIT_CODE : 1077 (0x435)
SERVICE_EXIT_CODE : 0 (0x0)
CHECKPOINT : 0x0
WAIT_HINT : 0x0

Parameters set in the registry correspond to functions in the driver, no intel driver supports "EnableHDDParking" so no intel device can benefit from the "EnableHDDParking" parameter

Service names respective to Intel include iaStorA, iaStorAV, iaStorAVC, iaStorV

redundant service iaStor used to respect the following parameters

"Port0AAM"=dword:000000fe
"Port0APM"=dword:000000ff
"Port1APM"=dword:000000ff
"Port1AAM"=dword:000000fe

and iaStorA
"EnableAPM"=dword:00000000

There is no confirmation that this parameter works for current versions of the Intel AHCI/RAID driver (AV, AVC) but it may.

That was a very eloquent response and just as equally wrong. There are a number of services and settings that share common platforms. This is due to the cross-licensing between Intel and AMD. This is also why that entry is set to enable and has to be disabled if you want to prevent drives from head-parking and motor spin-down. It works on ALL platforms that run the AMD X64 instruction sets which Intel licenses and uses in ALL of their CPU's and chipsets, which includes drive interfaces.

You were saying?