WD GPs are dying? Really?

Silencing hard drives, optical drives and other storage devices

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Vicotnik
*Lifetime Patron*
Posts: 1822
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

WD GPs are dying? Really?

Post by Vicotnik » Wed Apr 29, 2009 2:30 am

Lots of people seem to think so. Is the title of the sticky thread to blame? Or is there another source of this rumour? It's getting a bit frustrating and I would like to see some real data on this, if it exists.
Main: ASRock B85M-ITX | i3-4330 | 16GB DDR3 | Intel 730 240GB | HDPLEX H1-S | picoPSU | No moving parts | Idle 13.9W
HTPC: ASRock J4105-ITX | 4GB DDR4 | Intel 535 120GB | picoPSU | No moving parts
Gaming: Intel DH77EB | i5-3570K | GTX 1060 6GB | 16GB DDR3 | TJ08-E | RM750X
Server: ASRock N3150-ITX | ~30TB | G-360 | Idle ~25W

nutball
*Lifetime Patron*
Posts: 1304
Joined: Thu Apr 10, 2003 7:16 am
Location: en.gb.uk

Post by nutball » Wed Apr 29, 2009 10:36 am

I'd like to see some real data too.

The "what you can run on a 300W PSU" sticky here is a credit to the Internet. The WD GP sticky is just the opposite.

PartEleven
Friend of SPCR
Posts: 279
Joined: Sun May 06, 2007 10:37 am

Post by PartEleven » Wed Apr 29, 2009 11:52 am

Seems like it's all speculation based on SMART recommended limits. Unfortunately any rumor about hard drive unreliability gets converted into "fact" on the Internet.

Ch0z3n
Posts: 400
Joined: Thu Oct 02, 2008 5:48 am
Location: Orlando, FL

Post by Ch0z3n » Wed Apr 29, 2009 2:42 pm

Or, just don't use SpeedFan on a computer with a GP drive, or tell it not to poll the SMART info on it?
[size=75]Main Rig: [color=blue]Athlon II X4 620 @ 3ghz, w/ Ninja Mini, MSI R4830-T2D512, GA-MA78GM-S2HP, 4x2GB DDR2 800, WD5000AACS, 120mm S-Flex 1,200rpm & 800rpm, TV Wonder 650, Signature 650, NSK2400[/color]
Laptop: [color=green]Asus UL80VT-A1[/color]
[/size]

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Thu Apr 30, 2009 4:32 am

"WARNING: WD Green Power drives may kill themselves" may be the name of that thread but it surely should rather be just "WD unload count issue and how to reduce unload cycles" than the name it carries. Current name is just too FUD.

Also the general panic in that thread is pretty ridiculous. Some claim there's high failure rate of few months old WDs, and the same claim the unload count issue is a hazard, yet not many unload cycles can accumulate in just few months, leading to contradictory statements. Unload count issue and it's potential reliability concerns become relevant only on a longer time period. The drive may exceed minimum sustained unload count in less than a year, but that minimum cycles. It should sustain more. Will failures increase noticeably after 3 years? 5 years? By this point, we'll probably start to see other old-age failures not related to unload cycle count. (Though, with low power consumption, temperature accelerated degradation of internal parts and evaporation of spindle lubricant may be slower than with 7200rpm drives.)

Honestly, we can't tell if there will be a problem and any failures we've seen so far have been early deaths not related to this potential future issue. There's several people who have exceeded the minimum sustainable cycle count without encountering a single hiccup.

While high number of unload counts is a potential reliability concern and a certified noise issue (as they are more noticeable than regular seeks), GPs may still be one of the most reliable drives out there. Samsungs have been average at best. They've never exceled in reliability nor been as horrible as Deathstar or Maxtor DM+9. Seagate 7200.11 had some horrible firmware corruption (bricking) issues. Hitachi ain't as quiet as Samsungs (even the high-rpm models)... but then again, neither are those Seagates or 7200rpm WD Blacks.
[size=75]Antec 1200 | HX520W | Commando | Q6600 G0 @ 3.15GHz | Noctua NH-U12F | 8GB of RAM | HD 4670 (passive)
7 TB of storage: 1x 1st gen GreenPower (1TB), 1x 2nd gen GreenPower (1TB), 1x 3rd gen GreenPower (2TB), 1x 7200rpm F1, 2x 5400rpm F2 EcoGreen[/size]

PartEleven
Friend of SPCR
Posts: 279
Joined: Sun May 06, 2007 10:37 am

Post by PartEleven » Thu Apr 30, 2009 7:21 am

Exactly. The problem is still what I said before though. Any rumor about hard drive unreliablility becomes irrefutable fact on the Internet. I don't even bother convincing people otherwise anymore. Well at least if those people get driven away from buying WD GP drives, maybe WD will keep offering them at low, low prices for the US market :P . Might as well benefit from them.

Ralf Hutter
SPCR Reviewer
Posts: 8636
Joined: Sat Nov 23, 2002 6:33 am
Location: Sunny SoCal

Post by Ralf Hutter » Thu Apr 30, 2009 12:43 pm

I changed the title of that thread, and un-stickied it.

I never did much like it. Basically nothing but FUD, from the get-go.
Main Box: Intel i3-3225, Intel DH77EB, 16GB Corsair RAM, 256GB Samsung 830, SS360GP PSU, CM PS07 case.
Music Server: Intel DH77EB + i3-3220, 2xSamsung 2TB F4, Pico PSU, Fractal Define Mini, 2xScythe Fans @250 rpm.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Thu May 07, 2009 11:50 am

Ralf Hutter wrote:I changed the title of that thread, and un-stickied it.

I never did much like it. Basically nothing but FUD, from the get-go.
You need to put linux in the title as the issue, if any, doesn't exist in windows.
.
Please put a country in your profile if you haven't already.
This site is international but I'll assume you are in the US if you don't tell me otherwise.
RAID levels thread http://www.silentpcreview.com/forums/viewtopic.php?p=388987

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Thu May 07, 2009 9:43 pm

dhanson865 wrote:You need to put linux in the title as the issue, if any, doesn't exist in windows.
I agree with that. However I probably disagree with some of the other posts in this threads. Although I have used many WD drives and find them to be very reliable in general, the behavior of my Green Drive with Linux is quite different in terms of racking up unload cycles very quickly (and I would emphasize "very"). I applied the patch to eliminate the unloads, because I did not want to find out the hard way whether or not the drive would eventually fail. The issue is not the cost of the drive, but the amount of time it would take me to rebuild the system and cost of loosing any data that could occur (even with daily backups I could loose 8 hours of work).

Unfortunately, WD has the attitude that Linux is a very small part of the market, and they don't really care if their drives work probably with that OS. For me, the WD Green Drive made sense on my Linux server that I use for application development, because their is little synchronous disk I/O (most stuff is in memory when needed) and performance is not a high priority (and saving energy is a laudable objective IMO). As a previously loyal WD customer (last 8 drives purchased have been WD) I do feel I have been betrayed.

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Fri May 08, 2009 1:15 am

"You need to put linux in the title as the issue, if any, doesn't exist in windows."

It does exist in Windows to a lesser extent. To a very small extent, if you aren't
- running constant very low-I/O fileserving 24/7
- running SpeedFan without proper settings (i.e temperature monitoring is not disabled for GreenPower HDDs).

SpeedFan causes 1 unload cycle per minute. Linux background processes cause 3.

Instead of calling the drive not working with Linux, WD betraying you, etc., why don't you bitch about why Linux is so f'ed that it writes stuff to HDD on 20 second intervals even when it just sits there idle. Why don't you direct your bitching to *nix developers?

Why should you? Because unload technology can reduce power consumption and increase reliability (when used properly). To solve this issue with Linux from the HDD side, the only way is to effectively disable unloads, either by making hard disable or giving unload counter 21+ second interval, or by making the interval adaptive. All result in no unloads and increased power consumption.

The ONLY good solution is to fix your Linux. You can make a small file and do random write operation to it once in 19 seconds and... well, again achieve state of non-power saving but get around the unload cycle increment issue. But the best solution would be to un-break Linux itself.

Linux is NOT designed for ecological operation of the system. The programmers don't apparently care or they didn't care, as basic building blocks of these OSes predate eco-thinking and GreenPowers. Sure, you can argue that OS predates GreenPower and couldn't have prepared for such an issue. Yeah. But the same thing exists with every laptop drive.

So apparently the programmers didn't intend Linux to run on laptop or didn't care about unload count issue with the drive and also about reduced battery life resulted from non-working power saving.

Also, HDDs have had STAND-BY capability for ages. With 20 second IO intervals, you can't spin down a HDD. It's not only that Linux developers didn't just ignore all laptop, and predated GreenPower, they ignored all desktop power saving as well.

They - did - not - care.

So STFU about "betrayal" and all that shit, unless you want to direct (at least some of) it towards your loved *nix developers. Or is your *nix just too perfect to criticize?
[size=75]Antec 1200 | HX520W | Commando | Q6600 G0 @ 3.15GHz | Noctua NH-U12F | 8GB of RAM | HD 4670 (passive)
7 TB of storage: 1x 1st gen GreenPower (1TB), 1x 2nd gen GreenPower (1TB), 1x 3rd gen GreenPower (2TB), 1x 7200rpm F1, 2x 5400rpm F2 EcoGreen[/size]

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Fri May 08, 2009 10:24 pm

whiic wrote:So STFU about "betrayal" and all that shit, unless you want to direct (at least some of) it towards your loved *nix developers. Or is your *nix just too perfect to criticize?
I don't claim to be an OS expert, but I think you have got it backward. It is Windows that writes so often that the WD Green Drive does not unload very often. One reason it does this is for the indexing service, but there may be other reasons. Linux writes to the drive less frequently (when not actively performing a specific user request), which means that WD Green drive thinks it is OK to unload, this causing the problem.

I do not use Linux on my desktop or on a laptop. I use it as a server for software development, and Windows is not an option (others at my company make the architectural decisions). But I do know that Windows servers are simply not robust enough to run the type of enterprise applications we develop. Our Linux servers have up to 96 CPU cores with 128 GB of memory to process large database applications.

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Sat May 09, 2009 4:01 am

"I don't claim to be an OS expert, but I think you have got it backward. It is Windows that writes so often that the WD Green Drive does not unload very often. One reason it does this is for the indexing service, but there may be other reasons."

Indexing can be disabled. I use it only on my OS HDD as OS HDD will be thrashed by pagefile and installed programs so adding indexing is not going to worsen it. I don't use GP as OS drive.

For storage drive, drives that are accessed by user or third-party application, the drives may be totally untouched for hours and hours, enough to even justify use of STAND-BY (a much more drastic power-saving than just unload). This also applies to NTFS filesystem even though along with it comes filetable mirroring, compressable folders, etc., that add to extra access needs (compared to old FAT32).

Basically if you have a IO problem with Windows, it's either
- OS drive itself or
- caused by unnecessary settings
- caused by third-party application.

With Linux, this unload problem is pretty systematic from what I've read of it. I don't use Linux myself (I have tried some distros but never ended using them).

"Linux writes to the drive less frequently (when not actively performing a specific user request), which means that WD Green drive thinks it is OK to unload, this causing the problem."

Yeah, a very short IO every 20 seconds is very light IO, yet also VERY unnecessary and potentially harmful for long term durability of GreenPower and all laptop HDDs regardless of maker (unless that HDD has unload feature disabled at factory or by user).

Windows does such a thing only for OS drive (I dunno if it's a problem only with Linux OS drive as well) and while it can be a lot heavier (if indexing is enabled) it's not systematic. From power saving perspective, it's better to thrash the shit out of HDD for short period of time, and allow it to idle hours consecutively than write single byte every 20 seconds ad infinitum.

What is so important you need to write to HDD every 20 seconds? Computer sits idle. You're not downloading anything, you're not doing anything. Your OS isn't indexing to speed up future searches. Nothing is done. Why access it constantly? What sort of brainfart is that?
[size=75]Antec 1200 | HX520W | Commando | Q6600 G0 @ 3.15GHz | Noctua NH-U12F | 8GB of RAM | HD 4670 (passive)
7 TB of storage: 1x 1st gen GreenPower (1TB), 1x 2nd gen GreenPower (1TB), 1x 3rd gen GreenPower (2TB), 1x 7200rpm F1, 2x 5400rpm F2 EcoGreen[/size]

josh-j
Posts: 11
Joined: Sat May 17, 2008 9:13 am
Location: UK

Post by josh-j » Sat May 09, 2009 4:27 am

whiic wrote:What is so important you need to write to HDD every 20 seconds? Computer sits idle. You're not downloading anything, you're not doing anything. Your OS isn't indexing to speed up future searches. Nothing is done. Why access it constantly? What sort of brainfart is that?
I believe this more to do with the way UNIX was developed and its original uses - ie. server reliability, and so on - than a "brainfart".

Unix was not developed more for server environments than for home uses. The behavior is not difficult to change really, its more down to distribution-makers making it easy for users to do.

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sat May 09, 2009 8:17 am

The bottom line is that Windows writes to the drive more than Linux even when both are idle. It is possible to disable indexing, but very few people do that (and besides, it is not just indexing that causes the frequent disk writes in Windows in my experience), and that is why the WD Green Drives do not cause a problem with Windows. Again, the whole analysis of this situation is upside down and backwards. It is Linux that writes to the hard drive less often than Windows.

So WD doesn't care about Linux support of their Green Drives, since apparently the drives behave in an unexpected manner by unloading very frequently (my unload numbers after a few months were in the hundreds of thousands). They basically have written off Linux because it is such a small part of their market. That is their prerogative. They can do what they want.

But IMO they should have stated that the Green Drive was not suitable for Linux up front if it will not behave properly with that OS so that I would not have purchased 3 of them for Linux servers. That part is inexcusable to me (along with their frequent denial and stonewalling about the issue). There is a point where the behavior of a company toward it customers becomes so egregious, that I will not buy any of their products, even the ones that don't cause problems for my intended application.

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Sat May 09, 2009 9:43 am

"The bottom line is that Windows writes to the drive more than Linux even when both are idle. It is possible to disable indexing, but very few people do that (and besides, it is not just indexing that causes the frequent disk writes in Windows in my experience), and that is why the WD Green Drives do not cause a problem with Windows."

I don't think it's the sheer amount of background tasks that make it less of an issue with Windows. It's true that when Windows starts indexing, there is no 8 second periods of pause during the process. It thrashes and when done, may be idle for quite some time.

As for other processes... I think that's just BS. I took a peek with FileMon.exe to what happens on my OS drive and the only program that makes READs and WRITEs on filesystem is PeerGuardian, a third-party application. Didn't know it thrashes that much. Maybe I should disable logging since I don't even read the logs. No Windows process is doing IO on my OS drive while idle.

For some reason, Explorer.exe does OPEN e:\, QUERY INFORMATION E:\, CLOSE E:\ all the time, but it doesn't cause any IO as information is cached in memory as it reads the very same data over again. (Drive in question is a GreenPower and I can hear it's not actually reading from it.)

"So WD doesn't care about Linux support of their Green Drives"

Why make HDDs for them dipshits who don't care about environment (= Linux users)? Sorry for the flaming but you're really asking for it. You claim Linux was not intended for desktop/laptop and only for server. So be it. Why should desktop and laptop HDDs be optimized for server OS then? It doesn't make any sense.

You may argue that Linux has surpassed it's original purpose and is not (EDIT: "now", not "not") desktop mature. So... why is it not fixed yet? Linux is still as BROKEN as it was originally, it's just Linux users who forcible (EDIT: "forcibly") want to make it a mainstream.

"But IMO they should have stated that the Green Drive was not suitable for Linux up front"

Two points:
1) Yeah, blame WD. But why blame only WD... laptop drives from all manufacturers have similar, very aggressive power saving schemes. Have you read about IBM/Hitachi, Seagate, Samsung, Toshiba, etc. admitting that their laptop HDDs cannot be used with Linux? I believe not. At least I have not.

2) It's not HDD manufacturer's responsibility to get familiar with how different OSes work and what kind of filesystem accesses they make. How could they get familiar with all existing OSes? There's so ridiculously many OSes, especially on the *nix front.

Also, to "fix" the HDD to suit to poorly working Linux is to break the power saving scheme of the HDD. It's natural Linux should be fixed instead of breaking any funtionality to suit Linux's needs.

You may argue all you want about Linux/Unix server roots but those roots don't justify anything. We aren't in the 90's anymore. We must advance. Hardware has advanced. OSes have to keep up. I know it's difficult for public license (EDIT: "licence" (noun), not "license" (verb)) OSes and software. This is why free codec packs for example cannot utilize GPU's 2D acceleration and instead do it all with CPU. If only FFDSHOW could use DXVA...

"That part is inexcusable to me (along with their frequent denial and stonewalling about the issue)."

So, go ask Seagate to make a public announcement that their laptop drives are incompatible with Linux. Watch them deny the issue (via email, not publicly as they wouldn't bother a public statement). Do the same to Samsung. Watch the same happen. Hitachi and Toshiba. No-one will admit a problem. Why should only WD admit such an issue? Peer pressure? Yeah, that's the only reason. It's not about the drive itself and not about any announcement left unaccounced by the manufacturer, it's about what gets nerd masses raging.

And what does rage nerds? Well, SMART attributes the nerds don't understand but which they still monitor and try to interpret. Take for example GreenPower. It would not have become an "issue" had people not started worrying about the SMART attribute. When 2nd generation GP (00D6B0) hid it's SMART data (while unloads continued to occur) vast majority of (STUPID!) nerd thought it was fixed!

Same has happened before. Not with unloads. Remember raging people over Seagate's RAW Error Rate and Hardware ECC Recovered? With typical raw data values being in millions of errors and value & worst around 50...70 range (instead of 90-255 range for normally working drives made by other manufacturers) it caused quite a bit of FUD. Even when everyone posted screencaps of values around 60 and not a single person reported a near 100 value, almost every (STUPID!) nerd thought their drive was faulty.

"There is a point where the behavior of a company toward it customers becomes so egregious, that I will not buy any of their products, even the ones that don't cause problems for my intended application."

Then, don't buy any HDD at all. I'm fairly confident they all have had PR issues before.

Same applies to motherboards, CD drives, CPUs, GPUs, prebuilt computers. They all get arrogant sometimes when dealing with pricks. Because that is what we custumers usually become when we feel betrayed (regardless whether we are or if we just feel betrayed because we were too stupid to do our homework).

EDIT: and some missing pluralities, etc. Not going to fix them as they wouldn't cause any severe misunderstanding of what I meant to say.
Last edited by whiic on Sat May 09, 2009 2:36 pm, edited 1 time in total.
[size=75]Antec 1200 | HX520W | Commando | Q6600 G0 @ 3.15GHz | Noctua NH-U12F | 8GB of RAM | HD 4670 (passive)
7 TB of storage: 1x 1st gen GreenPower (1TB), 1x 2nd gen GreenPower (1TB), 1x 3rd gen GreenPower (2TB), 1x 7200rpm F1, 2x 5400rpm F2 EcoGreen[/size]

josh-j
Posts: 11
Joined: Sat May 17, 2008 9:13 am
Location: UK

Post by josh-j » Sat May 09, 2009 11:42 am

whiic wrote:Why make HDDs for them dipshits who don't care about environment (= Linux users)?
Wow.

Anyway, sorry if I annoyed you. I actually agree with you that for desktop use its not necessary to have frequent disk writes. I was just pointing out the reason for it :)

halcyon
Patron of SPCR
Posts: 1115
Joined: Wed Mar 26, 2003 3:52 am
Location: EU

Post by halcyon » Sat May 09, 2009 1:02 pm

m0002a wrote:The bottom line is that Windows writes to the drive more than Linux even when both are idle. It is possible to disable indexing, but very few people do that (and besides, it is not just indexing that causes the frequent disk writes in Windows in my experience), and that is why the WD Green Drives do not cause a problem with Windows. Again, the whole analysis of this situation is upside down and backwards. It is Linux that writes to the hard drive less often than Windows.
I have Win XP Pro 4 GB memory, swap on non-GP drive.

All indexing is disabled (both windows original and Windows desktop searches). I also have background defrag disabled.

My OS hits disks very infrequently and only when I give it reason to.

I have 2*1TB orig (1st batch) GP drives, on minimum 8hrs / day, 7days a week.

No problems so far. SMART drive fitness data shows very high health for all monitored disk variables.

Maybe I'm lucky.

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Sat May 09, 2009 3:25 pm

josh-j: "Anyway, sorry if I annoyed you."

You didn't annoy me. It was just the whole unload topic, and people who contribute on spreading FUD without giving their suspicions a proper thought. You know, the people who say: "This is the worst product ever. I will never buy from them again! The betrayed me! They abused me! They are murderers! They even eat panda bears!" You just got into crossfire.

josh-j: "I was just pointing out the reason for it"

It's a reason, but not a justification. I consider Linux "broken" for regular desktop use. It needs to be fixed to meet modern standards - not fixing modern standards to fit back to the outdated world of *nix computing.

Yeah, and I do acknowledge *nix is still very usable in server environment, when you don't need such things as CPU power management, HDD power management, hardware video acceleration, etc. It's like running Win95...

You may argue Win95 is not safe and it's true in a way. It's not very protected. The main reason for any OS to become dangerous is it's popularity. If *nix was mainstream, there would be more people creating viruses, troyans, etc. to *nix environment. Going back from Vista or XP to Windows 3.11 today would actually be pretty safe. No-one attempts to hack to Win3.11 because of it's low popularity. Troyans infecting and spreading from Win3.11 wouldn't find another similar system, so it's propagation would be severely hindered. Safer than *nix. Definitely safer than Mac. Mac users are quite ofter f'ed by viruses, because they know their OS is (a bit) safer than Windows... and as a result rarely actually use firewalls and antivirus software where as in Windows, any sane person uses them.

So *nix's "security" is in a way result of only OS itself but it's unpopularity.

Very sidetracked. I'll shrink it to small font.


halcyon: "..... No problems so far. SMART drive fitness data shows very high health for all monitored disk variables.

Maybe I'm lucky."


Well, from what I've read on these threads, many people are afraid of GPs dying but very few have actually had any problems with them. So it appears GreenPowers are capable of taking quite many unload cycles without substantial wear despite all the FUD circulating on the net.
[size=75]Antec 1200 | HX520W | Commando | Q6600 G0 @ 3.15GHz | Noctua NH-U12F | 8GB of RAM | HD 4670 (passive)
7 TB of storage: 1x 1st gen GreenPower (1TB), 1x 2nd gen GreenPower (1TB), 1x 3rd gen GreenPower (2TB), 1x 7200rpm F1, 2x 5400rpm F2 EcoGreen[/size]

ugrakarma
Posts: 23
Joined: Sun Dec 16, 2007 3:43 pm

Post by ugrakarma » Sat May 09, 2009 5:03 pm

Claiming that linux has no hardware video acceleration is factually incorrect. When ranting it helps to get the facts straight.

http://en.wikipedia.org/wiki/VDPAU

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sat May 09, 2009 6:26 pm

whiic wrote:Well, from what I've read on these threads, many people are afraid of GPs dying but very few have actually had any problems with them. So it appears GreenPowers are capable of taking quite many unload cycles without substantial wear despite all the FUD circulating on the net.
According to the WD specification sheet, GP drives are rated for 300,000 load/unload cycles.
http://www.wdc.com/en/library/sata/2879-701229.pdf
That does not mean the drive will destruct when it meets that number, but I assume that WD engineers have some idea of what they are talking about when they designed the drive and listed the specs.

My WD Green Drive accumulated several hundred thousand load/unload cycles in a matter of a few months, idling 99% most of the time (running Linux as an application development server). So after a few years, it would accumulate 8-10 times the rated spec.

So I don't think the concern is just FUD, it is a legitmate concern for those using it with Linux.

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sat May 09, 2009 6:31 pm

halcyon wrote:
m0002a wrote:The bottom line is that Windows writes to the drive more than Linux even when both are idle. It is possible to disable indexing, but very few people do that (and besides, it is not just indexing that causes the frequent disk writes in Windows in my experience), and that is why the WD Green Drives do not cause a problem with Windows. Again, the whole analysis of this situation is upside down and backwards. It is Linux that writes to the hard drive less often than Windows.
I have Win XP Pro 4 GB memory, swap on non-GP drive.

All indexing is disabled (both windows original and Windows desktop searches). I also have background defrag disabled.

My OS hits disks very infrequently and only when I give it reason to.

I have 2*1TB orig (1st batch) GP drives, on minimum 8hrs / day, 7days a week.

No problems so far. SMART drive fitness data shows very high health for all monitored disk variables.

Maybe I'm lucky.
I don't know if you are saying that your WD GP drive has not had a problem or it is not accumulating large number of load/unload cycles?

In either case, I would bet that you drive is getting hit much more often than you think (thus preventing unloads), even with those things disabled. If you look at the Task Manager (Processes), you can see the number of I/O's that are occurring (you may have to go to "View--Select Columns and check all the I/O counts).

Cryoburner
Posts: 160
Joined: Sat Dec 01, 2007 4:25 am

Post by Cryoburner » Sat May 09, 2009 7:05 pm

While I think whiic went a little overboard, I do agree that it shouldn't be up to a hard drive manufacturer to let potential customers know if a power-saving feature that's been common in other drives for years might not be fully compatible with an OS that's used on roughly 1% of desktops. While it would be great for them to go out of their way to release a utility to disable the feature for every applicable drive, it shouldn't be expected, and should be up to the desktop Linux developers to actually fix the problem at its source. If someone wants to disable the feature on their own, there's nothing stopping them from having a utility poll the drive more frequently to accomplish the same task.

In any case, I'm glad to see that thread finally renamed to something more reasonable. There's no need to incite internet hysteria and potentially hurt a company's reputation for something that only affects a relatively small percentage of systems and may not even be a significant reliability issue. I'd be more concerned about their inability to change model numbers when releasing new revisions of drives. : )

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sat May 09, 2009 7:32 pm

Cryoburner wrote:While I think whiic went a little overboard, I do agree that it shouldn't be up to a hard drive manufacturer to let potential customers know if a power-saving feature that's been common in other drives for years might not be fully compatible with an OS that's used on roughly 1% of desktops. While it would be great for them to go out of their way to release a utility to disable the feature for every applicable drive, it shouldn't be expected, and should be up to the desktop Linux developers to actually fix the problem at its source. If someone wants to disable the feature on their own, there's nothing stopping them from having a utility poll the drive more frequently to accomplish the same task.

In any case, I'm glad to see that thread finally renamed to something more reasonable. There's no need to incite internet hysteria and potentially hurt a company's reputation for something that only affects a relatively small percentage of systems and may not even be a significant reliability issue. I'd be more concerned about their inability to change model numbers when releasing new revisions of drives. : )
I don't know where people keep getting the idea that Linux is for desktops or that the WD GP driver are only for desktop machines. It can be, but I don't use Linux for that. I use Linux for an application development server. The vast majority of application and database servers in the business world are Linux or UNIX (I am not including file servers and print servers).

So yes, I would expect WD to disclose if there are problems using GP drives with Linux. My guess is that they did not know at first, but instead of fessing up and finding a solution, they are in complete denial. Having personally purchased 3 GP drives, I am not amused.

Cryoburner
Posts: 160
Joined: Sat Dec 01, 2007 4:25 am

Post by Cryoburner » Sun May 10, 2009 12:54 am

m0002a wrote: So yes, I would expect WD to disclose if there are problems using GP drives with Linux. My guess is that they did not know at first, but instead of fessing up and finding a solution, they are in complete denial. Having personally purchased 3 GP drives, I am not amused.
They did release a statement about the load/unload issue, so they aren't denying it. They also suggested the obvious solution of fixing your OS and software settings to not poll the drive at a rate that causes it to constantly go in and out of power saving mode. They even go as far as providing a utility for their enterprise class drives to increase the timer, effectively disabling the feature.

I never claimed that Linux was meant for desktops. It's just that if you're using a GP drive in a server environment, you should be using the enterprise class version, for which they provided a 'fix'. Even without that, they explain that the way to properly enable the drive's power saving mode is to adjust your settings on the software end of things. They can't make your computer stop accessing the drive at a certain rate, which is the actual root of the problem, if one even exists.

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sun May 10, 2009 11:38 am

Cryoburner wrote:They did release a statement about the load/unload issue, so they aren't denying it. They also suggested the obvious solution of fixing your OS and software settings to not poll the drive at a rate that causes it to constantly go in and out of power saving mode. They even go as far as providing a utility for their enterprise class drives to increase the timer, effectively disabling the feature.
I stand corrected on the fact that WD has now made the utility available to the public for enterprise class drives, but not sure what one is supposed to do if they have a consumer WD GP drive. For a long time before that they refused to release the utility, and usually claimed it did not exist (even though a few people on this forum have had for quite some time). I know about that because I spoke to WD tech support on multiple occasions, and exchanged at least 10 emails with them, and the whole time the denied any knowledge of the utility.

But let's review the current WD claims that the problem is with Linux. Here is their claim:

"Some utilities, operating systems, and applications, such as some implementations of Linux, for example, are not optimized for low power storage devices and can cause our drives to wake up at a higher rate than normal. This effectively negates the power-saving advantages of low-power drives, such as Western Digital’s RE2GP, and artificially increases the number of load-unload cycles"

1. The fact that WD has acquiesced and released the utility tacitly acknowledges that significantly exceeding the rated load/unload cycle rating for the drive (300,000) most likely is a problem that impacts the WD GP reliability. So all those who are questioning that on this forum are now in conflict with what WD is implying.

2. WD suggests that problem is that Linux is not optimized for power-saving because it causes the GP drives and artificially increases the number of load-unload cycles. This is backwards. Windows avoids the load/unload cycles by constantly writing to drive every few seconds, which means that the drive never (or rarely) unloads during normal use. Linux (on the other hand) writes to drives less frequently (even when not used by an external application), which still causes the drive to unload, but the frequency is just outside the maximum WD unload setting so the drive is constantly going from load to unload and back to load again. Linux is no more inefficient in this regard than Windows, and the claim that it lacks power-saving features is bunk.

3. If Linux were fixed to “not poll the drive at a rate that causes it to constantly go in and out of power saving modeâ€

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Sun May 10, 2009 12:40 pm

m0002a: "So yes, I would expect WD to disclose if there are problems using GP drives with Linux."

Like I already said, the same "issue" is present in all laptop HDDs (minus a few low-volume laptop drives intended specifically for Blade servers, that have the unload feature disabled at factory). The "issue" is relevant to ALL manufacturers' laptop HDDs. Why are you not raging against them?

Let me answer that on your behalf: because you bought 3 GreenPowers for Linux. You didn't buy laptop drives. If you had bought laptop drives, you'd possibly be raging against the manufacturer on why [pick any manufacturer] hasn't made a public announcement about their drive being "faulty" (= your opinion).

Likewise, quite many go into Holy Rage mode, swearing, cursing and spreading FUD if they lose a HDD with data. If they lose a HDD without data, they seldom rage. You are raging against WD because you didn't do your homework. People who lose data on HDD crash rage because they were too stupid to make back-ups. People take their hatred toward themselves against someone/something else because they can't stand the fact that they themselves made an error somewhere (whether that was OS decision, drive decision, decision of skipping regular back-ups, decision of using a HDD from same batch to back-up another, or whatever).

m0002a: "1. The fact that WD has acquiesced and released the utility tacitly acknowledges that significantly exceeding the rated load/unload cycle rating for the drive (300,000) most likely is a problem that impacts the WD GP reliability. So all those who are questioning that on this forum are now in conflict with what WD is implying."

In my opinion, reason for it's release may be different: they want to calm down the nerd rage... people like you. Apparently it's not working in all occasions.

They don't release the software => you rage about them denying the issue.
They release the software => you rage about them admitting the issue.

I think they just want you to shut the f' up. Honestly. And they are giving you carrots instead of beating you with a stick to get you quiet, yet you keep on raging even though they give you software to configure the power-saving feature. LOL.

m0002a: "2. WD suggests that problem is that Linux is not optimized for power-saving because it causes the GP drives and artificially increases the number of load-unload cycles."

But that is just so true.

m0002a: "This is backwards. Windows avoids the load/unload cycles by constantly writing to drive every few seconds, which means that the drive never (or rarely) unloads during normal use."

That's just total, utter, complete bullshit. At least with XP, there's no constant background processes accessing HDD when computer is idle. When you browse through folders, it reads. When you open files, it reads. When you let it sit idle, it may still continue accessing HDD for a minute or two, then it suddenly stops all HDD access. I have verified this with Filemon.exe on my desktop and by listening to my XP powered laptop. I can clearly hear the HDD unload and not reload until I actually use the laptop again.

Surely, I've configured my XPs on my computers to be light on HDD and on graphical bling and everything else that would consume memory, CPU cycles or HDD I/O, but I have not gone into lengths of registry hacking to do it. Just minor tweaks using XP's own user interface and some more minor tweaks using Powertoys for Windows XP.

m0002a: "3. If Linux were fixed to “not poll the drive at a rate that causes it to constantly go in and out of power saving modeâ€
[size=75]Antec 1200 | HX520W | Commando | Q6600 G0 @ 3.15GHz | Noctua NH-U12F | 8GB of RAM | HD 4670 (passive)
7 TB of storage: 1x 1st gen GreenPower (1TB), 1x 2nd gen GreenPower (1TB), 1x 3rd gen GreenPower (2TB), 1x 7200rpm F1, 2x 5400rpm F2 EcoGreen[/size]

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sun May 10, 2009 1:04 pm

It appears to me that whiic is the one who " go[es] into Holy Rage mode, swearing, cursing" (and also abusive) instead of discussing this in a rational manner. Not to mention using half-truths to make the weaker argument appear the better.

I am not going to address each one of the claims in detail, because I have already done so above. But to summarize, Windows does not accumulate high numbers of load/unloads because it writes to the drives very frequently, even when "idle." I don't know what you have done to your Windows installation, but this is true for 99.9% of all Windows installations in the world on desktop machines.

Laptops are a different subject, partly because laptop drives are much more robust (but slower and smaller) than desktop drives. The whole system will power down much when not in use more often than a desktop, making the whole question of disk drive load/unload a moot point.

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Sun May 10, 2009 2:22 pm

m0002a: "Not to mention using half-truths to make the weaker argument appear the better."

Half-truths are still better than complete BS like "Windows does not accumulate high numbers of load/unloads because it writes to the drives very frequently, even when "idle.""

Get that Filemon logging already. You may claim you've addressed it already in detail but I see no links to logs. In return, I can also upload logs, if you can't take my word of what I've concluded browsing through them. My disbelief toward your claim is not about whether I trust your honesty - it's because I don't think you have seen your own logs to back up your idea... you just ASSUME your disk access is due to OS itself. Even if you don't want to post logs (even partially censored if something delicate such ass access to C:\temp\kiddyporn.avi shows up in them) at least make the logs for YOURSELF to evaluate before you claim it's the OS itself that causes I/O. You are jumping to conclusion.

Of course, Vista may well be worse than XP, and Filemon might not work in Vista. I dunno. I have and I will avoid Vista like plague. XP is the best Windows so far... well, not including Windows 7 Beta. I saved myself the hazzle of trying it out since it would have been a limited time-period OS.

m0002a: "Laptops are a different subject, partly because laptop drives are much more robust (but slower and smaller) than desktop drives."

That was not the point. Point was that unload time-out is similar 8-second(ish) on both and if laptop drive does unload, GreenPower would have unloaded as well.

If you meant unload count tolerance... well, laptop HDD durability over desktop HDD durability might ... or might not apply to unload count tolerance as well. More massive head actuator arm assembly of course causes greater force to the metallic tip that functions as a landing gear to plastic unload ramps. But then again, bigger dimensions of this metallic tip and unload ramp size also make it more tolerant to these forces. Honestly we can't be certain GreenPower is less durable in this aspect than laptop HDDs.

Laptop HDDs are rated to higher unload count. Rather conservative unload count rating of GP is not proof of GP's vulnerability, it may just mean that there's not enough statistical data in WD disposal. The drive will survive AT LEAST hundreds of thousands, but most likely millions of cycles. The development team has not had the chance of performing tests with statistically relevant sample size over millions of cycles.

m0002a: "The whole system will power down much when not in use more often than a desktop, making the whole question of disk drive load/unload a moot point."

Only if you set it to power down on idle. I have made no such configuration. I bought the laptop without OS (to avoid being forced to use Vista, or pay for Vista and overwrite it anyway).

All I meant is that I got my laptop HDD to unload and stay unloaded (without computer entering stand-by, hibernation or powerdown) so using same settings in desktop XP should yield same result. XP causes no unnecessary idle activity.
[size=75]Antec 1200 | HX520W | Commando | Q6600 G0 @ 3.15GHz | Noctua NH-U12F | 8GB of RAM | HD 4670 (passive)
7 TB of storage: 1x 1st gen GreenPower (1TB), 1x 2nd gen GreenPower (1TB), 1x 3rd gen GreenPower (2TB), 1x 7200rpm F1, 2x 5400rpm F2 EcoGreen[/size]

whiic
Posts: 575
Joined: Wed Sep 06, 2006 11:48 pm
Location: Finland

Post by whiic » Sun May 10, 2009 3:09 pm

And about Task Manager IO reading (read, read bytes, write, write bytes, other, other bytes), they appear to be about read/write IO in general. So it includes not only HDD IO but also RAM IO.

I too my laptop on my lap while writing this stuff on my desktop. I'm making these observations for laptops Task Manager because it only has one HDD and I can clearly hear the noise of unload/reload (which are easily distunguishable from regular seeks)

lsass.exe causes ~200 bytes/sec of both read and write IO. Constantly. VLC causes ~300 bytes of "other" (not read, not write) IO while being PAUSED.

RMClockHLT.exe causes 20 kilobytes/sec of "other". (It's a utility to run HLT command on very low priority. Thus, if no other process uses CPU, HLT will be run. This saves more electricity than just regular idling. (RMClockHLT.exe is part of RMClock CPU monitorin and management suite.)

After I opened Task Manager, there was some HDD IO, HDD unloaded, after about half a minute, it reloaded, did something, unloaded, a few minutes later reloaded, did several seeks and unloaded again. No reloads in last 10 minutes (and counting)... but IO read, writes and "others" keep rising all the time.

To me, this is a definite conclusion that
- Task Manager does not make any difference between RAM IO and HDD IO.
- Win XP indeed does let HDD idle without small interrupts. And when it (rarely) interrupts the idling, it typically does a batch of many small tasks and enters idle again.

So your claim of XP behaving in similar manner as Linux (periodic access) but only in smaller intervals as to not let HDD unload is BUSTED.

So, if you want to prove something, get that Filemon running, Task Manager doesn't prove feces.
[size=75]Antec 1200 | HX520W | Commando | Q6600 G0 @ 3.15GHz | Noctua NH-U12F | 8GB of RAM | HD 4670 (passive)
7 TB of storage: 1x 1st gen GreenPower (1TB), 1x 2nd gen GreenPower (1TB), 1x 3rd gen GreenPower (2TB), 1x 7200rpm F1, 2x 5400rpm F2 EcoGreen[/size]

m0002a
Posts: 2831
Joined: Wed Feb 04, 2004 2:12 am
Location: USA

Post by m0002a » Sun May 10, 2009 3:57 pm

whiic wrote:So your claim of XP behaving in similar manner as Linux (periodic access) but only in smaller intervals as to not let HDD unload is BUSTED.

So, if you want to prove something, get that Filemon running, Task Manager doesn't prove feces.
You are correct about filemon. I have run filemon many times before (I had forgotten about it) and got the results I mentioned, i.e, that Windows does not allow the disk to unload/load with WD Green Drive, because Windows frequently writes to hard disk, even when user applications are idle. That is why the WD Green Drive is not a problem with Windows and it does not rack up a lot of load/unload cycles.

You may have disabled all those tasks, but 99.9% of Windows machines are still running them. If you have a WD Green Drive, what is your load/unload cycle count?

If WD had set the unload timer to 30 seconds or so, I don't think we would even be having this discussion because it would not be a problem (but neither Windows nor Linux would unload unless in sleep mode).

Post Reply