Slashdot Mirror
IBM 120GXP Revisited
Andrew sent us a link to an article about the
IBM 120gxp controversy.
This is about the fact that the drive has been declared unfit for server use,
and to back that up, IBM says you should only use it for 333 hours a month.
This is a good summary of the issues and worth a read.
I think part of the trouble here is that IBM is selling these drives as OEM parts, and not retail drives. OEMs generally don't sell systems to the enthusiast market (The group most likely to leave systems running all day). In your average Dell/Compaq/Cow computer, it's preset to go on standby after an hour or so, powering off the drive. Since all of us "power users" don't like those performance-detracting ACPI/APM functions, we always disable them.
Furthermore, the DeskStar isn't intended to be a server part - IBM makes the UltraStar for that.
So, in essence, it's buyer-beware with OEM parts. Just like with the ATI video card debacle - You're buying parts that aren't intended for *you* to use. It's your fault if you're tryin' to skimp a couple of bucks out of IBM/ATI/whoever by buying on the grey market.
Now, that said...it's pretty fscking ridiculous to be making these drives and all but marketing them as the fastest ATA drives on the planet. That's practically hyping it up to the enthusiast market right there. And I really think it's asinine to expect these drives to *only* be run 8 hours a day. Factoring in the average lunch break when the computer will most likely get left on, that means that the drives are generally running out of spec on a regular business day in your average workstation.
"Mod, mod, mod...and another troll bites the dust."
Beware of the Leopard.
-- @rjamestaylor on Ello
If they aren't good for the server market, the 120GXP aren't good for anything- since what regular home user ever needs that much space?
Oh, and BTW, the article also mentions problems with the 75X and 40X drives.
Conclusion- Somebody at IBM QA has screwed up- vote with your $$, folks, and make IBM take notice of this problem- we should not have to replace a HD after only 1 year (or less!) of use!
I have a light-duty server that has been running two 60GXP drives for the past 6 months with no trouble. After I heard about the problems with the 75GXP I switched from striping to mirroring in my raid configuration just to be safe.
When I heard about the 120GXP I figured that IBM was releasing the modified (glass plattered) version of the Deskstar drive in order to clear up any perception that the line had problems (due to the issues with the 75GXP). I decided to buy one to put in the new Athlon XP box I was building at the time. I've been using it for 2 months with no trouble (so far), but since I purchased it from a retailer I found on pricewatch, I doubt that I could follow the article's suggestion and return it. The performance benchmarks I've done suggest that the drive performs relatively well (135% of the 7200 RPM ATA100 reference drive in SiSoft Sandra's HD benchmark).
I probably won't buy another IBM drive for a while, however, based on the unresponsiveness of IBM to the problems as reported in the article.
Amazing magic tricks
Seriously, though, online polls are completely meaningless. They are swayed by rigging and mass canvasing, and that's ignoring the basic tenent that only the motivated (or bored) bother voting in the first place.
Having said that, I find this whole debate intriguing. Firstly there is the fact that the 75GXP was a very big seller (performance and value packed into one), so the industry standard failure rates indicate that with normal failure rates there will still be more people with failed drives. Anecdotally I can say that myself, and several other people I know, have had zero problems with our 75GXP, but following standard Slashdot-esque thought processes I should extrapolate that out and say that therefore no one has every had a problem, and therefore the drive is perfect. I have heard stories about people who had to "replace it X times!", but in almost all cases you'll find that they grossly inproperly installed the drive with no venting space on both sides (and this is a case with drives from any manufacturer. I had a Maxtor die and opened the case to find that the OEM had sandwiched it between two other drives).
I saw an interview with one of the plaintiffs against IBM, and I'd swear I saw them subtly shift gears from saying that the 75GXP had a higher failure rate (I would guess that that they can't find numbers to back that up, and no numbers determined by a Slashdot polling are not sufficient to convince anyone but the converted), to saying that instead this is a lawsuit expressing outrage about any failures, and it is really a bellwether against all hard drive makers. Uh huh. Now there's this article that is basically thrown off by standard marketing and reliability metrics: The drive IS made for desktop use, and desktop use is normally about 8 hours a day of infrequent use, versus 100% usage 24 hours a day for some server drives. Perhaps they simply realize that the latter will naturally have a higher failure rate so they built that into the server drive prices, but they don't guarantee that for desktops? The article makes the contention that it is a usage heat issue, but that seems a bit silly as the drive will reach maximum temperature minutes after going to 100% usage (i.e. It's not still creeping upwards after 8 hours).
I'm finding this story rather spooky as my 75GXP failed on me (ironically in the middle of my first real backup) yesterday after 18 months of "average" use (less than 6 hours a day). I use the IBM drive as my main system drive, and keep a 5GB for data backups "now and again".
I bought the drive way back in October 2000 with the confidence that "it's an IBM drive: these things are not only fast, but are meant to be reliable". When I started to see the horror stories of other peoples' drives failing I felt quite lucky (read: smug) that mine was still going strong. Now I realise how stupid I look: that whiz, whir, crunch, grind noise that I heard yesterday from the disk *above* the sound of my rather noisy fans scared the life out of me.
I'm now faced with the nasty task of not only attempting to salvage what data I can before I send it to IBM (yes, as other people have mentioned, at least IBM provide a fairly decent 3 year warranty), but also shudder in anticipation at what IBM decide to do with my drive. I believe there are 3 options:
1. Attempt to "fix" my drive and send it back (although having the "Drive Fitness Test" return "Defective Disk" should quash this option). That means I'm stuck with the same drive which is most likely to fail on me again sometime in the (not-too-distant) future.
2. Replace my drive with the same spec, from the same product line & production factory. Again, this worries me as I'm probably ending up with a new disk which has the same defects and thus is also going to die on me.
3. Replace my drive with a newer product with an equivalent spec. Yesterday I was hoping for this option. However, having read the ViaHardware article, this doesn't hold much hope for me either.
And that's the crux: it's alright having the drive under warranty & returning it, but who's to say that any drive they replace it with is not as faulty?
This whole fiasco with the GXP line has certainly put me off IBM drives, no matter how fast and "great" they may be. Shame.
from the article:
:)
While large numbers of readers responded to the questions I posed regarding drive reliability, their emails present very different pictures. Some of you swear by IBM drives and their reliability, while listing many of the Seagate, Maxtor, or WD drives you've seen fail in both a corporate and a consumer setting, while other readers had horror stories of seeing IBM drive after IBM drive bite the dust.
On the general topic of hard drive reliability, I've noticed a similar trend -- every sysadmin to whom I speak seems to have a poorly-founded personal hatred for one hard drive manufacturer. Sure, I admit, having a hard drive fail on you really sucks (esp. if you've been lazy with backups and don't have RAID).
What's weird about this is that people who are otherwise rational will take a single experience with a bad drive and use it to justify an opinion that all drives from that manufacturer are unreliable. It reminds me of D&D players who will, after rolling a d20 four or five times, decide that it "rolls high."
Here's the deal: hard drives fail. Get over it and design your systems such that your important data isn't relying on a single hard drive. In fact, two of my hard drives (a Quantum and an IBM) are slowly failing on me right now. Before that, the last one was a Seagate.
Now, I will admit that there must be some models from some manufacturers which are more prone to failure, just as there are probably some d20s which are prone to "roll high." Perhaps some manufacturers tend to make more reliable drives than others. However, in all the times I have heard someone bitch about a hard drive manufacturer, not once has someone referred to a study that did a statistically sound comparison of drives (I'm not sure that one even exists that compares, over time, all the various models of the manufacturers). It's always "Seagate sucks! A Seagate drive failed on me once, and I had to do a bare-metal recovery."
Of course, in this case, lots of people have reported problems with this drive, so it's a little different. If, sometime in the near future, someone tells me not to buy a cheap-ass OEM IBM IDE drive to use in a critical server, saying "remember the 120GXP?", I'll probably listen to them. However, based on my limited anecdotal evidence, I doubt that will happen
</rant>
Bad idea: Letting your drive melt.
I had problems with my 40GB Quantum AS (bad sectors, spinning down and refusing to speak to anything); I noticed it was getting rather hot (bare in mind this is in a well ventilated case at the bottom of the 3.5" mounting bays with plenty of space above it) and wondered if this had anything to do with it.
So I mounted a card cooler I had spare, put it on top of a small speaker just outside the case (I leave the side open) and had it blow over the system.
The result? CPU temp dropped a good 6c (43c fully loaded for a 1GHz Athlon clocked to 1.2GHz, compared with ~50 before) and now both my drives (Quantum Fireball 20GB and Quantum Fireball AS 40GB) are cool to the touch. There's been absolutely no sign of any problems with the drive since either.
How will people use their 1,000 free hours of AOL in 45 days if their hard drive, and hence their PC, can only be on 333 hours per month?
Tip: Check if the drive temperature is not exceeding the specified limits with something like this tool. Many case designs do not provide sufficient ventilation for 7'200 RPM drives, especially if they're mounted closely together. Use a HD cooler if the disks get too warm, it's still less expensive than reconstructing data and/or reinstalling.
Irony: big ad from IBM on my page of the "IBM 120GXP Revisited" article, saying "Time to update your critical systems security!". Yes, indeed 8-)
"There are already a million monkeys on a million typewriters, and Usenet is NOTHING like Shakespeare." - Blair Houghton
Late last year when it was Quantum Fireball drives that were dropping like flies in our office I got on the phone with a drive repair center in Canada. We were going through the "How much will it cost us to actually retrieve the data from these drives?" discussion and I thought to ask him what brand of drives he sees the *least*.
"SCSI" was his response. "Oh sure," he said "there are fewer of them out there -- but we hardly see any at all."
One of the other interesting things he told me is that the drives included in Quantum's Snap Server appliances, despite being IDE, are *NOT* drives you can buy off the shelf. And he hasn't gotten in a single Snap Server drive in the two years they had been selling them.
I switched from Quantum to IBM drives at the time (ugh!) but had the forsight to put them all in a RAID-1 configuration. We've sent three DeathStars back for repair so far. The good news? IBM had replacements to us in under a week.
I recently switched from IBMs to Maxtor (making sure I wasn't buying Quantum's old stock) and have already had one of their 80Gb drives fail. For the record they are not as responsive as IBM in the RMA department.
So what's the answer folks? You get what you pay for. If you care about your data buy an Adaptec 1200A RAID-1 controller and two drives, or spend the money on a SCSI controller and SCSI drives. So far I haven't found any IDE drive vendors that can sell you a reliable drive (I have dead fujitsu drives around here as well, but must admit that I still haven't tried Western Digital.)
I am not interested in articles about life extension advancements.
Yes, we all know IBM's IDE drives are shitty now,
OK, maybe you were just trolling, but with your low UID I will give you a chance.
What drive manufacturer do you think is good then? From random reviews I was starting to think that IBM was one of the better ones. I have purchases several Western Digital ones and have had too many of them fail (for me that is). Therefore, I have been looking for a better HD manufactuer. Who makes the most reliable HDs today?
IBM is giving a classic demonstration in how NOT to handle a problem. It has been demonstrated time and time again that the tactic of denying everything does not work. It doesn't work for politicians, it doesn't work for accounting firms, and it doesn't work for electronics manufacturers.
IBM could have come right out, admitted to a defect and paid the price for that. By admitting to it, and making sure to replace all of those defects, they would have bought themselves a huge amount of credibility. We'd all buy IBM drives knowing full well that if there was ever a design problem we'd hear about it and get it fixed.
Now, IBM is risking the reputation of their entire drive line through these shenanigans. Before IBM stood as one of the best drive manufacturers, but repeated issues with the GXP line are quickly submarining that. In the ultra-competitive hard drive market, this sort of problem could put that unit completely out of business.
I personally owned an IBM 75GXP, and it is the only hard drive I've ever owned that had a problem. I've been using hard drives since a 20MB box attached to my Atari 1040ST, and not a single one of them ever made a fuss. My defective drive has since been replaced, but it's of course with another IBM drive and now I continue to be concerned that maybe this drive will be defective too.
I wonder how long before people learn the lesson that covering your tracks, especially in this era of rapid distribution of information, is a bad policy.
This sig has been temporarily disconnected or is no longer in service
I have not really trusted IBM drives since my DGHS 18 U died. Not because it died, but because IBM Customer Service handled it extremely poorly. Not only was the drive purchased from an authorized IBM agent I had full documentation. They had initially said I needed documentation to replace it, but when I obtained documentation they said the warranty was only a year. The paperwork I had clearly showed otherwise, but they sternly refused. Since, I have acumulated about 14 dead IBM drives in the 10-30 gb range...
Anyway, I think we're all misisng something here. I've seen IBM drives installed in a Raid config die within hours of eachother, just days or weeks out of warranty.
I think the thinking at IBM drives is along this line "Lets manufacture the drive in such a way we can undercut our competition, but as a result, it will make the drive only last this many hours.." The failure rate could be related to the fatigue rate of metal of a certain purity used in the drive, stability of ceramics used, how good the air filter is inside, etc etc. From my experience seeing each class of drives die, The MTBF is amazingly similar between drives that die.
Lets say the warranty on these is 3 year. Isn't that IBM saying that the drive has a lifetime of 11,998 hours, or just about 499.5 days? If I'm right, even if you follow IBM's reccomendation, the drive will die, but more likely to be out of warranty. Will they replace the drive if I don't follow the reccomendation? I would like my drives to last 5 or 10 years.. or until I don't need it anymore. Period. Not a year.. or three years or whatever the warranty du jour is.
The oldest drives I have and am using are Seagate FH 5.25" 9 gb scsi drives. They're 10 years old. Their MTBF is clearly published, and about 800,000 hours, if memory serves.... this is far more acceptable.
Wendell
Ick...it's a wonder they're still in business. While the 130MB Maxtor in my parents' ten-year-old PS/1 still works, I have yet to see a newer Maxtor last a year. An 80GB drive in one of my work machines held out for 9 months before it started making weird head noises. A 20GB drive purchased a couple of years ago and a string of three 5.1GB drives purchased four years ago all crapped out after 1-3 months.
By comparison, I haven't had an IBM go bad on me. I've had a 45GB 75GXP at home for a little over a year and a couple of 60GB 60GXPs at work for the past few months. I just added a couple of 60GB 120GXPs to a machine at home (the same machine with the 75GXP) in RAID-0 to speed up video editing.
I suspect that most of the problems people see with IBM drives are brought on by inadequate cooling (stacking several drives with little or no separation and no forced-air cooling), crappy power supplies, or overclocking. The drives in the home machine have an 80mm fan in front of them to force air through the stack and are powered by a 330W Enermax. The work machines have only one drive each, installed in the lowest drive bay. (The power supplies are whatever was in the case...if it helps, they're AMD-certified for the 1.4-GHz Athlon XPs that they power.)
20 January 2017: the End of an Error.
As to HD prejudice: I agree it exists and it can go for or against any mfgr -- I think it's mainly a matter of what HDs one had that 1) first failed, and 2) next didn't fail. And if you don't use tons of drives, that's usually a matter of dumb luck. It could just as easily have gone the other way.
As to W.D. -- like everyone else, they've had bad production runs. The 1.6gb and 6.4gb come to mind, tho not all of those were bad either. Don't know for sure about the 3gb model, the only one I've seen died at the ripe old age of 5yrs by simply losing everything at once (but considering the system it was in, it may have had a bad power shock, and it resurrected well enough to use as a junk drive). However -- on the norm W.D. are reliable, and their warranty procedure is hassle-free (and they send you a NEW replacement drive, NOT a repair or refurb). As a result, I use W.D. in all my machines and my clients' machines. I've got about a dozen in use here right now, some over 7 years old.
As to Maxtor, they have a generally poor reputation around here, and if a local clone dealer uses Maxtor HDs, it's a sure sign they're cutting corners all around. YMMV.
As to IBM, the first time this topic came up, a guy who worked in their HD mfg'ing process posted a comment that explained why the recent IBM HDs are unreliable. Anyone got the link offhand?
~REZ~ #43301. Who'd fake being me anyway?
Only 8 hours a day? Next thing you know the drives will be demanding 15 minute coffee breaks twice a day, full medical and dental, three weeks vacation each year, job security, and so on.
-me
Love many, trust a few, do harm to none.
333 hours per month is only 13 days per month. 45% working time. Working 12-7 or maybe 24-3.5
I used to be in the clone mfg business. One thing we determined is that companies go through cycles. For instance, WD has at various times been among the best AND the worst of the manufacturers out there. Seagate has made some wonderful drives and some absolute crap.
Here's a resource I've been watching lately. If anyone has similar things (published reports of reliability from places that deal with dead drives) please follow up to this message.
http://www.driveservice.com/bestwrst.htm
Overclocking is causing the drives to fail? I'm sorry, but that seems pretty far out there. I've seen drives fail to respond when PCI timings were WAY out there but fail? Nah, I don't think that's a very likely cause. The drives have got to have their own internal clocks (okay, I know they do - I see the oscillators on the PCB) and should have checks inside to make sure that commands don't send the heads off into lala land. If tey don't have the latter then would you really want to use that drive?
;-)
Heat and power fluctuations sound like much more likely external issues. Either that or there's simply an internal flaw that didn't show up in testing. Wouldn't be the first time that's occured now would it? I still recall the grease problem seagate had years ago where heads would get mired in the stuff. A quick "twist start" would usually free them up but if you shut them down and allowed them to cool it would stick again. I replaced DOZENS of those damned htings doing field service. when I hit up a Seagate rep at a show about it he officially denied the problem - and then proceeded to tell me off-record just how bad it was. I didn't buy a Seagate drive for awhile afterwards
Build it, Drive it, Improve it! Hybridz.org
I guess that pixie dust IBM uses isn't so magical after all.