Poll: What’s the worst PC hardware bug?

Intel’s 6-series chipset bug has dominated the headlines this week, and for good reason. The flaw has been traced back to a single transistor and only affects the long-term performance of the chipset’s 3Gbps Serial ATA ports, but it’s effectively brought the Sandy Bridge platform to a grinding halt. You can’t run a Sandy Bridge CPU without a 6-series chipset, and the new stepping isn’t due to start shipping out to Intel’s customers until the end of the month.

This rather unique situation has us wondering where the chipset bug fits among other PC hardware flaws. Is it worse than the Phenom TLB bug or the problems Intel had with the old 820 chipset’s MTH? What about IBM’s notorious Deathstar 75GXPs and the Pentium FDIV bug? We’ve come up with a list of PC hardware bugs for our latest poll, and we’d like to know which you think is the worst

In last week’s poll, we asked where you get most of your video content. As expected, "Yarrr" proved the most popular answer, commanding 35% of the vote. 22% of folks rely primarily on a cable/satellite/digital TV service, while 20% stream most of their video online. Over-the-air TV broadcasts are the primary video source for 6% of those who voted, with the same percentage favoring DVD and Blu-ray purchases. Only 1% get most of their video through downloadable purchases, while 5% have other sources. Finally, 4% of voters would like to point out smugly that they don’t watch video.

Comments closed
    • erek
    • 9 years ago

    [url<]http://cgi.ebay.com/IBM-Deskstar-60GB-HDD-75GXP-DTLA-307060-IDE-ATA-/220731414717?pt=LH_DefaultDomain_0&hash=item33649e18bd#ht_500wt_1156[/url<] as bad as those other bugs were they didn't seem to be big enough to kindle selling off an entire HDD division, so i voted for the Deathstar 75GXP

      • indeego
      • 9 years ago

      IBM wanted out of low end/low margin hardware, the money is in services anyway. They made the right decision, and probably would have sold regardless.

    • LoneWolf15
    • 9 years ago

    Sahrin had some good points. The following bugs weren’t even mentioned.

    nVidia Purevideo bug (Geforce 6xxx series)
    nVidia nForce3-4 broken hardware firewalls
    nVidia laptop GPU failures

    nVidia did a great job of sweeping a number of their bugs under the rug. Fortunately, the laptop GPU bug was too big to ignore by both end-users and laptop vendors, and it forced nVidia to actually do something.

    While I think nVidia makes some great products, I think they’ve had some issues, which they haven’t always done a good job of addressing. Half of a “worst bug” is how the company handles fixing it for vendors and end-users. To this, I’d say that the Deskstar fiasco (some people never saw replacements for the drives they RMA-ed, as the queue got longer and longer and support was bungled badly) and some of nVidia’s issues (which I think were due to hoping to avoid responsibility) rank as the worst.

    I can’t count Sandy Bridge, because Intel found the error early, admitted to it, and has responded to its vendors, working with them to satisfy customers –exactly what should happen. IBM and nVidia, on the other hand, hid the errors (or attempted to dance around them), and botched handling their customers once issues came to light.

    • blacksteel
    • 9 years ago

    I voted Deathstar, it did the most damage because you lost data over it and was wide spread. I even bought a Deathstar before I found out it was a bad drive and yeah it failed on me. Sandy Bridge problem kind of minor since not a lot of people upgraded to it and Intel is working on fixing it.

    • The Egg
    • 9 years ago

    I had one of the dreaded 820 chipset motherboards. They were designed to use RDRAM, but since nobody wanted to buy that cr*p, they used an add-on translator chip (Memory Translator Hub) which allowed the use of standard SDRAM.

    I’m pretty sure every single one of the MTH chips had were defective, so all the SDRAM 820 boards (probably 90% of them) were severely unstable. I missed the tiny window to get my board replaced, so I ended up with a useless board. To give you an idea of how bad it was, this was for Pentium III Coppermine (S370) and after that I don’t think I built another Intel system for myself until Core2Duo.

      • yuhong
      • 9 years ago

      Yep, the problems with it also ended up cancelling the Timna project, which integrated graphics and a RDRAM memory controller into the CPU.

    • FireGryphon
    • 9 years ago

    I chose Deathstar because data integrity problems are the most severe and sinister of them all, and that particular issue was a problem only after the 75GXP was widely used by consumers.

    • slaimus
    • 9 years ago

    Ones I have had to deal with:

    1. VIA 686B southbridge IDE corruption (Temp fixed with the Bremse patch)
    2. Nvidia nForce3 “Active Armor” firewall and general ethernet bluescreens and other problems (Never fixed, ended up disabling it in BIOS and using a separate NIC)
    3. AMD Athlon X2 core synchronization problem (Fixed with dual core optimizer)
    4. Create X-Fi popping and sudden loud buzzing sounds (got better with newer drivers)
    5. Seagate 1.5TB sudden death

    • wiak
    • 9 years ago

    Opteron/Phenom TLB bug was not the biggest issue, why? you could fix by updating bios or install a patch, sandy bridge issue is alot bigger as its not fixable in software/bios and can make you harddrives disappear thats pretty much a bigger issue than a cpu crash etc

    anyway cant comment on the other bugs but still, something related to harddrives are alot bigger than a cpu, after all you data is on your harddrive most of the time

      • poulpy
      • 9 years ago

      Blows my mind as to why 13% (as of now) think it was the biggest bug as I am yet to hear about anyone being affected in the real world even without the patch.
      AFAIK you could get it to fail in synthetic test scenarios with shitloads of VMs loaded to the top with concurrent processes.

      I guess as TurtlePerson2 put it, it was probably the context and public reception that was bad in term of PR, more than the actual severity and/or impact (or lack of) in the real world.

        • UberGerbil
        • 9 years ago

        Yes, it’s impossible to know what metric of “worse” people are using. “Worst” in terms of dollars? Worst in terms of number of people affected? Worst in terms of downtime, or data lost, or how much it personally affected them? Or worst for the reputation of the company involved?

    • kamikaziechameleon
    • 9 years ago

    So many people upgraded to sandy bridge, I would definately see this as a big flub on intels part.

    • swalkenshaw
    • 9 years ago

    4-5 years ago Seagate made a run of 250GB hard drives that had a 20% failure rate. At the time I had about 100 of those drives in workstations and servers. I even had a RAID 6 array that kicked 2 drives out of the array and a 3rd failed as a hot spare. The array was 1 drive failure away from office data loss. I was on a cruise and checked my E-mail and saw the error messages. I had to VPN into the office from the middle of the Bermuda Triangle and get the data off the server and remap to a workstation. Got it done and never lost data, but no way I want to spend a cruise.

    • Arclight
    • 9 years ago

    On a side note, would you call “erectile dysfunctions” one of the greatest human males “hardware” bug?

    • rephlex
    • 9 years ago

    But the [b<]least[/b<] serious hardware bug would be the [url=http://en.wikipedia.org/wiki/F00f<]f00f[/url<] bug.

      • CheetoPet
      • 9 years ago

      Glad I’m not the only person whose mind drudged that one up. CMPXCHG8 FTW!

    • continuum
    • 9 years ago

    I’m pretty sure I’m still under NDA about the worst bug I’ve encountered…

    • Sahrin
    • 9 years ago

    I wouldn’t go for this one myself, but I think AMD’s K7 thermal protection (or lack thereof) deserves a mention. The THG video of a Duron being burnt alive was pretty stunning back in the day, and given the “well, duh” nature of the problem and Intel’s comparatively simple and elegant solution there was plenty of egg to go around.

    Also, nVidia’s “bad solder” notebook cards should make the list. Easily among the most expensive recalls.

      • Peffse
      • 9 years ago

      have a link to that video?
      I’ve never heard of Durons having a problem.

        • Sahrin
        • 9 years ago

        [url<]http://www.tomshardware.com/reviews/hot-spot,365-6.html[/url<] EDIT: The THG link doesn't work, this is the same video on Youtube: [url<]http://www.youtube.com/watch?v=NxNUK3U73SI[/url<] EDIT AGAIN: A better version of the same: [url<]http://www.youtube.com/watch?v=Xf0VuRG7MN4&feature=related[/url<]

      • UberGerbil
      • 9 years ago

      I don’t know, that always seemed a bit of a sensationalistic attempt to make mountain out of a molehill. It seemed far less significant than the Sony exploding batteries problem, which actually happened to real people in real life who didn’t work for THG.

        • Sahrin
        • 9 years ago

        You’ve never known someone who this happened to? Admittedly, it was a huge edge case and I don’t think it’s the worst on this list (any sort of data corruption error be it through execution or disk failure is fundamentally worse than just a plain old hardware failure) but it’s not ‘fake’ and clearly was an oversight.

          • flip-mode
          • 9 years ago

          I’ve never known anyone that happened to either in corporeal life or on the TR forums. Did I use corporeal appropriately?

          • UberGerbil
          • 9 years ago

          No, no one I knew online or IRL at the time had this happen to them. (But then I never hung out with overclockers). I’ve seen the magic smoke come out of lots of things — monitors, motherboards with bad caps or stuck northbridge fans, even a hard drive controller board once — but I never saw a Duron do that. But note that nowhere did I use the term “fake” and that’s not what I was implying or complaining about. I just thought the whole issue got played up — complete with sensational video — to get page hits for THG. The mindshare that “problem” thereby acquired was completely out of proportion to its likelihood or its significance, especially compared to some of the other bugs on this list.

            • Krogoth
            • 9 years ago

            Socket A era was full of cracked cores and improperly installed HSF killing CPUs. πŸ˜‰

            I was a victim of the latter with not one, but two chips. πŸ™ The second time wasn’t really my fault though.

            Fortunately, AMD caught up with Intel in their K8 platform and onward. (Headspreaders FTW)

            • UberGerbil
            • 9 years ago

            Yeah, I knew a couple of people who did that. That’s not exactly the [url=http://en.wikipedia.org/wiki/Halt_and_Catch_Fire<]HCF[/url<] failure mode we're talking about here, though.

            • just brew it!
            • 9 years ago

            In spite of building many Socket A systems, I managed to never get “had” by the cracked core issue. I had a couple where the corners got chipped (with no apparent ill effects) though, so there was definitely some risk of physical damage.

            Heatspreaders are good from the standpoint of being idiot proof. Unfortunately, they potentially hurt your thermal performance since they introduce a second thermal interface. I’m a little surprised that there aren’t more people who are “uncapping” their CPUs (removing the heatspreaders) to try and gain a little more thermal headroom.

            • UberGerbil
            • 9 years ago

            I seem to recall some sites experimented with that when the heatspreader was first introduced and the general conclusion (IIRC) was that it was difficult to remove without damaging the chip and the result wasn’t worth it.

            • Krogoth
            • 9 years ago

            The cracked core problem mostly affect people with oversized, aftermarket HSF not the stock HSF.

            • clone
            • 9 years ago

            I personally never cracked an AMD K7 but 1 friend of mine cracked 3 using the same stupid heatsink each time.

            it was a thermaltake and the guy didn’t have a lot of coin and I had to loan him a K7 duron cpu that he cracked, I asked “wtf Dan I’ve yet to break one of these cores what are you doing?”

            “I think it’s this thermaltake heatsink I’m using this is the third core in a row.”

            “then stop using the F’ing heatsink for F’ks sake, you just killed my spare and you’ve got no coin!!!!”

            I had given him the stock heatsink but he though if he could get the Thermaltake on that he’d be able to overclock the 950 Duron higher.

      • blacksteel
      • 9 years ago

      AMD’s K7 thermal Protection, was it really that large of problem. I kind of doubt it was as wide spread or as destructive as losing data from the Deathstar. Sandy Bridge problem is just there because it’s news. It only effects SATA ports 2 and up. You can rectify the problem by using a SATA card.

    • d0g_p00p
    • 9 years ago

    I had to vote the Deathstar 75GXP mainly because it hit everyone I knew who had that drive, myself included. Plus the amount of people who it affected. I consider that drive and the XBox 360 pretty much the worse hardware bugs I have ever encountered. Like I said everyone i know had been hit by the DeathStar and everyone i know is on at least their second 360. I myself am on my 4th.

    Plus the DeathStar not only affected your computer and wallet but the lost of data. I lost a pair of 45GB disks which was *huge* at the time..

    edit: looks like most people agree.

    • jackbomb
    • 9 years ago

    Some of the first LGA775 P4 Prescotts would overheat and throttle when used with the bundled Intel heatsink.

      • AssBall
      • 9 years ago

      They sure did, or the power transistors on the motherboard around the socket would bubble their solder and fry. Even with aftermarket sinks I was always fighting with the power deilivery, heat, and problems.

    • jpostel
    • 9 years ago

    I voted for the Deathstar, but after reading all the comments, I was also hit by the bad caps (2 boards) and the Zip drive click of death.

    The Zip was pretty crazy for me, because we had standardized on them at the company I was working. We had SCSI, Parallel, and IDE varieties installed. The company did statistical analysis of medical/pharma data and the datasets were 50-500MB, so we had lots of Zip and Jaz drives around to make them sneaker net capable. Click of death was a serious problem for us. I still have Zip disks and at least one working drive in my house, although not in a powered on PC.

    • willmore
    • 9 years ago

    Where’s the 32-bit i386 glitch? I’m pretty sure I can dig up a double sigma part if anyone wants to see one.

    Though, I’d have to say the bad caps was the largest impact on me. I had a number of systems die of that. It was industry wide and lead to the current trent of solid state caps.

      • bthylafh
      • 9 years ago

      What 32-bit 386 glitch is this, then?

        • yuhong
        • 9 years ago

        The 32-bit multiply bug. 386 B1 stepping errata was recently mentioned on Raymond Chen’s blog:
        [url<]http://blogs.msdn.com/b/oldnewthing/archive/2011/01/12/10114521.aspx[/url<]

          • willmore
          • 9 years ago

          I hadn’t seen that, yuhong, thanks. I’ve got a nice double-sigma i386 in the basement. Came out of a Samsung high end desktop from ‘back in the day’. Big old box had a DTP ESDI RAID controller in it. Huge sucker with a bar graph display that was fun to watch. πŸ™‚

    • FuturePastNow
    • 9 years ago

    Bad caps. I’ve lost some primo old hardware to them. Athlon XP and 64 systems that would still be good for web and basic tasks, now junk.

    Yes, I know they can be recapped, but they shouldn’t have failed in the first place.

    • rephlex
    • 9 years ago

    I have two 45 gig 75GXP Deskstars in a system I use for some basic digital audio work. They both still work fine. I have updated the firmware in them both though, as IBM recommends. This fixes at least one important bug and also supposedly does some stuff with head positioning which is supposed to make drive failure less likely. Simple job in DOS with a download from their website.

    • jensend
    • 9 years ago

    The Zip drive “click of death” is probably the worst I remember. Iomega looked quite promising until they made that mistake (wikipedia says the main problem was foolish cost-cutting of a necessary washer which cost a handful of cents). Afterwards Iomega was kind of a walking corpse of a company for a decade until they got bought out.

    Because of that problem the already-long-obsolete floppy drive stuck around for several more years- we’d neared a consensus on its replacement but that vanished with the appearance of the bug, and no new consensus on floppy alternatives arrived until USB flash drives became common.

      • holophrastic
      • 9 years ago

      yup, I voted other purely because of your comment. I was in the zip-drive camp for a whopping six weeks before my first hey-this-is-the-same-as-a-floppy moment, except this time I lost far more than 1MB of data.

        • bdwilcox
        • 9 years ago

        There was that GRC utility [url=http://www.grc.com/tip/clickdeath.htm<]Trouble in Paradise (TIP)[/url<] that would diagnose your zip drive for an impending 'click-of-death'.

          • Palek
          • 9 years ago

          Oh wow, GRC. Blast from the (not so distant) past!

    • mnecaise
    • 9 years ago

    I’ve lost a computer to a bad chipset and another machine to bad caps; but, the worst, hands down, was a DeathStar 75GXP failure and the resulting data loss.

    • HisDivineShadow
    • 9 years ago

    The number of nForce chipsets with the potential for failing, corrupting data on a hd, or just plain making me hate my computer were way too many to be forgotten.

    That said, I think the Deathstar failures were probably the worst case of computer hardware for enthusiasts failing on a large scale. People STILL hold Deskstars to task for that despite the fact that the old ones were made by IBM and the new ones aren’t even IBM’s anymore. So that staying power is just too hard to discount…

      • Voldenuit
      • 9 years ago

      I’m one of the lucky ones – my 75GXP lived with me for years. I heard about the failures and backed up my data, but the drive never quit.

      I’ve been bitten by nforce chipsets, though only in a very limited sense. Installing windows XP x64 and 64-bit chipset drivers caused my nf4 ultra mobo to corrupt ethernet throughput. I didn’t realise the entire network backup had been corrupted, and lost a lot of files that way. My next XP x64 build was a Wolfdale + G35 chipset that ran without a hitch.

        • clone
        • 9 years ago

        what size is your deathstar, the 30gb’s were the worst of the bunch, others lasted much longer.

          • Voldenuit
          • 9 years ago

          Mine was either a 60 or 75 GB iirc.

    • not@home
    • 9 years ago

    I had an old Nvidia chipset that would slowly corrupt all the data on my HD. I remember trying to figure out why it seemed like I had a virus or something when the PC had never been on the internet. A search online from my other PC told me it was the chipset. That has to be the worst because everything seems fine by it is not. I cannot remember what chipset it was though.

      • JumpingJack
      • 9 years ago

      Any bug that generates silent data corruption or destruction is at the top of the list. I would vote for almost any nVidia chipsets specifically the 600 series, which ruined many a file for me.

        • rephlex
        • 9 years ago

        Weren’t Nvidia’s buggy IDE drivers to blame, not the chipset itself? I’ve never had any data corruption issues with my nForce 2 board, but I’ve only ever used Microsoft’s IDE drivers.

        • Flying Fox
        • 9 years ago

        600 series? data corruption was reported even back in the nForce 4 days.

          • JumpingJack
          • 9 years ago

          That was what I mean by most of them — I only had experience with the 600 series.

    • Hattig
    • 9 years ago

    Bloody deathstar. That’s data loss. All the others aren’t as serious, SATA – stops working but data intact. TLB – knocks a bit of performance off, but still works. FDIV – annoyed some scientists and accountants, probably second worst, but didn’t affect normal users. 820 … forgotten what that was all about.

      • anubis44
      • 9 years ago

      I agree that failed hardware is potentially worse than anomolous calculations (assuming there is a software/bios patch work-around). However, anybody not backing up there data is simply stupid in the first place; it’s just that the worst case scenario was far more likely with that series of HD. On the other hand, the fact that my Nvidia-powered Toshiba Tecra simply burned itself up prematurely must surely rate as a more egregious design flaw than a failing hard-drive. After all, a hard drive generally costs less than $100. A $2000 notebook computer, on the other hand, is significantly more expensive to replace. Jesus, you’d think Nvidia could perform some thermal testing in their lab and notice there’s a problem.

      In my view, from an engineering standpoint, that’s far more inexcusable than missing some incredibly obscure errata in a multi-million or billion circuit CPU, and one of the reasons I’m still relunctant to buy any Nvidia products since.

    • DrDillyBar
    • 9 years ago

    I voted “other” as Bad Caps definitely wins the day.

      • ClickClick5
      • 9 years ago

      You won my vote!

      • KarateBob
      • 9 years ago

      Bad caps are why my Samsung 204B LCD monitor from 2005 and Sony KDS-55A2020 SXRD TV from 2007 aren’t able to display pictures @ full brightness anymore.

      Sure, the Panasonic plasma i bought has it’s various quirks and issues….but nothing like low quality capacitors, which are ruining electronics.

        • ludi
        • 9 years ago

        I’ll be happy to take those off your hands and repair them…

      • nstuff
      • 9 years ago

      big dittos

      • ew
      • 9 years ago

      Of all the problems bad caps is the only one that is user fixable though.

      • ColeLT1
      • 9 years ago

      Bad caps should be on this list. Working a state DOT job, I replaced ~100 motherboards (bad EPOX bad) in a year’s time. After the replaced boards started popping, they gave up on them, about 3 of my friends ended up with free boards/cpu/ram, even my home server happened to be one of these, w/athlon xp 2800+, it died less than a year ago, bad caps.

    • cynan
    • 9 years ago

    Worst hardware bug to affect me (excluding random failures not attributable to a systematic “bug”) would have to be the lack of crossfire support for some early HD4850 (and other HD48xx cards?) release cards (ie, all Visiontek HD4850s) due to bios incompatibility with recent drivers – an issue which I have yet see AMD even admit. Took me weeks of poking around to figure out the problem.

    In a nutshell, crossfire no longer works with drivers later than Catalyst 10.5 or so with some of the earlier HD4850s. The second GPU just doesn’t do anything – it’s like you only have 1 card. Flashing the bios to ones from a more recent release card seems to fix this issue for the most part, but that obviously voids any warranty you may have and is potentially risky due to voltage setting incompatibility depending on which memory/components your particular card has.

    I’ve complained about it on TR boards/comments before, so apologies for reiterating, but I thought it was well within the topic of the pole.

      • MrBojangles
      • 9 years ago

      That actually clears up alot , at least for me anyways.A few years back i picked up 2 visiontek 4870’s at microcenter during a firesell.After i got them installed the crossfire and everything would show as active but performance across the board was barely a step up form just using a single card.literally like i had just upgraded to 4890 rather than a second 4870 in cf.Never did figure out the problem just ended up selling one and using the other as a single card.

      • Krogoth
      • 9 years ago

      My CF 4850 setup (Second-generation Visiontek 4850s) never had any hardware-related issues. The only persistent software issue that I had was some weird resource allocation issue. It kept causing 7/Vista to BSOD (Device_Exception_Error) when I attempted to CF on the card that the BIOS selected as the default video output device. CF worked fine on the other card which led me to believe that it was some kind of low-level software issue. This was further supported by the fact that XP32, XP64, Linux never had any of the issues. 10.7s for 7/Vista and onward seemed to have fixed it though.

      It was sure frustrating to figure out it though trial and error.

    • mutarasector
    • 9 years ago

    While not exactly at the top of the list in any recent, or even semi-recent category, at the time, one of the worst hardware bugs I ran across and was extremely frustrating were the old REV6 Amiga 2000 motherboards.

    • Meadows
    • 9 years ago

    The Deathstar. By and large.
    None of the other options threaten your data, and there’s no higher offense.

      • cygnus1
      • 9 years ago

      Agreed. I clicked the article ready to vote pentium/fdiv and as soon as i saw deathstar, i had to vote for it. Data destruction is the worst bug possible. And I had couple IBM drives that died. Only drives i’ve ever had that died so abruptly like that. Has made me overly paranoid about data loss to this day.

      • JumpingJack
      • 9 years ago

      The problem is both the FDIV bug and TLB bug above could potentially corrupt data, silently.

      The occurrence granted would be extremely low to the point of negligible, however, data corruption is an unpardonable sin regardless of the infinitesimally low odds of ever happening.

    • jdaven
    • 9 years ago

    I voted Pentium FDIV but I would like to add a few more:

    Nvidia bumpgate (like some others have mentioned)
    Nvidia FX series (not a bug per se but damn those coolers were so loud I don’t know how it got past QA)
    TSMC 40 nm process (whole lot of mess there)

      • Rakhmaninov3
      • 9 years ago

      LOL aaahhhh the Dustbuster. Nvidia punished Damage for telling the truth about them and made him buy an FX card out-of-pocket to review. It was loud and walloped by the 9700 Pro at the time.

        • yuhong
        • 9 years ago

        Here is the worst thing about it, though:
        [url<]http://en.wikipedia.org/wiki/Nvidia#Stumbles_with_the_FX_series[/url<]

    • Laykun
    • 9 years ago

    Where is the nvidia 8 series laptop discrete GPU hardware issue?

    • Laykun
    • 9 years ago

    Where is the nvidia 8 series laptop discrete GPU hardware issue?

    • ShadowTiger
    • 9 years ago

    i am going to abstain from this poll since i have no clue what those other bugs are

    • Forge
    • 9 years ago

    Well, if you use the traditional metrics, measuring by likelihood to occur and likelihood to mangle data, it would be i820 MTH and DeathStar 75GXP way out in front. Second would be the ‘less likely to mangle’ FDIV and TLB, followed up by the relatively mild Sandy Bridge and TLB bugs, which generally just degrade performance.

    I’m voting MTH, since I owned one of those Gawd-awful things. I was a WD fan during the DeathStar days.

      • Krogoth
      • 9 years ago

      Still have that one board (affected by MTH bug) in your collection? πŸ˜‰

    • TurtlePerson2
    • 9 years ago

    TLB was probably the most devestating as far as CPU errata go. Sandy Bridge and Pentium came out and were great, then they found a bug. Phenom came out and was mediocre, then it had a bug. AMD underdelivered with Phenom; the TLB bug was just another nail in the coffin for the first generation of Phenoms.

      • clone
      • 9 years ago

      that’s exactly why the bug really didn’t matter, very few had even considered buying the cpu in the first place so the recall was minimal and the impact on sales regarding a crappy cpu was almost nothing.

      had AMD’s quad been incredible surpassing everything Intel had by 20% the sales would have been huge, the impact huge.

      the cpu was crap, that it also came with a bug just made it crappier.

    • TurtlePerson2
    • 9 years ago

    TLB was probably the most devestating as far as CPU errata go. Sandy Bridge and Pentium came out and were great, then they found a bug. Phenom came out and was mediocre, then it had a bug. AMD underdelivered with Phenom; the TLB bug was just another nail in the coffin for the first generation of Phenoms.

      • TaBoVilla
      • 9 years ago

      I didnΒ΄t know Craig Barrett had a profile on TR, nice!

    • bwcbiz
    • 9 years ago

    Ranked order:

    Deathstar caused the worst data loss and loss of system functionality, so that comes first.
    nVidia bumpGate (missing from list! (?)) was a big loss of functionality but not data.
    Sandy Bridge has a performance impact plus (unproven) possibility of data corruption.

    Probably the MTH would be pretty close to the Sandy Bridge bug, since it would cause random system crashes, but apparently it wasn’t that common.

    • ChunΒ’
    • 9 years ago

    What about the bad capacitors a few years back? I thought they would surely be on this poll

      • bthylafh
      • 9 years ago

      Geoff probably suppressed that memory. πŸ˜›

    • l33t-g4m3r
    • 9 years ago

    Amd’s bugs I thought was the worst. The old X2’s had some stupid bugs too. It’s almost like they didn’t have QA back then. Second has to be bumpgate.

      • Krogoth
      • 9 years ago

      What X2s? If you are referring to the original A64 X2s. I do not believe they had any sort of hardware bug that was problemic.The issue at the time was due to NT5-based OS (XP, 2000, 2003). They didn’t know how to properly deal with multi-core chips (dual-core P4s were also affected πŸ˜‰ ). Several hotfixes from MS and AMD resolved this problem.

      The only CPU related bug that was infamous from that time period was the issue with Winchester (AMD’s first attempt with 90nm, single-core) memory controller. The vast majority of them wouldn’t POST with four DIMM slots popluated. This bug was fixed with Venice ( second-gen 90nm, single-core) that came out several months later.

      • clone
      • 9 years ago

      X2’s didn’t have any significant problems aside from being new and being 2 processors on 1 package,, the problems were on the software side because not all of it was ready for dual core early on.

      I believe their are a few in here who don’t remember how bad the IBM deathstar issue really was, nor the Fujitsu FDB hard drive failures.

      to any not familiar think you buy 4 IBM deskstar75 hard drives, the fastest IDE 7200 rpm drives on the market at the time and 4 months to a year later they are all dead as are a cppl of the replacement drives IBM sent you.

      you buy 4 Fujitsu Fluid Dynamic Bearing drive and 6 months later 3 are dead and the 4th dies a month after that… although the replacement Fujitsu’s didn’t all die, a friend has a replacement 40 gb on it’s 12th year still going.

        • Dashak
        • 9 years ago

        Your friend has balls of steel.

    • potatochobit
    • 9 years ago

    my old first gen SCSI CDRW running windows 98 used to piss me off

      • jackbomb
      • 9 years ago

      Anything running Windows 98 used to piss me off.

    • anotherengineer
    • 9 years ago

    I voted other.

    Although the FDIV is up there, I think some of the ones that happened long ago (that were major bugs) probably aren’t as bad as more minor recent bugs due to the infiltration of computers everywhere relative to 8 yrs ago.

    I remember the panic about the y2k calender rollover bug, how many people and $$$ were spent trying to avoid that one??

    • bthylafh
    • 9 years ago

    The Deathstar, certainly, since most people don’t do backups and data is quite valuable. The Seagate 1.5 TB drive fiasco as well for the same reason.

    The Intel SATA thing has to be a close second, though. It’s their comeuppance, though, for not letting third parties make chipsets for their newer chips.

      • ludi
      • 9 years ago

      Ayep. The others are nuisances. Irritating and possibly expensive nuisances, but nuisances.

      Deathstar was the first major decline in the quality of a component whose failure may produce a nigh-irrecoverable loss.

      • Sunburn74
      • 9 years ago

      I don’t even remember the seagate 1.5tb fiasco (and I own one of those drives). What was the problem again?

        • derFunkenstein
        • 9 years ago

        Streaming data at low speeds would cause it to stop responding. You could update hte firmware a couple weeks after launch to get a fix, or you could send the drive to them for a new one. It was so short-lived (or should have been, if you owned a drive and paid attention) that it hardly registers compared to the Deathstar.

    • GrangerX
    • 9 years ago

    Are you guys kidding me? The FDIV bug meant that Pentiums couldn’t even do grade-school math correctly. They didn’t crash, they didn’t blow up, they just … silently introduced hideous corruption to any floating-point math application that was run on the processors.

    How many multi-billion dollar things like NASA missions and even scarier possibilities like corrupted measurements for nuclear reactor designs do we need before we rate what folks used to refer to as the Intel “Quality is Job 0.999739” bug as the most terrifying, and therefore, worst, hardware bug?

      • bthylafh
      • 9 years ago

      Wrong, fanboy. The Pentium FDIV bug only affected certain floating-point division operations, and almost nobody back then was affected. Sayith Wiki, only about 1 in 9 billion FDIV operations would have been affected. However, people expected their chips to /work/ no matter what, and Intel screwed up epically on public relations.

      PROTIP: Don’t speak of things which you don’t understand. You just made yourself look like an ass.

        • GrangerX
        • 9 years ago

        Fanboy? Of whom? What in the world are you talking about?

        Well, “1 in 9 billion operations” sounds pretty low-risk, except that only one transistor in Sandy Bridge is bad, yet it forced a recall of every chipset in the market. I think the Sandy Bridge debacle shows the evidence that even one in a billion can be pretty disastrous if it’s the right one in a billion.

        Intel’s PR problems in that situation are only secondary concerns. I guess they learned their lesson for Sandy Bridge, at least.

        Protip, though? That’s adorable. I’ve been affected by the Sandy Bridge, Deskstar, Motherboard Capacitors, and Phenom bugs, and several others in the “other” category, and none of those are as unsettling.

          • ludi
          • 9 years ago

          You still don’t understand what the FDIV bug actually affected, and your earlier statement about the breadth of the error is completely wrong.

          The damage was limited almost entirely to Intel’s klutzed PR response.

            • GrangerX
            • 9 years ago

            I understand the technical ‘what’ the bug actually affected. I just didn’t focus on breadth so much as a potential breadth. I will concede the point that luckily few people were probably affected. I was more focused on the ‘class’ of bug and what silent corruption represents. Each and every one of these bugs is depressing as an individual affected user. If you’ve lost a Deskstar, you were in a world of hurt, regardless of what the world was doing that day.

            • bthylafh
            • 9 years ago

            You went for the most dramatic misinterpretation, you mean.

            • GrangerX
            • 9 years ago

            Most dramatic interpretation perhaps. For some reason I was given to hyperbole when I posted. Probably comes from actually being interested given the Sandy Bridge bug personally affecting me. Anyway, you, uh, win? Please choose the bug of your choice from top shelf of the list.

            • just brew it!
            • 9 years ago

            I don’t know if anyone [i<]ever[/i<] demonstrated actual harm from the FDIV bug; it was discovered by a mathematician doing number theory calculations. The bug had an extremely low probability of occurring in practice, and even when it did occur the magnitude of the error was small enough that it was generally harmless in most real-world situations. As has already been mentioned, the biggest effect of the FDIV bug was Intel's bungled PR response. If you're really worried about the [i<]potential[/i<] for corrupted data, we should talk about a pet peeve of mine, specifically the fact that only a small percentage of desktop PCs use parity or ECC protection on their DRAM. Flipped bits can and do occur; about 20 years ago I was involved in a project where we had to measure the background error rate for what was then a fairly large DRAM array (20GB). We found that we were getting on the order of 1 randomly flipped bit per day. DRAM error rates have improved significantly since then; but the very nature of DRAM means they are still susceptible to random bit errors triggered by normal background radiation levels. Without any sort of data integrity check on your system's DRAM, these bit errors certainly have the potential to silently corrupt data. Why don't more systems use ECC DRAM? IMO it is a combination of A) people being cheap (ECC DIMMs are 72 bits wide instead of 64, and therefore cost a bit more); and B) Intel's desire to artificially segment their product line (if you want a system that supports ECC you generally need to buy something that uses their "server" chips). B is actually one of the reasons I tend to prefer AMD CPUs over Intel, and Asus over other desktop motherboard vendors -- nearly all of AMD's CPUs support ECC DRAM, as do most (possibly all) of Asus' AMD motherboards.

      • UberGerbil
      • 9 years ago

      I had a P90 with the bug. I never bothered replacing it, and I used it for several years without any problems. In part this was because the bug was obscure and wouldn’t and didn’t manifest in most of the things I was doing; but more importantly, [b<]it was patched in software soon thereafter[/b<], including a patch within the Windows kernel. So if you did happen to stumble into a calculation that actually would be affected, it was just silently fixed: it ran a little slower but still gave you the correct result. And that "little slower" was imperceptible in everything but benchmarks designed expressly to show it. So, now, can you tell me about how the Deathstar drives, or the Sandy Bridge issue, or the Dead Caps, were fixable in software?

        • GrangerX
        • 9 years ago

        The FDIV bug caused Intel to change their CPU architectures to allow CPU microcode updates, so, in a way, this particular bug created a whole architecture for fixing this kind of bug. Do any of the other listed bugs share that dubious distinction?

        Yes, the software solution to the P90 problem is probably fine. Sorry you had to deal with it, though. If you still had the P90, you could ship it back to Intel today. Be interesting to see what they’d send you back in return. I hear they have a few Sandy Bridge CPUs to spare right at the moment. πŸ™‚

        No, I can’t tell you how non-software-fixable bugs can be fixed in software. If they were, they wouldn’t have probably been in the list to start with.

          • UberGerbil
          • 9 years ago

          No need to extend your sympathies — like I said, the bug had essentially no impact on me (it didn’t even affect me enough to motivate me to get the free replacement). And I’m sure Intel could scare up a P90 (perhaps a salvaged one) if they needed to, though I suspect the term of that offer has run out. You’d be surprised at what they still make and sell from their older fabs — embedded equipment makers can still get their hands on sub-1GHz Celerons, for instance.

    • Krogoth
    • 9 years ago

    You forgot the infamous broken hardware firewall on southbridge of Nforce 3/4. πŸ˜‰

      • KarateBob
      • 9 years ago

      Wasnt that still broken in 5, and removed in 6? πŸ˜›

      And the Nforce 680i IDE failure? πŸ˜›

        • Krogoth
        • 9 years ago

        I thought it was removed from 5.

        i680 also had booting issues with Q6xxx/Q9xxx chips.

    • KarateBob
    • 9 years ago

    [x] Other. NVIDIA’s bad bumps, and Microsoft’s lead-free solder causing billions worth of RRoDing are pretty bad. Seagate’s 1.5TB 7200.11 firmware fiasco was a huge mess, but thats firmware right ;).

    • GraveDigger
    • 9 years ago

    From Wiki:

    “The drives were also known for an unusually high rate of head crashes, due to the magnetic coating soon beginning to loosen and sprinkle off from the platters, creating dust in the hard disk array and leading to crashes over large areas of the platters. The combination of two technologies that were quite new at the time, GMR heads and glass platters, are said to be largely to blame for the issues.
    In addition to the failure that had led to the lawsuit, additional flaws were found in the Deskstar 60GXP, 75GXP, 120GXP and 180GXP, caused by the way the Giant Magneto Resistive read/write heads interact with the stored data, and the easily corrupted NVRAM chip. It was even discovered that IBM had used a badly designed and laid out printed circuit board and had used a soldering alloy of a poor quality on them. Over time, the contacts of the chips loosened, in turn causing firmware corruption. The drives were also quite susceptible to burn damage to the controller board.”

    The Capacitor Plague got undergrad-level chemistry wrong, but Deskstars were failure prone in so many different ways that it would be comical if not for the data loss.

    Still, the Plague should be a late add to this poll.

    • flip-mode
    • 9 years ago

    Phenom / Opteron TLB was the worst, but only because it was so harmful to AMD, even though it probably never hurt anyone in the real world.

    As for the bug that cause the most harm to end users – probably the Deathstar or, as someone else mentioned, the capacitor crisis.

    I have no memory of the 820 MTH issue – what was that? I don’t really know about the Pentium FDIV beyond the fact that it happened.

    I don’t think the Sandy Bridge chipset SATA issue is really that big of a deal. It’s quite easily circumvented.

      • bdwilcox
      • 9 years ago

      820 MTH was a flawed Memory Translator Hub that was supposed to allow expensive Rambus boards to use normal SDRAM. The Intel FDIV bug was the Pentium Floating Point errata.

    • GraveDigger
    • 9 years ago

    I voted for the Deathstar, because Fujitsu’s HDD chipset phosphor coating issue is the *best* hardware bug of all time for making me so much money as a contractor. Replace the drive on client expense and keep the rebate cheque for the dead one.

    Also, how the hell is the Capacitor Plague not on there? In terms of volume it’s #1 of all time, and was often extremely difficult to diagnose if a tiny cap blew somewhere unseen like between your bottom PCI slot and the floor of the case, or obscured by the CPU cooler.

    To make matters worse, IBM had some desktops during that time using plagued boards AND Fujitsu HDDs.

    • DancinJack
    • 9 years ago

    75GXP for sure. Nothing worse than losing data. Ugh.

      • Sanctusx2
      • 9 years ago

      Completely agree. All the other bugs were fairly insignificant when it came to data integrity for everyday users. Academics of course might argue differently, but I don’t think any of the other design failures rival the Deathstars.

      I can deal with a slow SATA, or I can replace it. A dead or malfunctioning hard drive means hours of attempted recovery, restoration from backups, flipping out because I remembered I don’t keep backups(or they’re very old), followed by the slow, painful re-installation and reconfiguration of windows and all software.

      Luckily enough, I never had problems with my handful of IBM drives from way back when, nor my dozens of Seagates. Now WD is another story…

      • NeelyCam
      • 9 years ago

      Deathstar FTW!

    • indeego
    • 9 years ago

    Currently for me it’s the OCZ Vertex 2 that just died on me. “Other”

    Murphy’s law, as just last week I was bragging to my brother that I haven’t had a mechanical or SSD drive die in many years.

      • just brew it!
      • 9 years ago

      That’s not necessarily a bug or design flaw, it could just be a manufacturing defect.

    • Silus
    • 9 years ago

    Can’t vote on the poll, without informing myself on some of these other bugs, that I never heard of. So yeah, this post is useless πŸ™‚

    • kenclopz
    • 9 years ago

    I had a Asus P4G8X motherboard, one of the first with hyperthreading and dual channel ram, and AGP 8x… except that it only worked with Nvidia AGP 8x, not ATI. It was later stated by Intel that it was due to errata and never fixed. This is my problem with motherboard makers. YOU BUY A TOP OF THE LINE MOTHERBOARD, and THERE ARE STILL PROBLEMS WITH IT!!! I wish Asus or Intel would give me a free computer for all the hassle and grief they’ve put my through.

      • Ihmemies
      • 9 years ago

      I had a similar problem with AMD 750 or whatever chipset back in 2000.. it was stated to support some faster AGP speed but in reality supported only something like 1x without constant BSOD’s.

    • LauRoman
    • 9 years ago

    The worst PC hardware bug is, of course, the one you don’t find prior to release. πŸ˜€

    • Deanjo
    • 9 years ago

    The Seagate 7200.11 firmware debacle should have been in the poll as well. The fujitsu HD scandal is probably worth mentioning. It is one of the few times where I have seen a manufacturer say “we can’t fix our screwed product, here is some money back to purchase another brand”.

      • albundy
      • 9 years ago

      totally agreed. they never disclose problems with these drives.

    • evilpaul
    • 9 years ago

    The worst PC bug was that problem with capacitors that swelled and burst from 8 or 9 years ago. It hit motherboards pretty hard and across a wide variety of brands, chipsets, whatever.

      • Deanjo
      • 9 years ago

      +1000, one the flipside however if you were handy with a soldering gun you could usually replace those caps easily enough and they would keep going for a long long time.

        • albundy
        • 9 years ago

        hah, i was forced to pick up that skill when cheap@ass mobo makers like asrock sold socket 478 boards with low quality cappies.

      • kvndoom
      • 9 years ago

      Well my Epox Socket 754 board had more than a few leaked caps on it when I retired it last year, but amazingly I never had any performance issues out it, and I used it for 4 years. At least in my case, it wasn’t world-ending. A dead hard drive is a pretty big deal though.

      • mutarasector
      • 9 years ago

      I’m *still* running across old mobos dying from bad leaky swollen caps – at least 4-5 this year.alone.

      • derFunkenstein
      • 9 years ago

      It so much affected so many people that today that’s the first thing many people (and I) look for in a flaky computer.

    • clone
    • 9 years ago

    I was about to rank the Sandy Bridge debacle as pretty high in significance but then I remember the MTH Intel issue and you guys even went back far enough to mention the Deathstar debacle…… you forgot to note the Fujitsu FDB failures though….. and both the IBM and Fujitsu disasters resulted in masses of failed drives for both companies that eventually killed their hard drive business entirely.

    AMD’s TLB bug was a non issue that only hurt them because their quad core processor happened to suck royally compared to Intel at the time, it was the “extra” excuse not to buy it even though to this day I’ve yet to find someone somewhere who experienced it personally.

    should have included the only Intel Floating point Pentium MMX disaster as well and the Intel transition to 90nm as that was disastrous… actually you could have simply said the entire Pentium 4 series given how badly they have aged and that they were the only series released by Intel that cost them massive amounts of marketshare and during the transition to 90nm failed at a surprisingly high rate initially.

      • bdwilcox
      • 9 years ago

      The choice “Pentium FDIV” IS the Intel floating point unit errata. Another Intel muck-up was the 1.13GHz P3 that was recalled due to cache instability.

        • ludi
        • 9 years ago

        True, but there were so few of those out and about that it was practically a non-event in the consumer sphere. I think there was a joke going around that the scope of the recall was pretty much limited to Toms’ Hardware.

          • clone
          • 9 years ago

          lol, I remember that one and if memory serves indeed very few cpu’s ever made it to consumers.

          come to think of it wasn’t the P3 1.13 cpu supposed to be destined only to OEM’s most notably Dell at the time but never made it because of the bug?

      • d0g_p00p
      • 9 years ago

      What is the “the Sandy Bridge debacle”

        • clone
        • 9 years ago

        [url<]http://www.businessinsider.com/intels-billion-dollar-mistake-2011-1[/url<]

    • OneArmedScissor
    • 9 years ago

    Where’s Nvidia? Bumpgate…cards frying on menu screens…Fermi lol

    Destroyed hardware seems just slightly worse than hardware with an issue that can be worked around.

      • anubis44
      • 9 years ago

      Nvidia bumpgate had the worst personal impact on me: I sold my mother-in-law a perfectly good (so I thought) Toshiba Tecra M3 with an Nvidia 6200Go chip after using it for a couple of years. About 3 months later, the damn thing was fried. Bear in mind, my mother-in-law used it for web-browsing and word processing-not exactly GPU-intensive tasks. Needless to say, I had quite a bit of explaining to do, after I had assured her that it was a rugged machine that would likely serve for many more years to come, like my old T-series IBM Thinkpad T21. Good thing she really likes me and trusts me!

      • MrBojangles
      • 9 years ago

      Agreed That’s exactly why i voted other.Was wondering “Where’s the nvidia driver bug that was frying there cards??”

      The again in hindsight it may not be on the list since it was a driver issue, and not actually apart of the hardware itself.

    • moshpit
    • 9 years ago

    75GXP, easily. There’s no getting around a dead drive. Every one of the other bugs mentioned had an escape route for users to avoid the worst of the bug’s effects. The 75GXP issue had no such escape route, the drive would go belly up with no warning and that was all she wrote for the data on the drive.

      • Bombadil
      • 9 years ago

      Swapping the PCB got mine working again.

        • moshpit
        • 9 years ago

        Which is great for those lucky enough to have an spare PCB laying around…

      • Clint Torres
      • 9 years ago

      Here here! Nothing more catastrophic/expensive than lost data!

        • Mourmain
        • 9 years ago

        It’s “hear, hear”, as in “listen to this man” or in a more poetic phrasing, “this”. Why do 80% of people get this wrong?

          • indeego
          • 9 years ago

          [b<]Because[/b<] it's for the most part a verbally transmitted expression and it doesn't really matter how it's spelled, [b<]irregardlessly[/b<].

          • bdwilcox
          • 9 years ago

          Here, here!

            • ludi
            • 9 years ago

            There, there.

          • ludi
          • 9 years ago

          I fully agree. However, your comment still makes you the homophone police.

          • Clint Torres
          • 9 years ago

          I never even thought about it. I consider my English to be idiomatic but you learn something new every day. Just last year Damage used the phrase “beck and call” and I always thought it was “beckon call”. Maybe I just don’t read enough.

            • Palek
            • 9 years ago

            Yeah, native speakers of English are actually at a disadvantage when it comes to spelling compared to people who speak English as a second language. You learn a lot of words and phrases through hearing but may not come across them in writing. Someone like me who had to learn the language from [s<]video games, movies and Beavis and Butt-head[/s<] books and classes will be more likely to hit up a dictionary when they come across a new phrase.

          • UberGerbil
          • 9 years ago

          As indeego says, you rarely see it written. And you can make the case that the other way makes equal sense, as in ‘Here is someone who agrees with you, right here!”

    • mmp121
    • 9 years ago

    Chose Deskstar 75GXP because I was bitten by this HARD several times. I had built 3 PC’s with the 45GB version of that drive. 2 of them failed. Both were RMA’d and those failed in short order. Will never purchased another drive from them again (IBM / Hitachi).

      • adisor19
      • 9 years ago

      They funny thing is i still have an 80GB Deathstar working wonderfully as my FreeNAS installation drive. It has been doing so for the last 3 years..

      Adi

        • bthylafh
        • 9 years ago

        and I got lucky with the 1.5 TB Seagate I bought on fire sale from Dell. It’s got the good firmware and saved me about fifty bucks.

          • l33t-g4m3r
          • 9 years ago

          I have two of the original seagate 1.5 drives. Never had a problem, and eventually flashed them to the updated firmware.

      • just brew it!
      • 9 years ago

      Hitachi drives are actually pretty good these days. I’d put their quality above Seagate’s.

    • bdwilcox
    • 9 years ago

    I would say ‘Other’ would be “Every chipset made by SiS and ALi, as well as most made by VIA.”

      • thermistor
      • 9 years ago

      I second this going back to when USB was deployed in the late 90’s…I had a VIA chipset that would absolutely freeze when a USB header (USB ports weren’t even on the port cluster) was connected to a USB device. So even though I had ‘USB’, I didn’t really have USB.

      That those chipsets got released to the public is mind-numbing.

        • mutarasector
        • 9 years ago

        Or worse. USB headers on motherboards that lacked standard pinouts, and trying to find documentation on the pinouts was problematic. I killed a couple of PC-Chips boards that way trying to connect a USB L-bracket.

          • bdwilcox
          • 9 years ago

          Remember the PC-Chips Lottery website that helped you to decode what board you had, what drivers and BIOS were applicable and any (near non-existent) documentation existed for it?

            • mutarasector
            • 9 years ago

            I sure do. ‘Lottery’ website is completely accurate, too.

        • highlandr
        • 9 years ago

        Remember the VIA Soundblaster fiasco? I didn’t have a soundblaster, so I thought I was OK. except every now and then, my machine would just hard lock. Usually while watching DVDs.

        Turns out the whole PCI bus issues affected my MX300 too, just in a different (though no less frustrating) way.

          • Palek
          • 9 years ago

          I used to have a system built around a VIA KT133A based Abit board that included a Soundblaster Live! card and yet ran without a problem, but I understand that I was in the minority.

          I also had a Deskstar 75GXP that DIDN’T fail. If I recall correctly, the faulty “Deathstar” drives were built in IBM’s factory in Hungary (which was subsequently closed), whereas drives originated in IBM’s other factories (factory?) were not affected at all. I think mine had a “Made in Malaysia” sticker on it.

          It seems that right now I am making up for avoiding those disasters back then. I have a WD Caviar Green 1TB drive that at first started to perform very infrequent disappearing acts and now phases in and out of existence regularly. The flakiness of this drive prompted me to seriously beef up my data back-up system with a RAID1-ed NAS device.

          [EDIT] Oh yeah, I voted for the Deathstar. Like others have said, there is no software fix for lost data.[/EDIT]

      • KarateBob
      • 9 years ago

      Only half of SiS’s chipsets are bad…The 530 in my K6-2 box is still going strong πŸ™‚

Pin It on Pinterest

Share This