Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

Building High Reliability Computers

By Bad Harmony in Technology
Thu Feb 15, 2001 at 07:24:49 AM EST
Tags: Hardware (all tags)

If the computer industry can put a million transistors on the head of a pin, why can't they make an inexpensive and reliable personal computer? A great deal (insert hand waving) of time and money is wasted in dealing with flakey computers.

Over the past few weeks, it seems like every computer that I have touched has suffered from major or minor hardware problems. Removable media devices that lock up or randomly generate I/O errors. Outright hardware failures. Motherboards that freeze several times a day. There is all this sophisticated technology, but it just doesn't work reliably. I'm not going to cloud the issue by ranting about broken software.

What can be done to build reliable and reasonably priced computer systems? When you look at catalogs, the only systems sold with ECC memory are the servers. Desktop systems get non-parity memory. A high percentage of consumer class computers are DOA (dead on arrival), the manufacturers seem to think that this is acceptable.

Mainframe programmers have told me that system failures are treated like airplane accidents. The cause of the failure is investigated, identified and changes are made to try to prevent a reoccurrence. None of this "hit reset and hope the problem goes away" that is common with small computers.

Can a reliable personal computer be built with off-the-shelf components? If so, what components would you use?


Voxel dot net
o Managed Hosting
o VoxCAST Content Delivery
o Raw Infrastructure


Related Links
o Also by Bad Harmony

Display: Sort:
Building High Reliability Computers | 30 comments (30 topical, editorial, 0 hidden)
Two Reasons (2.50 / 4) (#1)
by ucblockhead on Wed Feb 14, 2001 at 07:28:05 PM EST

  • Because computers are by far the most complex things ever designed by human beings. When you are talking about millions of transisters, you are talking about millions of elements that all have to be exactly so. Getting that sort of design right is extremely difficult (as Intel found out with the Pentium)
  • Because computer technology is changing so rapidly that all computers are designed under extreme time constraints, which means that the time to design properly is often lacking.

Combine those two things and you have, well, what we've got.

This is k5. We're all tools - duxup

May I add... (2.80 / 5) (#5)
by Phage on Wed Feb 14, 2001 at 08:13:19 PM EST

The market itself is different in that a frame manufacturer deals in a market that is high margin, low volume, and therefore is very mindful of its relationship with the client.

The PC business, however, is high volume, low margin. The manufacturers don't give a flying <insert noun> about the customer. It is all about units sold...

The large number of different manufacturers for PC components out there presents technical problems.
These products must all work together consistently. They must be able to work with a range of software and configs.
Whereas the range of mainframe configs and range of manufacturers is much smaller. They also tend to have a slower obsolescence rate due to the capital investment required to update them. Accordingly they have longer expected lifespans.
This leads to better customer support and generally higher customer expectations in this area.

I don't find Heathens to be sexy, as a general rule.
[ Parent ]

My preliminary diagnosis... (2.60 / 5) (#2)
by babylago on Wed Feb 14, 2001 at 07:43:17 PM EST

...is that the problem is the user, in this case you, as you touched all the systems.

Nevertheless, this gets +1 for hamfistedly making the points that cheap hardware is unreliable, that expensive hardware gets better tech support, and that maybe it should be different, given the amount of brainpower we seem to throw at engineering in general in this country.

[ Blog | Hunnh ]

Different standards (3.85 / 7) (#3)
by onyxruby on Wed Feb 14, 2001 at 07:55:47 PM EST

It is possible to build a reliable computer with off the shelf comonents. If you know what your doing, you can have a very rock solid system. Unfortunately the masses are not enlightened. This leaves it up to the marketing types to do the enlightening, and consumer quality is not high on their list. Compare a SCSI drive to an IDE drive. Your SCSI drive is much more reliable than an IDE drive, but is going to cost several times as much per gigabyte. This is primarily due to economics of scale. If there was as large a demand for the Quantum ultra 160 18 GB I want to buy as there was for the IBM Deskstar 45 GB, I could probably get a much better price. John Q Public just sees that they can get that 80 GB drive for about 2/3rd's the cost of the 18 GB drive, and simply go by the bigger numbers.

It also doesn't help that much of the technology we have moves at a rate several times faster than any other industry. Cars are much more reliable because the basics have been being built on for decades. Compare this to computers where your advancement rate doesn't allow for something like that. Sure, the architecture's are pretty similiar, but everything else changes. Three years ago I would be happy with a 8 GB Hard drive, compared to 80 GB hard drives that can be bought off the shelf now. That's an increase in capacity of ten times in three years. If the automotive industry moved at that rate, they would have just as many problems as the computer industry. They are also able to develop their product years before it is released to the public. Before a new car is ever built, supercomputers crunch aspects for everything from crash worthiness, service, and the HVAC. In the computer industry, you can't afford that kind of lead time for your next product.

It doesn't help matters that most manufactures no longer even build their own product. Almost every laptop is built in one of a few factories in Taiwan. Dell's, Compaq's, Toshiba's and the like all go down the line. To the employees, what does it matter if one is better built than the other? They use the components that the manufacture specs, and beyond that could care less. By removing ownership & pride in your product, your going to remove some quality. When Iomega was sued by their consumers for the "click of death" problem, it sent a message to the industry - hire better lawyers. The only way to get better quality out of manufactures is for consumers to demand it. If something is broken, send it back, do not allow a manufacture into talking you into "living" with a problem.

The moon is covered with the results of astronomical odds.

Reliability off the shelf (3.66 / 3) (#8)
by rusty on Wed Feb 14, 2001 at 08:26:48 PM EST

It is possible to build a reliable computer with off the shelf comonents. If you know what your doing, you can have a very rock solid system.

Absolutely true. The first box I built totally from scratch was the first K5 server (and is still chugging along serving scoop.kuro5hin.org and being our primary DNS). It cost me about $1000 bucks, and I never had a spot of trouble with it.

It's a dual Pentium Pro 200 (overclocked to 250 for a long time). The P2 was the newest intel chip when I built it, but my choice was between a P2 600 for about $500 and a PPro 200 for 60 bucks. I went for cheap, and incidentally, reliable. The PPro was a late-lifecycle processor, and the bugs had already been worked out. Plus, the pro had a bigger L2 cache than the P2 at the time, as it was originally intended for the server market.

The MoBo is an Intel PR440FX ("Providence"). No problems there. I looked around online for dual PPro motherboards, and that one came highly recommended. That was about $100. It takes ECC memory only, due to the demands of dual processor operation, so memory was a little more expensive, but highly worth it, IMO. If you're gonna spend extra money on a system, get an OK processor and some really good memory. I also splurged for SCSI, because I didn't need a huge fat wad of drive space, and plus, IBM 4.5G SCSI UW's were really cheap at the time ($200 each, I got 2).

And there you have it. Put the whole thing together, turn it on, it worked the first time. I had some issues with the 2.2 linux scsi driver locking up sometimes, but no problems with the hardware.

The key thing to take from this is that you can have a reliable system, but it might not be the "pushing the envelope of speed" system you'd like to have. Think about it though, will you be happier with 450fps in Quake, or with 360 days continuous uptime? Let the answer to that guide your design decisions and you'll be happy.

Not the real rusty
[ Parent ]

Tough Choice ! (2.66 / 3) (#12)
by Phage on Wed Feb 14, 2001 at 09:22:43 PM EST

Think about it though, will you be happier with 450fps in Quake, or with 360 days continuous uptime?

No, Nooooooo...... Must have fps....and must be able to play 24/7.....

What am I going to do ?

Hold on, there is a solution to this conundrum.
Oh yes, I cannot afford either.

I don't find Heathens to be sexy, as a general rule.
[ Parent ]

backwards compatibility... (3.25 / 4) (#4)
by Justinfinity on Wed Feb 14, 2001 at 08:06:31 PM EST

causes a lot of hardware weirdness. the designers want to make everything backwards compatible. they'd rather hack on an existing system, building a bigger kludge out of the existing kludge.

they could build a brand new system, without any legacy components. or at the very least design a new system made specifically for current peripherals. but that would be too expensive for their tastes.

some PC manufacturers are already trying to get rid of legacy components. many modern motherboards don't come with ISA slots. note: this doesn't nessacarily mean that the ISA bus is totally gone, as things like, serial, parallel and PS/2 keyboard and mouse ports often still run off the ISA bus, but it's a step closer by not allowing older, less compliant ISA cards in the system.

Indeed (4.00 / 3) (#7)
by Anonymous 7324 on Wed Feb 14, 2001 at 08:22:01 PM EST

... and things like USB and Firewire can easily replace most parallel port / serial port / bus port type interfaces -- but to sell a computer that doesn't work with many existing peripherals is like selling linux to the general public -- they just don't like the taste.

Things like serial to USB adapters are available, but at least for the Mac, which is the only computer type I've tried ordering these for, is expensive -- range of $50+ --> again, making consumers unhappy at the extra $$ they have to pay.

Another answer is that most individual home users don't _care_ if their computer screws up -- mainframe crashes cost businesses millions -- AOL wiping Joe Blogg's hard drive makes him scratch his head and call Microsoft. When it comes down to it, I guess the question is the same one that Linux users ask all the time -- why won't Windows switch, even though Linux is so superior?

Answer: inferior reliability is not a big concern, but compatibility is.

[ Parent ]
I don't see a problem. (4.00 / 5) (#6)
by Inoshiro on Wed Feb 14, 2001 at 08:18:27 PM EST

The workstations and servers I build for myself and others are as reliable as the power grid lets them be. If they have UPSes of enough capacity, they are as reliable as the weakest component of the system (usually the harddrive). That's a reliability of 3-5 years of constant uptime and no downtime for most IDE HDs. 10+ for most SCSI drives.

Why you have a problem is that you probably buy the "bargain basement" hardware, such as Maxtor drives. Do not buy these pieces of hardware which are 5-10$ cheaper, and have an exponentially lower life.

[ イノシロ ]
Maxtor Hard drives... (4.00 / 2) (#9)
by Fyndalf on Wed Feb 14, 2001 at 08:39:14 PM EST

Mine still works, August 1998 ...

Plus they're really good about replacing them under warranty, a friend of mine had his die a week after buying it, got the replacement *3 DAYS* later and sent his defective unit back in the box the replacement came in. It has worked since ..

Then again, yeah, an IBM Deskstar is generally worth the extra $$$

[ Parent ]
maxtor drives (4.50 / 2) (#11)
by h2odragon on Wed Feb 14, 2001 at 09:04:34 PM EST

'98 is about when they went to hell. I've maxtor 120MB drives that are running strong after 10 years; in late '99 and early 2000 I went through a maxtor experiance that has made me rant against them ever since. The story's long and boring, the final score was 60% DOA, 30% dead after less than a month, and me scrambling to find any drives that'd do the job with a budget abbreviated by the fact that maxtor doesn't (or didn't) do refunds, just returns for "refurb" drives that sucked worse than the new ones.

I second or third the IBM endorsement, I've had good luck with the higher-end Quantums, too.

[ Parent ]

Re: Maxtor drives (3.00 / 1) (#13)
by Fyndalf on Wed Feb 14, 2001 at 09:22:49 PM EST

Hmm. My friend's refurb replacement is still working, *shrug*.

I've had good luck with the higher-end Quantums, too.

Heh. Unfortunately for you, Maxtor bought Quantum.

[ Parent ]

Amusing (4.00 / 1) (#18)
by Miniluv on Thu Feb 15, 2001 at 02:24:38 AM EST

I love watching people who adore IBM drives tear up Western Digital for their drives, when they're about 50% WD tech and 50% IBM in both cases. Western Digital used to have issues with quality control in their plans in Asia, but that's pretty well worked out at this point. Every WD I've had has lasted until I pitched it, in one case that's 12 years at this point.

"Its like someone opened my mouth and stuck a fistful of herbs in it." - Tamio Kageyama, Iron Chef 'Battle Eggplant'
[ Parent ]
Maxtor vs. Fujitsu, gg Maxtor. (4.50 / 2) (#22)
by Inoshiro on Thu Feb 15, 2001 at 04:32:04 AM EST

I advised Communista to buy Fujitsu. She bought Maxtor. After it crapped out on her, she got a Fujitsu. It's given no trouble. A friend of mine had a 6gb Maxtor drive. It was slow, and seemed to have problems. I encoded all my CDs and stored the Mp3s on the drive. I had it then play them in a loop. I periodically tuned into the icecast stream.

The result was hearable bitrot. The mp3s became more and more corrupted over time. Maxtor drives are not good. They have quality control issues. Fujitsu and IBM are good drives. I've only ever had one Fujitsu act bad on arrival -- it'd been maltreated. The other Fujitsu I had to replace was because I accidently kicked it during operating. A part of the middle of the logical sectors couldn't be accessed (neat whir-click noises resulted). Fujitsu Canada replaced it no questions asked with a new drive. Their warrantees, like their drives, are top notch.

[ イノシロ ]
[ Parent ]
Don't get me started on Maxtor drives (4.50 / 2) (#26)
by burton on Thu Feb 15, 2001 at 10:20:04 AM EST

I'll try to be short and not babble, as I have a really pathetic story that kind of quantifies the hardware end of system instability.

Last year I needed a new hard drive, so I went with 7200 rpm Maxtor 20GB model. It ran fine, but loud, and hot. Really, really loud. I know several people who had the same model, and apparently, mine was oddball.

My father, who is mechanically inclined, said, "that sounds like it has a bad bearing that might cease soon." I'm not mechanically inclined, however, a little common sense obviously gave me a good reason to issue an RMA to get a different drive. His explanation may be incorrect, as he knows nothing about the internals of computer hardware, much less computers at all, but who am I to question somebody with a strong mechanical background, or what I can hear and observe on my own and reference against others? Maybe it was just a cosmetic issue, but I quite frankly did NOT want to play the roulette game considering the sensitivity of the data which would eventually fill that drive.

So, I get around to shipping it back, get my new drive, plug it in, start gently phasing data onto it, and it fails within 30 days, a full hardware failure. Logically I hit up the Maxtor RMA process again, and that is where the story gets interesting. I order a new full tower case w/ a sweet 300 watt ps from Antec, a company whom everybody seems to rant and rave about when it comes to enclosures. This arrives to me the same day as my new replacement drive. I decide, in the interest of being precautious and setting things up right the first time the first thing i'm doing is mounting my mobo in the case, plugging in only the vid card, hooking up kb/m/crt, and plugging in that new drive to format it the way I want it...

I'm glad my friend who was over at the time didn't hold the drive in his hand as I originally intended... Power on, the drive doesn't spin, instead, sparks, and blows a few chips on its underside, and catches fire for a few seconds, as i madly yank the power. Cutting out details of my troubleshooting process, the Antec power supply is bad. So here I am lying though my teeth filling out my 3rd RMA to Maxtor for something not really their fault and cringing and thanking myself that I had the forethought to not hookup any of my other devices to the machine for that initial power up. The board and vid card survived fine, thank god. I am interested to know if anybody has any explanation or has ever encountered something similar.

Now, I have the 4th drive and it clicks rather loudly sometimes when reading data reminiscent of the drive that died on me... but I'm pretty much stuck with it, as I've already seen the (lack of) quality of this product demonstated several times over and what the RMA process yields so pretty much no matter what I do I'm playing Russian roulette. I made a bad product choice and I'm living with the consequences. My really sensitive data stays off it; I now rely mainly on my IBM Deckstars :) (Shameless plug: thou shalt buy a 75GXP)

Otherwise, that machine has been rocksolid; P3450 on a wonderful BX Chipset Asus P2B-F. I am currently slowly migrating to a Tbird 1.1 system that decides to randomly lock up completely every few days during no load whatsoever no doubt due to a hardware issue, which makes me very hesitant until I can track down the exact problem. Step one, get better ram... and the saga continues...

This is just one story about my bad luck with hardware, there are many, many more.

- throughout human history, as our species has faced the frightening, terrorizing fact that we do not know who we are, or where we are going in this ocean of chaos, it has been the authorities... -
[ Parent ]
it's a conspiracy, of course (4.25 / 4) (#10)
by h2odragon on Wed Feb 14, 2001 at 08:56:58 PM EST

This applies mostly to motherboards; hard drives are a different rant. The short version is: buy the second best available, the manufacturing line has had the kinks worked out; anything older is the victim of gradual wear in the manufacturing process and the bottom of the product line always sucks.

Motherboards are the problem, it doesn't matter how good the rest of your components are if the motherboard sucks. Along about 1995 or '96; Intel and the rest of the big boys of the computer industry realized that quality wasn't a big selling point for most of their customers. They wanted cheap, and fast; if it locked up after 24 hours of running because the chipset was poorly designed or implemented it didn't matter -- for the great majority of customers, it would have been rebooted at least twice in that time period because of software problems, anyway.

Additionally, as the internet boom took off, there were a lot of people putting together servers out of commodidity hardware. "They", the evil little gnomes who control everything, seem to have decided that anyone wanting a machine capable of running 24/7 for extended periods of time should be able to pay for the privledge. Consumer level hardware didn't get that much cheaper, but the quality went straight to hell.

I've been building servers and other "must be reliable" machines from eBay sourced parts since the PPro went out of production. I swear by the Tyan Tomcat pentium boards, and the 166{2,8} PPro boards. IMAO the pentium 430HX and PPro 440FX chipsets were the last made that are worthy of use.

To judge by the prices I'm not the only person who values them. 6 months after the Tomcat 4 dual pentium board became unavailable through retail channels, they were being sold on eBay for up to twice what their MSRP had been. They've gotten cheaper since, but if you own one it probably hasn't depreciated much.

Yes, there are explanations for the observed facts that don't require a malicious intent on the part of manufacturers; they could even be right. Arguing "why is it that affordable hardware sucks, and good hardware is expensive" doesn't help much. No matter what conclusion we arrive at; the real question we need to be asking is "How do we get some manufacturer to serve the market for reliable parts at an affordable price?"

Buy reliable components (3.25 / 4) (#14)
by Moneo on Wed Feb 14, 2001 at 09:56:45 PM EST

I haven't bought a computer recently, so I can't speak for modern hardware. However, my experience has been that, particularly with core components, buying from reliable brands makes all the difference (as Ino said). I'd like to add that I prefer building a machine myself to purchasing prebuilt ones -- that way I know about _every_ component that goes into it and don't get shafted with lower grade stuff. (Yes, I know that's impractical when you're deploying 20+ workstations...however, it's fine if you're setting up one, two or even five servers.)

My machine has an Asus P2L97 board and it's run for two years without a problem. When shopping for a CDRW I read a lot of reviews...when I found one I liked (Acer 6296A), I did a google search for "$MODEL problems". I found out it occasionally overheats, so I made sure to give it plenty of room to breathe. I went through a couple of IDE drives until I finally learned that IBM makes the best. Now I only buy drives from them (and make that recommendation to friends, too).

Propaganda plays the same role in a democracy as violence does in a dictatorship. -- Noam Chomsky

It's a combo problem.. (3.75 / 4) (#15)
by Dr Caleb on Wed Feb 14, 2001 at 10:04:09 PM EST

Reliability vs cost + luck.

Semiconductors like anything else that is volume manufactured have error factors built in. When a semiconductor is produced, the overall error is determined. The circular substrate they are produced on is divided into a grid. Statistics show that chips that are formed from the centers of the substrate are more reliable than the chips produced on the edges. All chips are given a basic functionality test and shipped. That's what warranties are for.

In the real world, they follow the bell curve. .1% fail immediately, .1% will never fail, or will fail after X number of years, at a given voltage and temperature.

The 50% or so below the medial line are shipped in home systems. The 50% above, workstations and servers.

Take memory for example. All the working chips from the substrate are shipped. Some companies test and rate the chips, and find the ones that closely match each other, and build reliable, guaranteed memory sicks from them. They will probabally end up in a high end server or workstation. At the other end of the spectrum other companies just throw things together SIMMs and bang them out as fast as they can. Those end up in Radio Shack PC's. ;-)

I know for example, that HP (and other companies) will warranty RAM in their machines if it is HP ram, because HP takes the time to match tolerances of it's individual chips so that the deviation is less than .1%. That makes for reliable RAM.

Sometimes you just luck out. I bought a Slot A Athlon .25 micron process (Aluminum traces), rated for 700MHz. AMD says it's theoretical maximum is 800MHz. I'm running mine at 850 with air cooling. I could get better with liquid or Peltier cooling.

You want reliable, it's rare to get it cheap for new technology. You want performace. You gotta pay.

Plus there is always the way it was assembled. If whoever assembled it wasn't paying attention to static safe practices...you never know when it will fail...

Vive Le Canada - For Canadians who give a shit about their country.

There is no K5 cabal.

My system is very stable (3.50 / 4) (#16)
by NightHawk on Wed Feb 14, 2001 at 11:35:50 PM EST

I built my rig myself. I used an abit mainboard (bh6 first revision) and an intel celeron 300a oem (and despite running at 450mhz it never gives a hickup). My ram is pretty generic, but I do test using SGI's x86memtest to make sure it is 100% error free. As for storage, I currently use an IBM drive (22gxp). I'm a huge fan of IBM drives, they are so great, and just ooze with quality. I also tend to like toshiba cdroms. Plextor is also good.

System stability is all about the components you use. If you get elcheapo aopen parts, your bound to have trouble. If you use a cheesy power supply, you will have trouble. Drivers are also very important! Good hardware with bad drivers is still bad hardware.

I run win2k on my workstation and have seen uptimes well over two weeks before power failure (I plan to invest in a UPS in the near future).

When building a system for yourself, you should always aproach it like restoring a classic car. You want to use the best parts, because it is worth it in the long run, and the extra dollars spent are well justified.

A penny saved on hardware is 10$ in headache in the future.


Fast, Inexpensive, Stable: Pick any two. (2.66 / 3) (#17)
by gblues on Thu Feb 15, 2001 at 02:11:42 AM EST

My opinion jives with the general consensus: if you use quality parts, the PC will be more stable. My personal list:

Motherboards: Asus or Tyan, depending on features/chipset needed.
CPU: AMD or Intel
Memory: Crucial and Samsung.
Video: nVidia cards using reference drivers
Sound: Creative Labs (only player left, really)
SCSI: Adaptec
Network: 3com, though I've had good luck with Linksys products too.

Memory is tricky by itself, because you have a speed rating _and_ a latency rating. I.e. PC100 CAS2 vs. PC133 CAS3. You have to be sure the memory suports the CAS setting your motherboard needs. I.e. PC133 CAS3 memory should not be used in a motherboard that needs PC133 CAS2 (CAS = Cycles to AccesS).

... although in retrospect, having sex to the news was probably doomed to fail from the get-go. --squinky
Nitpicking (5.00 / 1) (#28)
by profesor on Sat Feb 17, 2001 at 09:10:08 PM EST

Technically, CAS stands for 'Column Address Select'. DRAM memory has two address latch signals - RAS (Row Address Select) and CAS. CAS2 or CAS3 refer to the number of clock cycles after CAS that the data will be available on the data bus.

Anyways, gblues is right that you need to pay attention to the latency rating, the acronym was just incorrect.

[ Parent ]
Nitpicking (5.00 / 1) (#29)
by profesor on Sat Feb 17, 2001 at 09:10:09 PM EST

Technically, CAS stands for 'Column Address Select'. DRAM memory has two address latch signals - RAS (Row Address Select) and CAS. CAS2 or CAS3 refer to the number of clock cycles after CAS that the data will be available on the data bus.

Anyways, gblues is right that you need to pay attention to the latency rating, the acronym was just incorrect.

[ Parent ]
My experience with a great PC retailer: (3.50 / 4) (#19)
by klash on Thu Feb 15, 2001 at 02:25:46 AM EST

When I was in the market for a new computer a few years ago, I wanted to find someone who would sell me a solid system, letting me choose exactly the components I wanted. I didn't want to be surprised to find that the video card was integrated on to the motherboard, or that the "16-bit Wavetable sound card" was crap and would refuse to play nice with Linux. The result of my search was pcsforeveryone.com.

Though I have a few minor complaints, all in all I am overwhelmingly impressed with them, and I will probably buy my next computer there. The things I like best about them:

  • They tell you what you actually want to know about the parts they're selling you. On how many retailer's sites will you find comments like "The Duron is also much faster than the Celeron-II for the basic reason of cache size and speed and the AMD's better floating point design," or "For Intel-CPU based computers, almost all the motherboards, if they are capable of taking a 133MHz FSB processor will run the memory bus at 133MHz if you give it PC133 SDIMMs?"
  • Though they will make statements on what they believe to be better parts, they don't sensationalize or try to make you buy things you don't want. Most parts have a link to at least one or two reviews next to them, and they'll be perfectly frank about stuff that sucks. About the ATI RAGE PRO, they write "Can process about 1.2 million triangles per second (which is only so-so these days) for 30fps+ gaming performance (albeit at a fairly low resolution). Fairly mediocre graphics performance for Pentium and Athlon systems..." This is an extreme plus in my book--I feel like I can browse their lists of parts and know exactly what I'm getting into.
  • Lifetime labor and 1 year parts warranty. And popping the hood doesn't void it.
  • The site is updated very frequently.
  • Many other reasons listed here.

This is the kind of company I'm eager to give my business, and I personally wish we saw more of this kind of company around.

PCFE is great...but only if you are in Boston! (4.00 / 1) (#20)
by ephibian on Thu Feb 15, 2001 at 03:47:39 AM EST

I'd like to second this opinion -- without question, PCFE is the best computer retailer that I know. Extremely knowledgeable people, and quality hardware. Unfortunately, they do not sell over the Internet, and are located in Cambridge MA. Unless you are in the Boston area, you are outta luck...

[ Parent ]
not true--I don't live in Boston! (4.50 / 2) (#21)
by klash on Thu Feb 15, 2001 at 04:09:30 AM EST

On this page, they say that they'll accept orders over the phone or over e-mail. They have some restrictions on the conditions under which they'll accept a credit card, but under those conditions a check will still be just fine.

I lived in Idaho when I bought from them, so I can personally attest to this. The shipping was kind of expensive, but the 2% discount for the cashier's check helped offset it some.

[ Parent ]
Failures are 99% software related - IMHO (4.50 / 2) (#23)
by NKJensen on Thu Feb 15, 2001 at 06:56:50 AM EST

I've been chasing a lot of bugs in PC hardware. And found most of them to be software problems.

More often than not, upgrading the drivers for the hardware solves the instability problems.

Writing low level driver software is not an easy task. I've been doing this since 1989.

You just have to pay attention to even the most miniscule details of the hardware. Interrupt timing. Multiple access to shared memory. DMA's running when they are not supposed to. New OS versions require you to rework old drivers - release date was yesterday, you know.

The hardware vendors often are in a rush to get the stuff to market and therefore ship crappy drivers - knowing that the customer can download a new driver, once he starts to notice the problems.

Reliable hardware exists, you just would not like to pay for it. Some millitary boards will continue to work even with bullet holes. The same techniques can be applied to any piece of hardware - at a certain cost.
From Denmark. I like it, I live there. France is another great place.

Why PCs crash and Mainframes Don't (4.33 / 3) (#24)
by wiredog on Thu Feb 15, 2001 at 08:28:47 AM EST

Was the cover story in the April 1998 Byte. Pretty much covers everything that's been posted here, and answers the question.

The idea of a global village is wrong, it's more like a gazillion pub bars.

You might want to check (4.00 / 2) (#25)
by wiredog on Thu Feb 15, 2001 at 09:50:40 AM EST

Over the past few weeks, it seems like every computer that I have touched has suffered from major or minor hardware problems.

Had that problem a couple of years ago. Then I started grounding myself before I touched the equipment. The problem went away. Static can be much more of a problem in the winter due to low humidity.

The idea of a global village is wrong, it's more like a gazillion pub bars.

Even more insideous is the time delay.. (none / 0) (#30)
by Zukov on Wed Mar 14, 2001 at 12:59:32 PM EST

I have heard that some static discharge will damage a chip in a way that makes it fail many months later, not right away. By then, of course, you have forgotten all about having worked on the machine to begin with..

ȶ H (^

Yes, I have just bumbled upon Gnome Character Map. Please ! me.
[ Parent ]

The computers ARE getting less reliable... (5.00 / 2) (#27)
by Armaphine on Thu Feb 15, 2001 at 03:35:44 PM EST

...and I doubt that the software is at fault as we think.

I work in a coal-fire power plant, and we have very fine coal dust EVERYWHERE. Now, we have a bunch of older Dell Dimension from circa 1995 still around the plant that have got coal dust, normal dust, floor wax, and God only knows what else in them, to the point of looking like Hiroshima if you try blowing out the power supply. Now think about this: These things have been, for the most part, stored in abysmal conditions, abused beyond reasonable usage, and have temperature shock taken to them numerous times. And they still work six years later.

Contrast this with some of the newer machines, which have had power arcs and blown themselves out after two years. And the same applies to the monitors. Now given, a power plant is not exactly going to be the friendliest place for a monitor, what with turbines generating hundreds of megawatts less than two hundred feet away. But the old ones held up well. After six years, some of them have one of the electron guns go bad, or start getting a little fuzzy around the edges, but the newer ones do this after two. Case in point, finally had an older monitor die on me after eight years in one of the shops. When I opened up the casing, I had over a pound of coal dust empty out of it...

Personally, I believe that the problem lies with the ever-shrinking circutry... if that pathway is getting very thin, then breaking it with a stray bit of static becomes much easier.

Question authority. Don't ask why, just do it.

Building High Reliability Computers | 30 comments (30 topical, 0 editorial, 0 hidden)
Display: Sort:


All trademarks and copyrights on this page are owned by their respective companies. The Rest 2000 - Present Kuro5hin.org Inc.
See our legalese page for copyright policies. Please also read our Privacy Policy.
Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
Need some help? Email help@kuro5hin.org.
My heart's the long stairs.

Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!