mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware

Reply
 
Thread Tools
Old 2007-11-27, 09:01   #1
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

169D16 Posts
Default Are commercial PC's becoming more reliable?

Can the data collected on failed LL tests be used to find a trend on failure rates of tests over the last few years?

My last two computers, bought ~4 years and ~1 year ago respectively, both have never had a detected fault in any test. But previous PC's I've had did encounter test failures and general unreliability over the long term.

So, are board and chip designs (and packaging) becoming inherently more reliable?

Why doesn't the standard off-the-shelf PC have ECC RAM? Is it only because of price considerations?
retina is online now   Reply With Quote
Old 2007-11-27, 17:20   #2
jasonp
Tribal Bullet
 
jasonp's Avatar
 
Oct 2004

3·1,163 Posts
Default

Quote:
Originally Posted by retina View Post
Why doesn't the standard off-the-shelf PC have ECC RAM? Is it only because of price considerations?
Because manufacturers do not perceive a demand for it, and so can save a few pennies laying out motherboards and a few more pennies leaving off the ninth DRAM chip on a DRAM stick
jasonp is offline   Reply With Quote
Old 2007-11-28, 09:34   #3
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

7×827 Posts
Default

Quote:
Originally Posted by jasonp
Because manufacturers do not perceive a demand for it, and so can save a few pennies laying out motherboards and a few more pennies leaving off the ninth DRAM chip on a DRAM stick
So it is simply to save cost then. But I don't know about the difference being only a few pennies, the MB's and RAM's with ECC capability seem to cost a lot more than your explanation would suggest.
retina is online now   Reply With Quote
Old 2007-11-28, 16:15   #4
jasonp
Tribal Bullet
 
jasonp's Avatar
 
Oct 2004

1101101000012 Posts
Default

Quote:
Originally Posted by retina View Post
So it is simply to save cost then. But I don't know about the difference being only a few pennies, the MB's and RAM's with ECC capability seem to cost a lot more than your explanation would suggest.
The last time I purchased an Opteron system, the only motherboards that supported ECC memory (even though AMD's chipset always supported ECC) were for workstations and servers, and this is a demographic that's used to paying more for hardware, so the manufacturers oblige them :) Also, the components are made in smaller volume and have less economy of scale.

ECC memory is actually not much more expensive than regular memory now; the RAM chips themselves are identical, the DIMM just contains one or two more of them to store the ECC bits and its circuit board contains a few more wires to publish the ECC bits. Even a few years ago the extra cost for ECC memory amounted to about $5 per stick. Motherboards that claim to support ECC are targeted at markets that don't care about the huge list of features in consumer motherboards, and so don't sell as many units to consumers, and so have less economy of scale as well.
jasonp is offline   Reply With Quote
Old 2007-11-28, 22:20   #5
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

22×5×13×17 Posts
Default

I could supply evidence to the contrary.
My first PC, a 1983 c-64 was very reliable:
- never saw a 'blue screen of death'
- most lock-ups were my fault (cutesy programming) ... yes, I am man enough to admit it

However, if you are specifically talking about Microsoft/Intel/AMD,then yes they are getting better from my point of view.
petrw1 is offline   Reply With Quote
Old 2007-11-29, 21:07   #6
jasonp
Tribal Bullet
 
jasonp's Avatar
 
Oct 2004

3×1,163 Posts
Default

Quote:
Originally Posted by petrw1 View Post
I could supply evidence to the contrary.
My first PC, a 1983 c-64 was very reliable:
- never saw a 'blue screen of death'
- most lock-ups were my fault (cutesy programming) ... yes, I am man enough to admit it

However, if you are specifically talking about Microsoft/Intel/AMD,then yes they are getting better from my point of view.
When everybody programmed on the bare metal, you never saw corrupted memory, because there was no indication it happened (unless you corrupted video memory :).

OTOH, a modern CPU+memory has several thousand times as many transistors as an old CPU+memory, each thousands of times faster, hundreds of times smaller and dozens of times lower-power compared to their old counterparts. If anything, that makes a modern CPU more prone to cosmic ray bit-flips and such. Still, the crashes you see would be overwhelmingly more likely to be caused by drivers and application software than by any intrinsic hardware instability.

I used to have a win98 system until about this year, which would BSOD maybe once a year. The system that replaced it runs winXP Pro, and the only BSODs I've ever seen from it were directly attributable to the network driver from a software firewall I use.

Regarding the thread starter's question, to use GIMPS failure rate data to get an idea of hardware reliability we'd also need to know how many GIMPS computers are overclocked or otherwise running outside their specifications; I have a feeling the percentage is much higher than the computer-using background percentage :)

Last fiddled with by jasonp on 2007-11-29 at 21:10
jasonp is offline   Reply With Quote
Old 2007-12-04, 14:47   #7
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

7·827 Posts
Default

Quote:
Originally Posted by jasonp View Post
to use GIMPS failure rate data to get an idea of hardware reliability we'd also need to know how many GIMPS computers are overclocked or otherwise running outside their specifications; I have a feeling the percentage is much higher than the computer-using background percentage
Yeah, you are right about that. But perhaps we could make an assumption that the ratio of overclockers/standardclockers today and in the yesteryears is approximately the same. Then the results could still be useful.
retina is online now   Reply With Quote
Old 2007-12-04, 22:43   #8
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

5·701 Posts
Default

Quote:
Originally Posted by retina View Post
Yeah, you are right about that. But perhaps we could make an assumption that the ratio of overclockers/standardclockers today and in the yesteryears is approximately the same. Then the results could still be useful.
You're assuming that's a valid assumption, when it might not be. Think of all the people who are crunching on 2-3 year old computers who, up to that point, were upgrading way more frequently in the decade before. The whole situation is changing radically, Moore's Law effects aren't as important to some people as they once were.
jasong is offline   Reply With Quote
Old 2007-12-05, 05:03   #9
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

10110100111012 Posts
Default

Quote:
Originally Posted by jasong View Post
You're assuming that's a valid assumption ...
Well, of course, that's what an assumption is! You're assuming that I'm assuming that it's a valid assumption. How far do we go with this?
Quote:
Originally Posted by jasong View Post
... when it might not be. Think of all the people who are crunching on 2-3 year old computers who, up to that point, were upgrading way more frequently in the decade before. The whole situation is changing radically, Moore's Law effects aren't as important to some people as they once were.
Not sure what you mean there. Yes, the situation is changing, but what can be done about it? The idea being to make some assumptions and see what the results are. If it all turns to crap then so what, we learned that our assumptions were no good. If it turns out to look like we found something interesting then we learned, well, something interesting. Either way we learn something.
retina is online now   Reply With Quote
Old 2007-12-06, 03:53   #10
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

5×701 Posts
Default

Quote:
Originally Posted by retina View Post
Well, of course, that's what an assumption is! You're assuming that I'm assuming that it's a valid assumption. How far do we go with this?
I really made an ass of myself last night, not sure what was going on. Maybe an off day, maybe I need a good slap, I don't know. It could also be that I've done this in the past and was just too stupid to notice, which may mean my character's improved slightly.(hopefully) Nothing to be proud of, I'm just trying to approach this honestly. :)

Last fiddled with by jasong on 2007-12-06 at 03:53
jasong is offline   Reply With Quote
Old 2007-12-06, 10:34   #11
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

169D16 Posts
Default

I have found there is little been said in the market about the expected reliability of basic off-the-shelf PC's. Microsoft claim that their "Please report this problem to Microsoft" dialog box thingy shows a significant number of computers have memory corruption which they attribute to memory (i.e. hardware) failures. But they fall short of actually giving a percentage, let alone some sort of long term trend.

I kind of get the impression that most people now days fully expect their PC to misbehave and don't much care that they have to reboot it on a regular basis. Quite a shame really. If more people demanded better reliability then the ECC capable stuff might become standard (and cheaper). That would also have the side effect of giving OS makers no more wiggle room when their OS crashes.

@jasong: Please don't hijack my thread with talk of your psychological difficulties.
retina is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Getting reliable LL from unreliable hardware preda GPU Computing 293 2020-03-02 13:30
Call for reliable LL and DC workers... chalsall GPU to 72 69 2012-03-30 22:18
CPU is not reliable enough to get *blah* willmore Software 2 2009-01-15 19:33
Own Project / Commercial License friedo Factoring 23 2006-06-29 06:40
Most Reliable Memory SLalley Hardware 10 2004-04-22 21:39

All times are UTC. The time now is 03:29.

Thu Oct 22 03:29:21 UTC 2020 up 42 days, 40 mins, 0 users, load averages: 1.69, 1.72, 1.79

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.