Tom's Hardware is hilarious

[e]

What does running a good memory test at boot have to do with anything?

Transistors can burn out at any time.


"Bill Davidsen" <davidsen@darkstar.prodigy.com> wrote in message
news:9wJCd.5023$8U5.1515@newssvr31.news.prodigy.com... Yousuf Khan wrote: George Macdonald wrote: On Sun, 26 Dec 2004 03:19:02 GMT, Johannes H Andersen <johs@sizefitter.com> wrote:> I run a Northwood 2.8 P4 overclocked to 3GHz with a Tagan 380W PSU. CPU> cooler> is a Zalman duct HS fan, speed is 2000 rpm. Prime95 no problems> whatsoever. Huh? You overclock? I thought your work was "valuable"? Don't tell me this is a new approach to Stochastic Programming!:-) Let me guess, he was one of those "my work is too important and valuable to trust to AMD processors" types? %^) A fair bit of my work is too I&V to run on non-ECC memory. Server boards and memory add about $200 to system price, typically. However, memory is far better than it was a few years ago, and running a good memory test at boot is enough for non-critical use in many cases. -- bill davidsen (davidsen@darkstar.prodigy.com) SBC/Prodigy Yorktown Heights NY data center Project Leader, USENET news http://newsgroups.news.prodigy.com

[George Macdonald]

On Wed, 5 Jan 2005 20:46:39 -0800, "e" <DoNotRespond@fake.net> wrote:
What does running a good memory test at boot have to do with anything?

I dunno whether he meant build time but running memory stress tests is an
essential part of a system build; with a repeat of this at regular
intervals it gives some confidence level. Given a reasonable operating
environment, I've rarely seen any memory chips/modules degrade with time &
use... and certainly not in the past 10 years or so.
Transistors can burn out at any time.

Detecting that is not the main purpose of ECC memory though.
"Bill Davidsen" <davidsen@darkstar.prodigy.com> wrote in messagenews:9wJCd.5023$8U5.1515@newssvr31.news.prodigy.com...
A fair bit of my work is too I&V to run on non-ECC memory. Server boards and memory add about $200 to system price, typically. However, memory is far better than it was a few years ago, and running a good memory test at boot is enough for non-critical use in many cases. -- bill davidsen (davidsen@darkstar.prodigy.com) SBC/Prodigy Yorktown Heights NY data center Project Leader, USENET news http://newsgroups.news.prodigy.com

--
Rgds, George Macdonald

[Bill Davidsen]

e wrote: What does running a good memory test at boot have to do with anything? Transistors can burn out at any time.

While that statement is true, the implication that failure is equally
likely at all times isn't. Like lightbulbs, failure on power cycle is
more likely because the system is subject to both electrical and thermal
transients not present at other times. So unless you run a memory test
all the time (you could think of ECC that way), right after power on is
a good time to do the test.

And I'm told that memory failure in dynamic memory is usually a
breakdown in the little capacitors, not the transistor. Random failures
are often one-time alpha particle hits, modern memory is far better in
that regard as well, due to smaller structure and better substrates.

Nothing is perfect, the question is how reliable can you afford to be?
"Bill Davidsen" <davidsen@darkstar.prodigy.com> wrote in message news:9wJCd.5023$8U5.1515@newssvr31.news.prodigy.com...Yousuf Khan wrote:George Macdonald wrote:>On Sun, 26 Dec 2004 03:19:02 GMT, Johannes H Andersen><johs@sizefitter.com>>wrote:>>>>I run a Northwood 2.8 P4 overclocked to 3GHz with a Tagan 380W PSU. CPU>>cooler>>is a Zalman duct HS fan, speed is 2000 rpm. Prime95 no problems>>whatsoever.>>>>Huh? You overclock? I thought your work was "valuable"? Don't tell me>this is a new approach to Stochastic Programming!:-)Let me guess, he was one of those "my work is too important and valuableto trust to AMD processors" types? %^)A fair bit of my work is too I&V to run on non-ECC memory. Server boardsand memory add about $200 to system price, typically. However, memory isfar better than it was a few years ago, and running a good memory test atboot is enough for non-critical use in many cases.


--
bill davidsen (davidsen@darkstar.prodigy.com)
SBC/Prodigy Yorktown Heights NY data center
Project Leader, USENET news
http://newsgroups.news.prodigy.com

[Robert Redelmeier]

In comp.sys.ibm.pc.hardware.chips e <DoNotRespond@fake.net> top-posted: Transistors can burn out at any time.

They're _not_ lightbulbs. Look up the bathtub failure curve.

Another datum for the correlation: top-poster=ignoramus.

-- Robert

[keith]

On Thu, 06 Jan 2005 15:59:14 +0000, Robert Redelmeier wrote:
In comp.sys.ibm.pc.hardware.chips e <DoNotRespond@fake.net> top-posted: Transistors can burn out at any time. They're _not_ lightbulbs. Look up the bathtub failure curve.

Well, to be fair... Lightbulbs fail in a bathtub fashion too. They fail
even faster in a bathtub. ;-)
Another datum for the correlation: top-poster=ignoramus.

Ignramus, perhaps. Self-centered asshole, certainly. One writes
for the reader, not the writer.

--
Keith

[Robert Redelmeier]

In comp.sys.ibm.pc.hardware.chips keith <krw@att.bizzzz> wrote: Well, to be fair... Lightbulbs fail in a bathtub fashion too. They fail even faster in a bathtub. ;-)

Doh! <smacks forehead> Quite correct. I guess the phrase
"transistor burnout" just makes me see red. I think most
electronics fails from electrolytic capacitor leak/dry-out.

We have some 486s and RS4000s that have been in continuous
use for 15 years. What is that, 125,000 hrs? You're lucky
to get 1-2000 hr from a lightbulb.
Another datum for the correlation: top-poster=ignoramus.
Ignramus, perhaps. Self-centered asshole, certainly. One writes for the reader, not the writer.

A profound insight, certainly valid excepting the egotist.

I was pushing a little further, correlating egotism with
ignorance. Not as certain, but probable.

-- Robert

[keith]

On Fri, 07 Jan 2005 03:03:31 +0000, Robert Redelmeier wrote:
In comp.sys.ibm.pc.hardware.chips keith <krw@att.bizzzz> wrote: Well, to be fair... Lightbulbs fail in a bathtub fashion too. They fail even faster in a bathtub. ;-) Doh! <smacks forehead> Quite correct. I guess the phrase "transistor burnout" just makes me see red. I think most electronics fails from electrolytic capacitor leak/dry-out. We have some 486s and RS4000s that have been in continuous use for 15 years. What is that, 125,000 hrs? You're lucky to get 1-2000 hr from a lightbulb.

Ok, but one's forced to ask; What's an RS4000?
Another datum for the correlation: top-poster=ignoramus. Ignramus, perhaps. Self-centered asshole, certainly. One writes for the reader, not the writer. A profound insight, certainly valid excepting the egotist. I was pushing a little further, correlating egotism with ignorance. Not as certain, but probable.

That works. Its surprising how the more one learns, the less one knows.

--
Keith

[daytripper]

On Thu, 06 Jan 2005 22:33:25 -0500, keith <krw@att.bizzzz> wrote:
On Fri, 07 Jan 2005 03:03:31 +0000, Robert Redelmeier wrote: In comp.sys.ibm.pc.hardware.chips keith <krw@att.bizzzz> wrote: Well, to be fair... Lightbulbs fail in a bathtub fashion too. They fail even faster in a bathtub. ;-) Doh! <smacks forehead> Quite correct. I guess the phrase "transistor burnout" just makes me see red. I think most electronics fails from electrolytic capacitor leak/dry-out. We have some 486s and RS4000s that have been in continuous use for 15 years. What is that, 125,000 hrs? You're lucky to get 1-2000 hr from a lightbulb.Ok, but one's forced to ask; What's an RS4000?

A now elderly MIPS RISC system, perhaps?

/daytripper

[chrisv]

keith wrote:
We have some 486s and RS4000s that have been in continuous use for 15 years. What is that, 125,000 hrs? You're lucky to get 1-2000 hr from a lightbulb.Ok, but one's forced to ask; What's an RS4000?

I think it's on old IBM Power-based computer, usually running AIX.

[Nate C.]

Robert
In fact the life of a light bulb depends on how it is built, e.g., standard, rough service, traffic signal, photo flood, etc., and how it is used, e.g., on/off cycle, relationship of lamp voltage rating to lamp supply voltage, environmental conditions, etc. In unusual conditions such as a 135 volt lamp burning continuously at 115 to 120 volts several years is not unusual. In more extreme circumstances such as a 230 volt lamp operated at 125 volts for use as an indicator (as opposed to providing illumination) with a limited number (less than 5) of on/off cycles per year the life is in the twenty to thirty year range. In contrast a photo flood lamp under normal use is rated in tens of hours.
However I agree with your 1-2000 for a standard 120 volt lamp in normal household use.
Stuart

"Robert Redelmeier" <redelm@ev1.net.invalid> wrote in message news:7_mDd.11954$wi2.658@newssvr11.news.prodigy.com... In comp.sys.ibm.pc.hardware.chips keith <krw@att.bizzzz> wrote: Well, to be fair... Lightbulbs fail in a bathtub fashion too. They fail even faster in a bathtub. ;-) Doh! <smacks forehead> Quite correct. I guess the phrase "transistor burnout" just makes me see red. I think most electronics fails from electrolytic capacitor leak/dry-out. We have some 486s and RS4000s that have been in continuous use for 15 years. What is that, 125,000 hrs? You're lucky to get 1-2000 hr from a lightbulb. Another datum for the correlation: top-poster=ignoramus. Ignramus, perhaps. Self-centered asshole, certainly. One writes for the reader, not the writer. A profound insight, certainly valid excepting the egotist. I was pushing a little further, correlating egotism with ignorance. Not as certain, but probable. -- Robert

[Robert Redelmeier]

In comp.sys.ibm.pc.hardware.chips keith <krw@att.bizzzz> wrote: Ok, but one's forced to ask; What's an RS4000?

Sorry, RS6000 - 60 MHz Power2
That works. Its surprising how the more one learns, the less one knows.

Precisely true, and a mini-intelligence test.

-- Robert

[Robert Redelmeier]

In comp.sys.ibm.pc.hardware.chips Stuart <xyz@hotmail.com> wrote: In fact the life of a light bulb depends

Of course.
In unusual conditions such as a 135 volt lamp burning continuously at 115 to 120 volts several years is not unusual.

20,000 hrs? I would have thought tungsten sublimation
would have killed a std 135V/2000h bulb long before then.
Unusual conditions like halogen fill for redeposition?
In more extreme circumstances such as a 230 volt lamp operated at 125 volts for use as an indicator (as opposed to providing illumination) with a limited number (less than 5) of on/off cycles per year the life is in the twenty to thirty year range.

I've run Euro bulbs. They hardly give off enough light at
115V to qualify as indicators except in darkened surroundings.

I didn't think cycles mattered that much on tungsten-filament
bulbs with good design for thermal expansion & filament
support. Las Vegas signage certainly gets hammered.
In contrast a photo flood lamp under normal use is rated in tens of hours.

Of course. So do Japanese 100V bulbs at 115V. Voltage ratings
are a bit of a farce. Filaments have a resistance curve and
reach a particular temperature at a given voltage.

-- Robert

[keith]

On Fri, 07 Jan 2005 19:24:49 +0000, Robert Redelmeier wrote:
In comp.sys.ibm.pc.hardware.chips keith <krw@att.bizzzz> wrote: Ok, but one's forced to ask; What's an RS4000? Sorry, RS6000 - 60 MHz Power2

Ok. ;-) ...but I didn't think the Power2 was quite *that* old. Dunno,
before "my" time. ;-)

That works. Its surprising how the more one learns, the less one knows. Precisely true, and a mini-intelligence test.

OTOH, don't confuse mastery of trivia for "learning". ...either way.

--
Keith

[Robert Redelmeier]

In comp.sys.ibm.pc.hardware.chips keith <krw@att.bizzzz> wrote: OTOH, don't confuse mastery of trivia for "learning". ...either way.

"Never confuse ignorance with stupidity.
The former can be cured.
The latter is forever."

-- Robert

[Joe Pfeiffer]

keith <krw@att.bizzzz> writes:
On Fri, 07 Jan 2005 19:24:49 +0000, Robert Redelmeier wrote: In comp.sys.ibm.pc.hardware.chips keith <krw@att.bizzzz> wrote: Ok, but one's forced to ask; What's an RS4000? Sorry, RS6000 - 60 MHz Power2 Ok. ;-) ...but I didn't think the Power2 was quite *that* old. Dunno, before "my" time. ;-)

Thre may have been an RS6000 based on Power2, but I never heard about
it -- the original RS6000 was a POWER.
--
Joseph J. Pfeiffer, Jr., Ph.D. Phone -- (505) 646-1605
Department of Computer Science FAX -- (505) 646-1002
New Mexico State University http://www.cs.nmsu.edu/~pfeiffer

[keith]

On Fri, 07 Jan 2005 23:25:55 -0700, Joe Pfeiffer wrote:
keith <krw@att.bizzzz> writes: On Fri, 07 Jan 2005 19:24:49 +0000, Robert Redelmeier wrote: In comp.sys.ibm.pc.hardware.chips keith <krw@att.bizzzz> wrote:> Ok, but one's forced to ask; What's an RS4000? Sorry, RS6000 - 60 MHz Power2 Ok. ;-) ...but I didn't think the Power2 was quite *that* old. Dunno, before "my" time. ;-) Thre may have been an RS6000 based on Power2, but I never heard about it -- the original RS6000 was a POWER.

Sure. I had a 590 under my desk in '94.
http://www-1.ibm.com/servers/eserver/pseries/hardware/whitepapers/power2_models.html

--
Keith

[Nate C.]

In line comments.
Stuart

"Robert Redelmeier" <redelm@ev1.net.invalid> wrote in message news:rGBDd.12106$wi2.11128@newssvr11.news.prodigy.com... In comp.sys.ibm.pc.hardware.chips Stuart <xyz@hotmail.com> wrote: In fact the life of a light bulb depends Of course. In unusual conditions such as a 135 volt lamp burning continuously at 115 to 120 volts several years is not unusual. 20,000 hrs? I would have thought tungsten sublimation would have killed a std 135V/2000h bulb long before then. Unusual conditions like halogen fill for redeposition?
Just standard US lamps except for the 135 volt.
In the cases I was referring to they are in difficult places to reach and are not turned off. In more extreme circumstances such as a 230 volt lamp operated at 125 volts for use as an indicator (as opposed to providing illumination) with a limited number (less than 5) of on/off cycles per year the life is in the twenty to thirty year range. I've run Euro bulbs. They hardly give off enough light at 115V to qualify as indicators except in darkened surroundings.

I can only say that my former employer used/uses them in hundreds of installations and they do the job. I didn't think cycles mattered that much on tungsten-filament bulbs with good design for thermal expansion & filament support. Las Vegas signage certainly gets hammered.

In a standard lamp operated at its design voltage lamp on/off cycles are a major factor in lamp failure.
You may have noticed that most lamps fail when you turn them off.
As to Las Vegas I have no personal knowledge but I do know that lamps used for traffic signals are of special construction.
They have limitations such as "Burn base down to horizontal" and have, relatively, a low light output, i.e., a 100 Watt traffic signal bulb lamp delivers a light output approximately equivalent to a 60-75 watt normal lamp. As a side note, the local jurisdictions are on a program to replace the lamps in traffic signals with LED arrays. They state the savings in lamp replacement labor costs and energy costs will offset the higher original cost.
In contrast a photo flood lamp under normal use is rated in tens of hours. Of course. So do Japanese 100V bulbs at 115V. Voltage ratings are a bit of a farce. Filaments have a resistance curve and reach a particular temperature at a given voltage.

Voltage ratings are the result of a design tradeoff.
For a given physical lamp if you lower the voltage you will lower the light output and the efficiency and increase the light output because, as you state, the filament is at a lower temperature.
Conversely, if you raise the voltage the temperature increases, the light output and efficiency increase and the life decreases.
The voltage rating assigned to the physical lamp is, more or less, a marketing decision balancing the cost of manufacture, the cost to produce a lumen and customer expectations as to light output, life, and purchase price.

-- Robert

[Robert Redelmeier]

In comp.sys.ibm.pc.hardware.chips Stuart <xyz@hotmail.com> wrote: have noticed that most lamps fail when you turn them off.

I see more failures at start-up -- a short brief flash
when turned on.
As to Las Vegas I have no personal knowledge but I do know that lamps used for traffic signals are of special construction.

I would expect so! Very expensive bulbs to replace, often
subject to high vibrations when lit (only very short, stiff
lightpoles might be immune).
They have limitations such as "Burn base down to horizontal" and have, relatively, a low light output,

The yellows have higher output because they have a
shorter required life. Many places, there are two reds.

I believe the prohibition on "base up" is to reduce the area
exposed to cold (untreated?) glass or to reduce obscuring
from tungsten deposition.
program to replace the lamps in traffic signals with LED arrays. They state the savings in lamp replacement labor costs and energy costs will offset the higher original cost.

Likely a very good move. Some trucks also have LED tail-lamps.
The voltage rating assigned to the physical lamp is, more or less, a marketing decision

Yes, to meet cost, life, lumen & color specs.
No problem over/underclocking :)

-- Robert

[keith]

On Sun, 09 Jan 2005 18:30:46 +0000, Robert Redelmeier wrote:
In comp.sys.ibm.pc.hardware.chips Stuart <xyz@hotmail.com> wrote: have noticed that most lamps fail when you turn them off. I see more failures at start-up -- a short brief flash when turned on.

That's when you see the failure, but power cycles aren't really what does
them in. Tungsten boils off the filament, over time, making it weaker.
As to Las Vegas I have no personal knowledge but I do know that lamps used for traffic signals are of special construction. I would expect so! Very expensive bulbs to replace, often subject to high vibrations when lit (only very short, stiff lightpoles might be immune).

Most traffic lights are LEDs these days. They are expen$ive, but last
longer, are more efficient, and have a higher apparent brightness.
They're better and the TCO is lower.
They have limitations such as "Burn base down to horizontal" and have, relatively, a low light output, The yellows have higher output because they have a shorter required life. Many places, there are two reds.

There are *hundreds* of reds. ;-) The reds were the first to change over
to LEDs.
I believe the prohibition on "base up" is to reduce the area exposed to cold (untreated?) glass or to reduce obscuring from tungsten deposition. program to replace the lamps in traffic signals with LED arrays. They state the savings in lamp replacement labor costs and energy costs will offset the higher original cost. Likely a very good move. Some trucks also have LED tail-lamps.

They've been using LEDs in traffic lights for ten years of so. ...and I
thought I lived in the sticks. ;-)
The voltage rating assigned to the physical lamp is, more or less, a marketing decision Yes, to meet cost, life, lumen & color specs. No problem over/underclocking :)

--
Keith

[Nate C.]

Bob & Keith
My fingers did type what my mind said!
As you both said most lamp failures are at lamp turn ON not OFF.
This, according to what I was told by GE engineers, is primarily due to the high initial current and the thermal shock when the cold filament is energized.
This is the same thing you see with vacuum tubes and is why computers using vacuum tubes were left on 24/7.
My point was, if the lamp is on continuously (no off/on cycles) it lasts until the filament actually boils away and fails mechanically.
Stuart


"Robert Redelmeier" <redelm@ev1.net.invalid> wrote in message news:qLeEd.12594$wi2.11948@newssvr11.news.prodigy.com... In comp.sys.ibm.pc.hardware.chips Stuart <xyz@hotmail.com> wrote: have noticed that most lamps fail when you turn them off. I see more failures at start-up -- a short brief flash when turned on. As to Las Vegas I have no personal knowledge but I do know that lamps used for traffic signals are of special construction. I would expect so! Very expensive bulbs to replace, often subject to high vibrations when lit (only very short, stiff lightpoles might be immune). They have limitations such as "Burn base down to horizontal" and have, relatively, a low light output, The yellows have higher output because they have a shorter required life. Many places, there are two reds. I believe the prohibition on "base up" is to reduce the area exposed to cold (untreated?) glass or to reduce obscuring from tungsten deposition. program to replace the lamps in traffic signals with LED arrays. They state the savings in lamp replacement labor costs and energy costs will offset the higher original cost. Likely a very good move. Some trucks also have LED tail-lamps. The voltage rating assigned to the physical lamp is, more or less, a marketing decision Yes, to meet cost, life, lumen & color specs. No problem over/underclocking :) -- Robert

[keith]

On Mon, 10 Jan 2005 14:15:26 +0000, Stuart wrote:
Bob & Keith My fingers did type what my mind said!

I don't think there was any doubt there...
As you both said most lamp failures are at lamp turn ON not OFF.

Sure, the difference of opinion was *why* the failures. On/Off cycles or
hours. My understanding is that the hours boil off the Tungsten and the
thermal stress of the power-on does the bulb in. That is, the total hours
in use is the real killer, not the power cycles (forgeting vibration and
other abuse). Certainly different bulbs are designed for diffeent
environments, but I trust we're all talking about the typical "household
60W bulb" as the standard.
This, according to what I was told by GE engineers, is primarily due to the high initial current and the thermal shock when the cold filament is energized. This is the same thing you see with vacuum tubes and is why computers using vacuum tubes were left on 24/7. My point was, if the lamp is on continuously (no off/on cycles) it lasts until the filament actually boils away and fails mechanically. Stuart

That's where "your people" and "mine" disagree. <shrug>

--
Keith