Temps that shorten drive life

[Arno Wagner]

Arno Wagner wrote:

one thought for you: i suspect you have
a case hot spot where your video card is
coz it shouldn't be running at 80C
(unless you are overclocking it)

No, no, my card is fine. I was experimenting here with
a new cooler and actually turned the fan off to see
whether the 105C had any validity...

Arno

put your hand on the case and if it is
noticeably warm then you have a hot spot
(with poor air flow at that point);
assuming you've got positive (or negative)
air pressure within the case, remove
the slot plate immediately below the
video card and you'll remove the hotspot


i simply don't know

what amazes me is that nVidia seems to
be ok with running their GPUs at *much*
higher temps (at least 20+C higher) than
anything i've ever seen for any modern CPU

 

[-09]

My drives are rated for 5C to 55C; I just asked Western Digital if there's a
temperature within the spec range at which their drive would experience a
shorter life. For instance, running in the 40's instead of the 30's. They
said this won't happen until above 55C. But I wonder if anyone here knows
differently? TIA.

Using an antec SLK3000-B case with a 120 mm fan blowing air in on
the hard drive cage orientated so the air blows front to back
across the drives not side to side, my two WD drives are ~31 idle
and 33, 34 max load per their smart temp diodes and speedfan. End
of discussion.

I picked up the case $54 local.

http://www.newegg.com/Product/Product.asp?Item=N82E16811129152

 

[Guest]

(e-mail address removed) wrote:

Semiconductors have about 20-30 years lifetime at 25C.


Well, depends. Actually it is difficult to get current information.
The figures are from Intel's memory components Handbook from 10-15
years back.


As I said, the figures are from a memory handbook from way back.
Of course different semiconductors have different characteristics.
But don't expect 30 years at 200C. You will not get it.


Does not matter. This is not a hard phase change. 10C below maximum
you get nearly the same behaviour as at maximum.

Ok, power semiconductors are different. My numbers are for CMOS.
Power semiconductors have larger structures and are therefore far
less succeptible to aging. They usually die by melting. For small
scale cmos, this is different.

The shorter estimated lifespans are generally for power semiconductors,
especially bipolars, where damage occurs right at the P-N junction.
Estimates for dense ICs seem to be longer, in the range of 50-200
years, but I don't know if this includes 3 GHz processors
(electromigration?).

Today you often get something like 5 years at 80C or so, if you get any
information at all. The 30 year point would then be around 50C.


What please is "ultra reliable equipment"?


And lower power allowances. The trade-off done here for non-cooled
chips is that e.g. military TTL is slower, and can tolerate
higher ambient temperatures.

Electrolythes age differently than semiconductors. Their "lifetime"
is the time when capacity has dropped to 50%. For semiconducors
it is when the failure rate rises again. Also electrolytes
age all the time, just a bit faster when used. Semiconductors
age almost not at all when not used.


Just saying your capacitor comparison is bogus.


I don't dispute that. The problem is low-quality capacitors combined
with a different ageing mode.

It's not just low-qualtiy capacitors but even Japanese brands, like
Rubycon and Sanyo. In my 20-30-year-old TVs, I've replaced only 1
semiconductor (vertical output module) but several electrolytics, and I
doubt this is atypical for such old products.

 

[Arno Wagner]

The shorter estimated lifespans are generally for power semiconductors,
especially bipolars, where damage occurs right at the P-N junction.
Estimates for dense ICs seem to be longer, in the range of 50-200
years, but I don't know if this includes 3 GHz processors
(electromigration?).

I have no idea. And, as I said, I was unable to find current data. I
have one data point though from observation: A bunch of Netgear GA302T
Gigabit cards. They were running at around 60C (heatsink temperature,
maybe 80C chip temperature)and the first failed after 2 years, now it
is 15/22 failed after about 3.5 years. (Incitentially Netgear never
replaced them, despite a 5 year warranty.)

If I assume 3 years average lifetime for these and scale down to 30C,
I get 100 years @ 30C, so you might be quite right about the 50-200
year figure.

I also had one likely termal failure of a chipset in the same PCs,
were the chipset runs at maybe 60C. That would have 12 years average
lifetime, assuming 100 years @ 30.

Ok, I agree with you about the 50-200 years, but at 25C or so.
I don't think the dreating has changed, since there is no reason
for it to. But the base lifetime seems to have gone up significantly.

Today you often get something like 5 years at 80C or so, if you get any
information at all. The 30 year point would then be around 50C. [...]

Nonsense, and a large sample of any electronic equipment will almost
always show more failures among the electrolytic capacitors than the
semiconductors. For equipment at least 10 years old, it's common for
most of its electrolytics to have slightly high ESR and none of the P-N
junctions of the power semiconductors to have excessive leakage.

I don't dispute that. The problem is low-quality capacitors combined
with a different ageing mode.

It's not just low-qualtiy capacitors but even Japanese brands, like
Rubycon and Sanyo. In my 20-30-year-old TVs, I've replaced only 1
semiconductor (vertical output module) but several electrolytics, and I
doubt this is atypical for such old products.

Not surprising. After 20-30 years even an unused electrolyte
capacitor fails. I thought you were talking about 5 years or so.

Basically if you use ceramics and modern foil, they cam match
the semiconuctors. Electrolytes are a cheap alternative
with limited lifetime.

Arno

 

[Folkert Rienstra]

I have no idea. And, as I said,

I was unable to find current data.

Yet you couldn't resist the urge to babble.

I have one data point though from observation: A bunch of Netgear GA302T
Gigabit cards. They were running at around 60C (heatsink temperature,
maybe 80C chip temperature) and the first failed after 2 years, now it
is 15/22 failed after about 3.5 years.
(Incitentially Netgear never replaced them, despite a 5 year warranty.)

If I assume 3 years average lifetime for these and scale down to 30C,
I get 100 years @ 30C, so you might be quite right about the 50-200
year figure.

Whatever the logic behind that is. Babblebot logic?

I also had one likely termal failure of a chipset in the same PCs,
were the chipset runs at maybe 60C. That would have 12 years average
lifetime, assuming 100 years @ 30.

Ok, I agree with you

Whatever that is worth.
Your babbling changes with whichever direction the wind blows.

about the 50-200 years, but at 25C or so.

I don't think the dreating has changed,

Doctor, quick, massive brainfarct here.

since there is no reason for it to. But the base
lifetime seems to have gone up significantly.

Today you often get something like 5 years at 80C or so, if you get
any information at all. The 30 year point would then be around 50C. [...]
Nonsense, and a large sample of any electronic equipment will almost
always show more failures among the electrolytic capacitors than the
semiconductors. For equipment at least 10 years old, it's common for
most of its electrolytics to have slightly high ESR and none of the P-N
junctions of the power semiconductors to have excessive leakage.

I don't dispute that. The problem is low-quality capacitors combined
with a different ageing mode.

It's not just low-qualtiy capacitors but even Japanese brands, like
Rubycon and Sanyo. In my 20-30-year-old TVs, I've replaced only 1
semiconductor (vertical output module) but several electrolytics, and I
doubt this is atypical for such old products.

Not with my Blaupunkts.
Twice the same capacitor, controlling the switch-on of the main Power Sup-
ply. Once the reservoir capacitor in the linear (standby) Power Supply.
Main problem usually: HT transformer, usually the diode at the business
end, causing either tripping the failsave or cause arcing/blazing.
One transformer had the coil insulation failing and spew smoke.

Btw, the failing capacitors were Philips plastic can bi-axial types while
95% of all the other electrolytic caps are Siemens, aluminum canned.

Not surprising. After 20-30 years even an unused electrolyte
capacitor fails. I thought you were talking about 5 years or so.

Pity about my 25 year old TVs full of those little buggers still working.
When they fail it is usally the same capacitor in the same application.
According to one TV repair man they don't like high frequencies, like in
switching power supplies.

Basically if you use ceramics and modern foil, they cam match
the semiconuctors.

Electrolytes are a cheap alternative with limited lifetime.

Babblebot, you really have no shame.
You churn out one idiot statement after the other and are proud of it.

 

[Guest]

Previously (e-mail address removed) wrote:

Ok, power semiconductors are different. My numbers are for CMOS.
Power semiconductors have larger structures and are therefore far
less succeptible to aging. They usually die by melting. For small
scale cmos, this is different.


I have no idea. And, as I said, I was unable to find current data. I
have one data point though from observation: A bunch of Netgear GA302T
Gigabit cards. They were running at around 60C (heatsink temperature,
maybe 80C chip temperature)and the first failed after 2 years, now it
is 15/22 failed after about 3.5 years. (Incitentially Netgear never
replaced them, despite a 5 year warranty.)

Do you mean Netgear wouldn't honor the warranty? If so, you should be
able to get a refund through your credit card, although your card
company may wrongly claim that there's a 60-day limit. The reasoning
behind this is that you paid for the warranty as well as the product
but had no way of knowing that the warranty was missing until you had
to exercise it. Do those cards have voltage converters on them? I've
seen them fail in D-Link cards and USB wireless adapters.

Today you often get something like 5 years at 80C or so, if you get any
information at all. The 30 year point would then be around 50C.
[...]

Nonsense, and a large sample of any electronic equipment will almost
always show more failures among the electrolytic capacitors than the
semiconductors. For equipment at least 10 years old, it's common for
most of its electrolytics to have slightly high ESR and none of the P-N
junctions of the power semiconductors to have excessive leakage.

I don't dispute that. The problem is low-quality capacitors combined
with a different ageing mode.

It's not just low-qualtiy capacitors but even Japanese brands, like
Rubycon and Sanyo. In my 20-30-year-old TVs, I've replaced only 1
semiconductor (vertical output module) but several electrolytics, and I
doubt this is atypical for such old products.

Not surprising. After 20-30 years even an unused electrolyte
capacitor fails. I thought you were talking about 5 years or so.

Even in just 5 years, capacitors are more likely than anything else to
fail.

Basically if you use ceramics and modern foil, they cam match
the semiconuctors. Electrolytes are a cheap alternative
with limited lifetime.

Where can I find a 10uF, 250V or 470uF, 25V ceramic or foil capacitor
in reasonably small sizes? Electrolytics are often the only practical
choices.

 

[Arno Wagner]

Do you mean Netgear wouldn't honor the warranty? If so, you should be
able to get a refund through your credit card, although your card
company may wrongly claim that there's a 60-day limit. The reasoning
behind this is that you paid for the warranty as well as the product
but had no way of knowing that the warranty was missing until you had
to exercise it. Do those cards have voltage converters on them? I've
seen them fail in D-Link cards and USB wireless adapters.

Not a credit card buy. The usual mode of payment here is bank
transfer. Far cheaper. What I mean is that Netgear has 10 of these
defective cards and did not bother sending replacements for the
last 5 months. Eventually we decided it was cheaper just letting
the matter rest. We also had trouble getting a replacement for
a 24 port Netgear GbE switch. They just gave us faulty shipping
instructions, the guy supposed to pick up the unit never showed
and the like. It took 3 months and many, many phone calls to get
the replacement. Netgear used to have very good warranty support. I
guess these days are past.

Today you often get something like 5 years at 80C or so, if you get any
information at all. The 30 year point would then be around 50C.
[...]

Nonsense, and a large sample of any electronic equipment will almost
always show more failures among the electrolytic capacitors than the
semiconductors. For equipment at least 10 years old, it's common for
most of its electrolytics to have slightly high ESR and none of the P-N
junctions of the power semiconductors to have excessive leakage.

I don't dispute that. The problem is low-quality capacitors combined
with a different ageing mode.

It's not just low-qualtiy capacitors but even Japanese brands, like
Rubycon and Sanyo. In my 20-30-year-old TVs, I've replaced only 1
semiconductor (vertical output module) but several electrolytics, and I
doubt this is atypical for such old products.

Not surprising. After 20-30 years even an unused electrolyte
capacitor fails. I thought you were talking about 5 years or so.

Even in just 5 years, capacitors are more likely than anything else to
fail.

Where can I find a 10uF, 250V or 470uF, 25V ceramic or foil capacitor
in reasonably small sizes? Electrolytics are often the only practical
choices.

The thing is that you usually don't need them. For example in a 1MHz
swithcing power supply a 10uF creamic may do as well as a 1000uF
low-ESR electrolyte or even better. But the ceramic part costs a
bit more.

Arno

 

[Curious George]

<chuckling>

Do you now feel better now?

Do you think that somehow "evens the score" for all the many times
I've corrected you, oh all-knowing Folkert?

Mr. Hooker, err I mean Folkert. Would you like me now to shine light
on your personality issues and trollism.

I didn't think so.

 

[willbill]

^^^^^^^^^

turns out you are right

my 7600GT also has a slowdown-point of "only" 105C!

makes me think that maybe i can OC my 6600GT.

so i downloaded the latest RivaTuner,
and went back (from 91.47) to 84.21 (which
is supposedly ok with the latest RivaTuner)
on my slower machine (with the AGP 6600GT)

it remains to be seen if i really can OC
my 6600GT, but i'm far more willing to try
that, than to OC my CPU (can't do much here
given that my 2 mobos are Tyan and
SuperMicro (Opty940 on both))

bill

 

[John Turco]

Pity about my 25 year old TVs full of those little buggers still working.
When they fail it is usally the same capacitor in the same application.
According to one TV repair man they don't like high frequencies, like in
switching power supplies.

<edited>

Hello, Folkert:

When an electronic device has been setting, unused, for many years,
it's a good precautionary measure to connect it to a "Variac" (variable
voltage transformer), before switching it on, initially. This allows
the capacitors to "reform" -- otherwise, they could be destroyed and
possibly, even catch fire, when "juice" is first applied!


Cordially,
John Turco <[email protected]>

 

[Arno Wagner]

Arno Wagner wrote:

turns out you are right

my 7600GT also has a slowdown-point of "only" 105C!

makes me think that maybe i can OC my 6600GT.

Dangerus thought! ;-)

Arno

so i downloaded the latest RivaTuner,
and went back (from 91.47) to 84.21 (which
is supposedly ok with the latest RivaTuner)
on my slower machine (with the AGP 6600GT)

 

[Ed Light]

makes me think that maybe i can OC my 6600GT.

Try the coolbits registry hack. It has a feature to "detect optimal
frequencies". Of course then you have to stability test it. It took my
6600GT AGP up to almost 600/1100.

add a DWORD named coolbits with hex value 3 to registry key:
"HKLM/Software/NVIDIA Corporation/Global/NVTweak/"
--
Ed Light

Smiley :-/
MS Smiley :-\

Send spam to the FTC at
(e-mail address removed)
Thanks, robots.

Bring the Troops Home:
http://bringthemhomenow.org

 

[willbill]

Try the coolbits registry hack. It has a feature to "detect optimal
frequencies". Of course then you have to stability test it. It took my
6600GT AGP up to almost 600/1100.

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^


what does your GPU *temp* max out at,
with 3D game play, at 600/1100?

did you replace the stock heatsink?

if you did, with what?

fwiw, my 6600GT has a decent stock heatsink
(about the same as what's on my 7600GT),
and gives off noticeably less heat. it runs
(stock) at 300/900, which is a lot less than
your current 600/1100

add a DWORD named coolbits with hex value 3 to registry key:
"HKLM/Software/NVIDIA Corporation/Global/NVTweak/"

thank you for that.

fwiw, i put coolbits on my 7600GT, and
the attraction was that it was easy to do

would i benefit there from setting the above
registry key to hex value 3?

on my machine with the 6600GT i've loaded
RivaTuna v2.0 RC 16 (and the nVidia 84.21 driver)
and i like what i see. the downside is that
it is complicated and takes some serious
reading before you do anything

bill

 

[Ed Light]

what does your GPU *temp* max out at,

with 3D game play, at 600/1100?

With the stock cooler it would get to 80C instead of 70C at stock 500/900.

I'm now using a Zalman
http://www.newegg.com/Product/Product.asp?Item=N82E16835118116
It comes with an adapter so you can choose to run it on 5v or 12v. I have it
on a Zalman fan controller. On slow (about 5v) it gets up to 70C. But on
halfway it's much cooler. No need to run it full-bore, though it's not loud
that way.

fwiw, my 6600GT has a decent stock heatsink
(about the same as what's on my 7600GT),
and gives off noticeably less heat. it runs
(stock) at 300/900, which is a lot less than
your current 600/1100

They run at 300 in 2D and in 3D they jump to 500.

thank you for that.

fwiw, i put coolbits on my 7600GT, and
the attraction was that it was easy to do

would i benefit there from setting the above
registry key to hex value 3?

Doing that *is* the coolbits hack. You then, in the nvidia panel, get a
clock speeds tab.

on my machine with the 6600GT i've loaded
RivaTuna v2.0 RC 16 (and the nVidia 84.21 driver)
and i like what i see. the downside is that
it is complicated and takes some serious
reading before you do anything

I haven't tackled Riva myself.


--
Ed Light

Smiley :-/
MS Smiley :-\

Send spam to the FTC at
(e-mail address removed)
Thanks, robots.

Bring the Troops Home:
http://bringthemhomenow.org

 

[Ed Light]

fwiw, my 6600GT has a decent stock heatsink

They run at 300 in 2D and in 3D they jump to 500.

BTW -- I wouldn't overclock the 2D gpu speed. Just the 3D.
When you speed up the memory it stays that way in 2D.


--
Ed Light

Smiley :-/
MS Smiley :-\

Send spam to the FTC at
(e-mail address removed)
Thanks, robots.

Bring the Troops Home:
http://bringthemhomenow.org

 

[Arno Wagner]

With the stock cooler it would get to 80C instead of 70C at stock 500/900.

I'm now using a Zalman
http://www.newegg.com/Product/Product.asp?Item=N82E16835118116
It comes with an adapter so you can choose to run it on 5v or 12v. I have it
on a Zalman fan controller. On slow (about 5v) it gets up to 70C. But on
halfway it's much cooler. No need to run it full-bore, though it's not loud
that way.

A pretty good cooler. But it does not fit every card. On my 7600GT I
have a newer VF900-Cu, which is even better and fits more
cards. Before I had a passive cooler from Zalman on my older card. No
problemas at all. These people are really good engineers!

Arno

 

[willbill]

I'm now using a Zalman
http://www.newegg.com/Product/Product.asp?Item=N82E16835118116
It comes with an adapter so you can choose to run it on 5v or 12v. I have it
on a Zalman fan controller. On slow (about 5v) it gets up to 70C. But on
halfway it's much cooler. No need to run it full-bore, though it's not loud
that way.

thank you for your two replies

i took an extended look at the above Zalman 700,
the Zalman 900 that Arno mentioned, and also
the Vantec Iceberq 5 (sp?)

willbill wrote:

They run at 300 in 2D and in 3D they jump to 500.

ah ha! i knew that but had forgotten it.
thank you again

fwiw, i liked what i saw of the
Zalman 900. (weight for one thing,
a current instant $10 rebate for another,
as well as it appears that it will fit
either of my AGP 6600GT and/or PCI-e
7600GT.) from what i see of the
bolts/screws supplied with the 900,
i'll likely have to get to a hardware
store for 4 "normal" bolts/screws coz
on one machine i've got my video board
in the middle with another card tight
on it's backside

thank you also to Arno, coz i'd
otherwise not have looked at the 900

bill

 

[Arno Wagner]

Ed Light wrote:

thank you for your two replies

i took an extended look at the above Zalman 700,
the Zalman 900 that Arno mentioned, and also
the Vantec Iceberq 5 (sp?)



ah ha! i knew that but had forgotten it.
thank you again

fwiw, i liked what i saw of the
Zalman 900. (weight for one thing,
a current instant $10 rebate for another,
as well as it appears that it will fit
either of my AGP 6600GT and/or PCI-e
7600GT.) from what i see of the
bolts/screws supplied with the 900,
i'll likely have to get to a hardware
store for 4 "normal" bolts/screws coz
on one machine i've got my video board
in the middle with another card tight
on it's backside

thank you also to Arno, coz i'd
otherwise not have looked at the 900

No problem.

Arno