Can IC`s which run hot last longer if heatsinks are glued onto IC`s when new ?

[TE Chea]

I found every hdd ( Seagate, Maxtor ) has at least 1 IC far hotter
than other IC`s. I glue hsinks onto these hot IC`s, to cool them : 1
ata33 Seagate hdd already works noticeably faster, after I glue a
hsink onto its hottest IC. In 12-03 I glued a hsink onto a nVidia
nForce 2 south bridge ( hot ), its IDE controller then worked 5-8
% faster. Now I even fit hdd`s @ an angle ( 45 - 70° ) in casings
so these hsinks can cool faster ; air heated by hsinks can flow
upward faster.
I suspect after IC`s are used for a few years, glueing hsinks will
be too late to extend IC`s' lives ; last week 1 ata100 Maxtor hdd (
bought in 5-02 ) 's IC`s died right after I used Disk Checker to
check all of hdd's 19Gb ( in 1 partition ), despite hsinks glued onto
all the 3 hot IC`s.
Can hot IC`s last longer, from this added cooling ( i.e. hsinks are
fitted ) when IC`s are still new ?

 

[John Larkin]

I found every hdd ( Seagate, Maxtor ) has at least 1 IC far hotter
than other IC`s. I glue hsinks onto these hot IC`s, to cool them : 1
ata33 Seagate hdd already works noticeably faster, after I glue a
hsink onto its hottest IC. In 12-03 I glued a hsink onto a nVidia
nForce 2 south bridge ( hot ), its IDE controller then worked 5-8
% faster.

Neither a hard drive nor a synchronous clocked bridge chip will run
any faster if you cool one chip. I just ain't happening.

Can hot IC`s last longer, from this added cooling ( i.e. hsinks are
fitted ) when IC`s are still new ?

IC failure rates double for about every 10 deg C increase in junction
temperature, so heatsinking a chip can reduce its failure rate. But
hard drives usually fail for mechanical reasons, and most chips are
designed to last a long time as-sold. The people who design hard
drives understand all this stuff.

John

 

[Mr. J D]

I found every hdd ( Seagate, Maxtor ) has at least 1 IC far hotter
than other IC`s. I glue hsinks onto these hot IC`s, to cool them : 1
ata33 Seagate hdd already works noticeably faster, after I glue a
hsink onto its hottest IC. In 12-03 I glued a hsink onto a nVidia
nForce 2 south bridge ( hot ), its IDE controller then worked 5-8
% faster. Now I even fit hdd`s @ an angle ( 45 - 70° ) in casings
so these hsinks can cool faster ; air heated by hsinks can flow
upward faster.
I suspect after IC`s are used for a few years, glueing hsinks will
be too late to extend IC`s' lives ; last week 1 ata100 Maxtor hdd (
bought in 5-02 ) 's IC`s died right after I used Disk Checker to
check all of hdd's 19Gb ( in 1 partition ), despite hsinks glued onto
all the 3 hot IC`s.
Can hot IC`s last longer, from this added cooling ( i.e. hsinks are
fitted ) when IC`s are still new ?

Ofcourse! You can buy IC heatsinks, I myuself own a couple of 14 DIP
Heat Sinks,

 

[Homer J Simpson]

I found every hdd ( Seagate, Maxtor ) has at least 1 IC far hotter
than other IC`s. I glue hsinks onto these hot IC`s, to cool them

Do the words "Warranty Void" mean anything to you?

 

[Puckdropper]

Do the words "Warranty Void" mean anything to you?

Yeah, but just TRY getting replacment under warranty. Depending on the
company, it's not a fun (or anything short of unpleasant) experience.

Puckdropper

 

[Homer J Simpson]

Yeah, but just TRY getting replacment under warranty. Depending on the
company, it's not a fun (or anything short of unpleasant) experience.

Most all drives now seem to come with a 5 year warranty.

 

[[email protected]]

Most all drives now seem to come with a 5 year warranty.

You might be able to get the manufacturer to replace your failed drive,
but you will never be able to get them to repair and return your broken
drive, even if it needs noo more than a replacement electronics card.
You will lose all of your data, even if the drive were repairable.. If
there is some way to do it, it is better to keep the original drive
working.

If you want to keep the drive and its components cool, you could add
another fan and direct its flow toward the drive.

 

[Chris]

Neither a hard drive nor a synchronous clocked bridge chip will run
any faster if you cool one chip. I just ain't happening.


IC failure rates double for about every 10 deg C increase in junction
temperature, so heatsinking a chip can reduce its failure rate. But
hard drives usually fail for mechanical reasons, and most chips are
designed to last a long time as-sold. The people who design hard
drives understand all this stuff.

John

And the business of mounting the hard drive on an angle doesn't sound
quite right. As John says, drives almost always fail for mechanical
reasons, as long as the enclosure ventilation is keeping the case at a
reasonable temp.

It's SOP to make sure rack-mount servers are mounted on the square to
maximize reliability, AFAIK for the HD. I believe drive bearings
work best when the drive is mounted either straight vertically or
horizontally.

Good luck
Chris

 

[Homer J Simpson]

If you want to keep the drive and its components cool, you could add
another fan and direct its flow toward the drive.

That'd be my choice.

Managing the data seems to get harder and harder - tape is out for half
terabyte storage system it seems.

 

[Eeyore]

Most all drives now seem to come with a 5 year warranty.

Do they ?

Graham

 

[Eeyore]

You might be able to get the manufacturer to replace your failed drive,
but you will never be able to get them to repair and return your broken
drive, even if it needs noo more than a replacement electronics card.
You will lose all of your data, even if the drive were repairable.. If
there is some way to do it, it is better to keep the original drive
working.

Use RAID mirroring if it bothers you.

Graham

 

[TE Chea]

Are IC`s always hottest @ their centres ?

| Neither a hard drive nor a synchronous clocked bridge chip will run
| any faster if you cool one chip.
s7 mboards' PipelineBurst sram too work faster with added cooling,
very noticeable even without a benchmark. WinTune98 ram speed
rose from 830 to 960 mb/s.
After I added Arctic Silver 1 paste onto a K6III400 's die ( nothing
else was changed ), this cpu was noticeably cooler & faster.

 

[Eeyore]

Are IC`s always hottest @ their centres ?

| Neither a hard drive nor a synchronous clocked bridge chip will run
| any faster if you cool one chip.
s7 mboards' PipelineBurst sram too work faster with added cooling,
very noticeable even without a benchmark. WinTune98 ram speed
rose from 830 to 960 mb/s.
After I added Arctic Silver 1 paste onto a K6III400 's die ( nothing
else was changed ), this cpu was noticeably cooler & faster.

How does a CPU run faster without changing the clocking ?

Graham

 

[TE Chea]

| How does a CPU run faster without changing the clocking ?

I never saw an explanation of this. What was weird was WinTune
98 & Sandra indicated little / no extra cpu speed. My guess is
resistance drops with temperature ( electrons move faster ) : I
believe if 1fits a peltier & water cooler to a cpu with A-Silver
paste, 1 will see / feel an even bigger rise in speed @ the same
clock speed.
sdr sdram chips ( TSOP / BGA ) which run hot will also work
noticeably faster @ same clock speed, if cooling is added.

 

[Bob Myers]

| How does a CPU run faster without changing the clocking ?

I never saw an explanation of this. What was weird was WinTune
98 & Sandra indicated little / no extra cpu speed.

That's because there wasn't any. For any conventional
CPU, it doesn't matter how quickly things get done within
a given clock cycle - it will still take whatever number of
clock cycles, at whatever the clock rate is, to complete a
given instruction. The only way you can improve performance
without increasing clock speed is if you make external
things run more efficiently such that the processor isn't
wasting clock cycles waiting for something external to
get done. But if those "external things" are themselves
clock dependent, well...

Bob M.

 

[Don Klipstein]

Are IC`s always hottest @ their centres ?

| Neither a hard drive nor a synchronous clocked bridge chip will run
| any faster if you cool one chip.
s7 mboards' PipelineBurst sram too work faster with added cooling,
very noticeable even without a benchmark. WinTune98 ram speed
rose from 830 to 960 mb/s.
After I added Arctic Silver 1 paste onto a K6III400 's die ( nothing
else was changed ), this cpu was noticeably cooler & faster.

Did this cooling merely enable use of a higher clock rate or use of
smaller allowances for latencies and delays such as RAM latencies and
delays?

Cooling an IC that was working successfully before will not
significantly reduce its error rate that was already zero or close enough
to zero for your machine to adequately work. But cooling may allow higher
clock rates and smaller alowances for latencies/delays.

If cooling alone speeds something up, then you have one of two
situations:

1) There is a significant error rate along with automatic error
correction. If this is the case, you should notice a high inverse
correlation of performance to room temperature. That means that the
design is marginal, mounting is marginal, or allowances for
latencies/delays is marginal (which I would count as marginal design) or
the clock rate is marginal (which I would count as marginal design). And
by marginal I mean "probably so bad as to be outright lousy".

2. You have something automatically checking for how aggressively to set
clock rate and/or allowances for latencies/delays.

I consider 1 to be more likely. If 2 is the case, I would be leery that
whatever is automatically changing any clock rates or allowances for
delays/latencies to be pushing very close to the limits.

Another consideration: Most computer usage is now done with all
software and hardware working so fast that improvements are
non-improvements.
Notable exceptions are modems often being a limiting factor, hard drives
being a limiting factor getting work done usually less than 1% of the
time, and processing of software loops requiring a large number repeated
operations (generally billions-plus given modern computer capabilities).

- Don Klipstein ([email protected])

 

[TE Chea]

Are IC`s always hottest @ their centres ?

| Did this cooling merely enable use of a higher clock rate or use of
| smaller allowances for latencies and delays such as RAM latencies and
| delays?
Nothing else was changed by me.

| 1) There is a significant error rate along with automatic error
| correction. If this is the case, you should notice a high inverse
| correlation of performance to room temperature
No such correlation was noticeable.

| 2. You have something automatically checking for how aggressively to set
| clock rate and/or allowances for latencies/delays.
Do PLL IC`s ( e.g. ASD ae71c604-37 & Winbond w83194r-58
for SiS530 & VIA mvp3 ) or mboard chipsets do this ?

 

[Eeyore]

| How does a CPU run faster without changing the clocking ?

I never saw an explanation of this. What was weird was WinTune
98 & Sandra indicated little / no extra cpu speed.

Because there was no extra speed.

My guess is
resistance drops with temperature ( electrons move faster )

No, resistance increaes with temperature and It's nothing to do with 'electron
speed' either.

: I
believe if 1fits a peltier & water cooler to a cpu with A-Silver
paste, 1 will see / feel an even bigger rise in speed @ the same
clock speed.

No you won't.

sdr sdram chips ( TSOP / BGA ) which run hot will also work
noticeably faster @ same clock speed, if cooling is added.

You can 'clock' them faster and get waway with it perhaps, but they won't work
any faster without changing that.

What you're seeing is your belief in an idea taking precedence over the reality.
In short, you're 'kidding yourself'. The same kind of thing happens with those
bogus high-price audio and video cables.

Graham