INTL 3:770/1 3:770/3
REPLYADDR Pancho.Jones@proton.me
REPLYTO 3:770/3.0 UUCP
MSGID: 125f741c
REPLY: a8b766ef
PID: SoupGate-Win32 v1.05
On 12/9/24 13:11, The Natural Philosopher wrote:
> On 09/12/2024 12:28, Pancho wrote:
>> On 12/9/24 10:50, The Natural Philosopher wrote:
>>> On 08/12/2024 19:50, David Higton wrote:
>>>> In message
>>>> The Natural Philosopher wrote:
>>>>
>>>>> It's an interesting thought as to why one would use a fan at all.
>>>>> If its
>>>>> such a high compute task that you need one, maybe a bigger Pi or an
>>>>> Intel based machine is indicated.
>>>>>
>>>>> I dislike fans. They fail.
>>>>
>>>> PC fans run pretty much all the time. A fan on a RasPi is likely to
>>>> run less of the time, and could well last longer overall.
>>>>
>>>> Fans fail. Disc drives fail. SSDs fail. Batteries fail. Reservoir
>>>> capacitors fail. But before they do, they are very useful.
>>>>
>>>
>>> Such an ArtStudent™ view of life.
>>>
>>> Do you know what MTBF means?
>>>
>>
>> I was surprised you'd use MTBF for a component which is expected to
>> steadily deteriorate due to wear and tear.
>>
>> I though MTBF was more a random failure thing.
>>
> No it isnt that at all.
>
>
>> For some relatively reliable components, such as people, you initially
>> see a relatively low failure rate, but come 80 or 90 years they start
>> dropping like flies, due to wear and tear.
>
> Yup. MTBF of peole is about 70 years.
>
I read MTBF for people is about 700 years.
>>
>> For some things like atomic an atomic nucleus, the failure does seem
>> random, so MTBF seems applicable.
>>
> Never used.
>
Well yes and no,
Given f = failure rate (e.g probability of nuclear decay)
half life = ln(2)/f.
MTBF = 1/f.
Therefore:
half life = ln(2) * MTBF = ~0.69 * MTBF
essentially they are the same, apart from the 0.69 conversion factor.
>> I don't know which it is for PC fans, but would assume it is more wear
>> and tear than random.
>>
> In general fans fail for one reason only. Bearing failure. The cheapos
> use phosphor bronze plain bushes and these dry out and seize up, wear
> out and get noisy and start slowing down or get clogged with people's
> cruft.
>
> You can go for sealed ball races if you like, as in hard drives, but
> the price goes up.
>
> In terms of drying out, its time elapsed, not time spent running. Same
> for cruft. Only bearing wear is time dependent.
>
> None of these are random., All if them are however dependent on
> conditions and maintenance
>
For a stingy old man like me, yes. I wait till things break. So MTBF is
not particularly useful. I want long average life, not reliability until
they get old.
> MTBF is an attempt to get a handle on how long a collection of parts
> should stay operational given the spreads of failures in a spread of
> conditions of the individual parts
>
> There will always be variations in conditions and manufacturing quality
>
MTBF is the type of metric used by someone who will replace bits before
they get old.
If I were managing a data centre and I knew human intervention was very
costly (I suspect it is), I might have a policy of replacing parts
before the failure rate became high due to old age.
I know the companies I worked for hated investigating failure, or fixing
stuff. They much preferred replacement. Essentially components were
cheap, human intervention was not.
Obviously if the equipment is expensive, or you have cheap slave labour,
the equation may change.
> My experience of cheap fans is that 5 years was about the MTBF.
>>
>
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)
SEEN-BY: 19/38 105/81 106/201 128/187 129/305 153/757 7715 218/700
SEEN-BY: 218/840 220/70 226/17 30 100 227/114 229/110 111 114 200
SEEN-BY: 229/206 300 317 400 426 428 550 616 664 700 705 266/512 267/800
SEEN-BY: 282/1038 291/111 292/854 310/31 320/219 322/757 342/200 396/45
SEEN-BY: 460/58 633/280 281 412 418 420 509 2744 712/848 770/1 3 100
SEEN-BY: 770/330 340 772/210 220 230 902/26 5020/400 5075/35
PATH: 770/3 1 633/280 229/426
|
