0
\$\begingroup\$

In the near future we will be probably flooded with graphic cards they were used for crypto mining. There are many not very well-founded discussions about their reliability. And similarly many un-founded comparisons between used and used for crypto mining. Hence I would like to dig deeper, educate myself and find more information's regarding this service life topic. Any technically/physically focused comments and sources describing components service life welcomed. I propose to skip electrolytic caps in this discussion as its reliability is already well described. However ICs, ceramic caps, resistors and magnetics is part of my focus.

As a starting point I cite Murata MLCC chart where with X5R they declare that 20C difference in case case temperature (range 60C-80C) is a question of almost 10 years of service life. Woo-ow!

Muratta MLCC temperature vs service life

\$\endgroup\$
1
  • \$\begingroup\$ What is your question? 20C difference in case case temperature... 10 years of service life so that means: if you want a design which last long, you should design it such that it remains cool. Also realize that statistics are involved here. An estimated lifetime of 1 year does not mean that all devices will fail after 1 year. It means a certain percentage of failures can be expected. How many percent will fail after 1 year is not mentioned. Some devices will fail in less than 1 year. \$\endgroup\$ May 27, 2021 at 10:12

1 Answer 1

0
\$\begingroup\$

There are many not very well-founded discussions about their reliability.

Hm, seems to be a very normal computer component, not unlike say, a server mainboard. Knowledge on the reliability of those after continuous use should transfer very smoothly.

I propose to skip electrolytic caps in this discussion as its reliability is already well described.

Like all other components. It's not like we've been building complex computer equipment for only 5 years or so; caps might be more in public discussion, but they're not "special" in terms of being considered in terms of reliability. So, yeah, you should include capacitors, RAM, SMPSes, the PCB itself, … everything.

Any technically/physically focused comments and sources describing components service life welcomed

You're starting at the wrong end. Learn how systems composed of components fail, then you'll learn which components you need to model better. Really, looking at individual components (not even classes of components: you're literally looking at a single type of capacitor from a single vendor!) before understanding failure models is like trying to understand the rainforest by looking at every beetle.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.