Do high power CPUs and GPUs fail earlier?

[chtorogu]

I just got stiffed with an i5 4790k which seemed like a bargain but turned out to have blown caps. I guess it must have been overclocked exessively or badly and or cooled poorly. This got me thinking about these extreme high power parts that are common in the high end when it comes to GPUs and Intel CPUs. Does the higher wattage flowing through those tiny circuits, caps and resistors mean that they will or are likely to fail sooner than parts with much more moderate power draw? Or are they, providing they're adequately cooled, as reliable long lasting as lower powered stuff?

[jaslion]

1 minute ago, chtorogu said:

turned out to have blown caps.

What caps? The ones at the bottom? If those are broken that is NOT form ocing but from physical damage or other catastrophic damage from like a power spike.

 

1 minute ago, chtorogu said:

fail sooner than parts with much more

So far it's been a no

[RONOTHAN##]

It's complicated. Higher current draw can definitely increase the amount of wear and degradation on a piece of silicon, but at the same time it's pretty rare for the silicon itself to be the piece that dies. Usually if there's something that dies in a setup, it's power delivery circuitry. 

 

Top of the line motherboards and GPUs will have VRMs that are capable of powering twice the amount of current that the CPU is physically able to draw, while lower end options can be anywhere from OK to unable to provide half of the power the chip can draw at full load without overheating or blowing up. Lower end GPUs and motherboards have worse power delivery, but if they're paired with lower power pieces of silicon they don't necessarily need to be better and will have similar lifespans to higher end chips. For comparison, 980 Ti reference cards are fairly likely to blow up as the VRM on them is rather underpowered for what they will pull, while a 980 Ti KP is about as overkill as you can get and practically never fail. Meanwhile, GTX 960s with fairly sensible VRM to current draw ratios don't fail all that often either. 

 

One other thing to mention is that if the VRM on a motherboard or GPU fails, that can in a lot of cases blow up the silicon as usually it will send 12V to the chip and most CPUs/GPUs don't like having 12V thrown at them. It's not necessarily that they'll immediately die, it depends on how long that happens for as well as how resistive the power rail is, but it's never a great idea and is likely to kill stuff. 

 

If your chip has actual blown capacitors on it (I.E. burn marks on the bottom), odds are that's what happened, the motherboard blew up and took the CPU with it. That's pretty rare for Haswell CPUs like yours as they drew so little power, though not impossible if the board was slightly defective or something. 

[chtorogu]

6 minutes ago, jaslion said:

What caps? The ones at the bottom? If those are broken that is NOT form ocing but from physical damage or other catastrophic damage from like a power spike.

Two of the caps (or maybe resistors?) at the bottom are blackened and there's slight signs of damage around. I've bought many used Intel CPUs and never had an issue so I didn't look too closely at it before approving the purchase. Stupid mistake and won't do it again.

 

Perhaps this would be relevant for LTT labs to test as many people are spending high sums of money on these extreme high power parts. They could put together systems with moderate power draw and systems with extreme power draw and set them to do CPU/ intensive tasks over and over 24/7 and see which fail first and when.

[Crunchy Dragon]

Higher power draw does not equate faster death.

 

As long as a given part is kept under control with regards to power input and cooling, it should run basically forever. Overvolting usually won't pop off resistors on the bottom of a CPU, but it doesn't sound impossible if the motherboard crapped itself and sent entirely too much voltage to the CPU.

[jaslion]

47 minutes ago, chtorogu said:

Two of the caps (or maybe resistors?) at the bottom are blackened and there's slight signs of damage around. I've bought many used Intel CPUs and never had an issue so I didn't look too closely at it before approving the purchase. Stupid mistake and won't do it again.

 

Perhaps this would be relevant for LTT labs to test as many people are spending high sums of money on these extreme high power parts. They could put together systems with moderate power draw and systems with extreme power draw and set them to do CPU/ intensive tasks over and over 24/7 and see which fail first and when.

Yeah this sounds like a board that blew and they sold you a known dead cpu

[chtorogu]

Thanks for the info, guys. I'm not so knowledgeable when it comes to the technical and electrical side of tech but find it interesting.

 

Seems she did sell me a dead CPU, yes. Had such a good rating too, but goes to show that's not a guarantee. But could be it wasn't originally hers and she found it in some drawer or something, I guess.

 

I hope I'm wrong about it. I wonder though since there are so many parts involved if some minor QC problems are not more likely to turn catastrophic with like an i9  than an i5 or a Ryzen CPU. Hope I'm wrong. Either way I don't think it's worth hundreds of watts extra to gain more performance and I woudn't feel comfortable with such a system under my desk. Fine if you do like it though, hope they last you a long time.

[Crunchy Dragon]

22 hours ago, chtorogu said:

I wonder though since there are so many parts involved if some minor QC problems are not more likely to turn catastrophic with like an i9  than an i5 or a Ryzen CPU. Hope I'm wrong.

You probably are wrong. If this was the case, we would've been hearing about i9s blowing up way back in 2017 on X299 when they first became a thing, if not earlier on the higher end Xeon chips with high double digit corecounts on X99. Same for Ryzen and Threadripper, if higher power draw was a clear negative all around, it would've been a problem much earlier.

 

As stated above, you're not very likely to have problems like that on an older Haswell CPU(not sure if you have an i5-4690K or i7-4790K based on your original post). Those chips ran pretty cool and didn't consume a ton of power, even with some higher overclocks.

 

As far as higher power draw for more performance....there's not really a way to increase performance without increasing power draw. There's only so much optimization that can be done on the software side and once a program is as optimized as it can be, the only way to increase performance in said program is to increase your power draw, either by overclocking or upgrading hardware. Higher performance will almost always equal higher power draw, which is why power rating for CPUs and GPUs has only gone up.