Average lifetime of your discrete video cards?

[momaka]

Re: Average lifetime of your discrete video cards?

GeForce 7-series are known for having shit caps! Even eVGA was caught using Sacon FZs!

Not only eVGA. XFX, PowerColor, VisionTek, and possibly a few others used them as well... though nowhere near as much as eVGA and XFX.

For eVGA, they've been using GSC/Evercon/Sacon for quite a long time on their low- and mid-range GPUs all the way up to the GeForce 7 series. Then it appears they stopped.

For XFX, it looks like they started using them a bit around the GeForce 6 series, and very heavily on their low- and mid-range GeForce 8 series (particularly on the 8600 GT). For the low- and mid-range GeForce 9 series, they started using a new brand by the name of Elcon, particularly the EEZ series. Upon a quick search, I found that these Elcon EEZ caps liked to blow up very similarly to Sacon FZ. While they are not as terrible as FZ, it seems, they are still pretty bad. My theory is that they are just Sacon FZ caps with a new name and possibly a slightly more stable electrolyte (but still crap.)

[eccerr0r]

Re: Average lifetime of your discrete video cards?

Based on this poll so far, people either aren't reporting their failures or they aren't failing as frequently as expected. A video card that lasted 5 years is pretty good, despite it's still annoying.

If the fan failed, I'd say the GPU failed. Granted it can be repaired much more easily than if the die broke, but it's still a failure.

Ultimately I was concerned about the ones that last fewer than 3 years. Fans are a weak point, alas I've had CPU fans that last 3 years just fine (dust removal is the only maintenance that should be necessary.) GPU fans, even with dust removal, have been awful - though in general, those fans 60mm and smaller just don't last...

(and the stock fan of my 5770 IIRC is at least 60mm, so maybe that's why it has lasted so long...)

[Curious.George]

Re: Average lifetime of your discrete video cards?

On a finishing note, here's something to think about: how come (desktop) CPUs fail so rarely, but GPUs fail so often? Does the load temperatures have anything to do with this? Or perhaps the socket? (The fact that most desktop CPUs don't have to go through a harsh heat-cycle for BGA soldering???)

CPUs tend to have a lot better thermal management -- they have room for it (vs. "add-in" video cards) as they aren't constrained to the geometry of a bus slot.

Also, don't underestimate the relative horsepower of GPUs vs CPUs. Or, the duty cycles they have to support

[Heihachi_73]

Re: Average lifetime of your discrete video cards?

My Radeon 9550 (installed in 2005) is still going 24/7 in my 2004 Celeron system. Just like Momaka's Radeon 9200, this one also saw a lot of gaming, from GTA3 and Vice City (I never liked San Andreas) through to console and arcade emulators. Today though, even MAME has outgrown this card's specs (0.179 was the last version which can run on this machine without it falling back to slow OpenGL or even slower GDI), requiring a more modern chipset which supports non-power-of-two textures and pixel buffer objects and everything else modern gamers take for granted.

[Vishera]

Re: Average lifetime of your discrete video cards?

But for newer ones that vary the fan speed according to the temperature, the maximum load temperature will depend on the pre-programed cooling profile... and with that, some GPUs will run at somewhat decent temperatures and others just a hair below their maximum allowable temperature (think single-slot cooler GeForce 8800 GT and 9800 GT here, as well as 8800 GTS/GTX/Ultra and single-slot cooler Radeon HD4850, just to name a few - hence the failure rate on those).

Great point about default cooling profiles not being aggressive enough. I imagine this is to reduce noise output, but I don't like it.

In the past I've used custom fan curves on GPUs that ran hot at stock clocks.

[ChaosLegionnaire]

Re: Average lifetime of your discrete video cards?

On a finishing note, here's something to think about: how come (desktop) CPUs fail so rarely, but GPUs fail so often? Does the load temperatures have anything to do with this? Or perhaps the socket? (The fact that most desktop CPUs don't have to go through a harsh heat-cycle for BGA soldering???)

it could be due to the fact that gpus have a much higher transistor count and thus transistor density than cpus. this makes gpus more susceptible to electromigration failure from high voltages and/or high temperatures. if u look up gpus and cpus built on the same process node and look up the transistor count for those gpus and cpus, u'll see it makes perfect sense.

[mockingbird]

Re: Average lifetime of your discrete video cards?

it could be due to the fact that gpus have a much higher transistor count and thus transistor density than cpus. this makes gpus more susceptible to electromigration failure from high voltages and/or high temperatures. if u look up gpus and cpus built on the same process node and look up the transistor count for those gpus and cpus, u'll see it makes perfect sense.

I'd like to get tom66's take on this. He claims that they die for other reasons. To be fair to him, he did say "CPU", not GPU.

Electron migration? On a consumer CPU? Bollocks.

AMD farm out all their fab stuff to TSMC anyway - same fabs make Nvidia and Apple chips.

I can assure you as a hardware engineer, electron migration is not a common cause of failure. Yes it might be an issue in an aerospace device due to the high cosmic ray energies imparted, but really unlikely in a consumer chip

[Snayperskaya]

Re: Average lifetime of your discrete video cards?

I had a Radeon HD 4870 1GB (Sapphire) that I used from 2009 to 2011, sold it to a friend that used it till 2014 (IIRC). Bought it back, cleaned it and resold it. It is prolly still running strong.