GPU wear and tear

Hi everyone,
I’m wondering how much my GPU will wear and tear if I use it for plotting?
Is it a good idea to let my GPU “rest” for a while? For instance plot for an hour, rest for a couple of minutes and plot again for an hour? Or is it fine to just let my GPU run 24x7 for several days or even weeks?

1 Like

If the GPU can stay cool (u can tell if the fans stop regularly and often), that’s way enough. Heat is the enemy of all electronics, but they are built to run hotter than you’d think, and why they have fans.

IOW, don’t sweat the (small stuff) heat too much, plot-a-lot…be happy :grin:

PS They do get to rest each and every plot…they are not used continuously anyway…so they get to rest.

2 Likes

Thanks.
I’m monitoring the temperature of my GPU (as reported by “nvidia-smi”) and it’s around 70C during plotting which seems fine to me

1 Like

Spare a thought for GPUs that were/are used for mining. I have a 3080, used to reach just over 100c, stripped it down, replaced the thermal pads and it ran a lot cooler, added an additional fan and cooler still. Used to run 80 - 90c 24/7 until I stopped mining a year ago.

It will be used for plotting when I get time to sort it all out.

1 Like

Your GPU will be fine! When I used to GPU Crypto Mine, my GPUs were running flat out for months without a rest. All were running a hefty overclock, literally being tortured for months on end and I never had a single one die on me. I never noticed any ill effects ever. How long will you be plotting for? A week, maybe 2? It’ll be fine!

1 Like

The chances are high that it will be a couple generations behind and out of warranty by the time it dies or needs to be repadded (which has me worried for you guys buying all the 1070’s and 1080’s off ebay, you know they’ve been flogged for 4 years). It’s like tires on a muscle car or a Costco sized pack of energy drinks: use them and enjoy them :sunglasses:.

Flash memory is much more susceptible to hot temps and run cycle wear, as it is meant to keep the given charge for long time. Other components are much less susceptible. If you run nvidia-smi -q, there should be info there at which temp throttling will start (close to 100C on my 3060 ti).

That said, some people replace those thermal pads with copper, and report big temp drops. I have not done it, so cannot say whether this is worth doing. Maybe, those people that do it are trying to overclock their boards, so such mods are helping to do that. Here is a YT video about it - I Copper Modded An RTX 3070 Ti. Memory Temperature Dropped 45 Degrees! This Is How. (110C to 64C) - YouTube, and here is eBay listing for such thing - NVIDIA GeForce RTX 3060/60Ti 3070/70Ti Cooling Thermal Pad GPU Copper Shim Plate | eBay.

2 Likes

@Jacek, How on earth have you sat on that guide and not been tempted to do that LOL :rofl:??? If there’s one thing I love, it’s a good mod that involves lots of tinkering! They were some good results!

shouldn’t be a problem since gpu gonna heat up only on some phases not all the way like when mining eth

lower temp better. I have seen old RX580 raped by years of mining. They still worked, but sometimes unstable.

I kept all my RX5700 XT under 70°c on VRAM…there was slow VRAM wear…they couldn’t hold same speed…you had to reduce clock from time to time.

ASUS RTX 3060 Ti running <60°C during plotting