How much the card can be loaded 24/365

I have such a project with the use of GPU for video analytics so I immediately point out that the specifics of this system is almost 24/7 for 365 days with virtually no stopping.
For the last 10 months or so, I’ve been using a Lenovo RTX Quadro A4000, which I think just died as its already the second card… The card was set to run at about 60-70% [nvidia-smi dmon -s pucm] and was getting about 60-65 degrees Celsius in an air-conditioned room.
There may have been some more features added that may have raised the consumption a bit.
In any case, one day the functions suddenly stopped working, and the card’s performance was continuously running at 100% Memory Controller 70% and getting constantly 35 degrees Celsius.
After replacing the card with another copy, the system started properly.

And now yes, because it was only testing. There were even supposed to be 3 cards x 4 servers in one server. And so a pile of money could go up in smoke…
The question is how much % you can safely set the GPU to run 24/365 like this.
Currently sm is 50% and reaches 65 degrees in an air-conditioned room.

As prices have dropped quite a lot I was still thinking of putting in an RTX 3090 only I don’t know if it would work in a Dell T7910, because at least currently the power plugs don’t allow it. The question is whether such a gaming type of card will better handle 24/365 load? To what % could it be loaded?

*Server run on Windows 11

Can you confirm what is the application that you are using for video analytics?
I ask, as you mention this is running on a Windows 11 machine.
NVIDIA SDKs for Video Analytics run on Linux.

Gaming cards are not recommended for application deployments.

How are you securing your RTX A4000 GPUs? Have you contacted your seller/rep?

This is a commercial VMS program: Axxon ONE

What do you have in mind regarding security features? Do you mean installing additional cooling, for example? I have seen such, but did not want to violate the original card. For my part, I cranked up the fans in the workstation, which were spinning from 60-70% at once, and positioned the station close to the air supply from the air conditioner.

I have not yet contacted the seller either. Should the warranty be honored for working in such conditions where cooling was provided? This card was purchased under BULK conditions.

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one.

RTX Quadro is intended for use in workstations running professional graphic related work, it’s not suitable for 24/7 use cases. RTX 3090 is consumer-grade graphic card and not suitable for 24/7 either. We suggest you consider Tesla (data center) cards such as A100/A30 series.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.