Steel Nomad Stress Test - Not passed

[Akela]

Does this mean my computer is malfunctioning?
I use W11 with an NVIDIA 3090 (driver 366.03) and Intel Core i9-12900K

Steel Nomad Stress Test - https://www.3dmark.com/3dm/119330390
Steel Nomad - 4800 pts - https://www.3dmark.com/3dm/119330555

Also in Time Spy Extreme it used to score 9500 like here (https://www.3dmark.com/spy/48967208) now it doesn't go over 9,300 https://www.3dmark.com/spy/50919057

[XYX]

doesnt mean the pc is malfunctioning just means the frame time lag is slightly below to pass this can be down to even the driver not being optimal

[UL_Jarnis]

This tells you that during a 20-minute run, the system had variable performance (in your case, went down by 5%)

Or in other words, the score of the best loop and the worst loop during 20 minutes was off by 5%. A "pass" requires staying within 3% and setups that do not thermal throttle or have background program interference usually stay within 1-2%. Note that even "perfect" systems rarely get better than 99%, so some variance is always inevitable and 97-99% is completely normal.

As a baseline, this does not mean that anything is malfunctioning. However, it may indicate one of these potential issues.

- Your GPU model has, by design, a cooler and/or fan curve that cannot keep the GPU out of thermal throttling. Some GPU models are like this. Palit 3090 that you have is an example of a model that is known not to be able to stay out of thermal throttling at stock fan curve. Setting fans to run at a higher speed can help, but the caveat is that the fans will then be super noisy. Granted, the difference in performance is small, so it wouldn't really matter in day-to-day use, but the reality is that they shipped, by choice, a model that has a peak performance level that cannot be sustained over a longer period at stock fan speeds.

Note that above applies to desktop GPUs. Laptops are a whole different story and there are models that can throttle over 10% without anything being technically wrong, with every unit of the same model doing the same thing - the model just has GPU cooler that is wholly inadequate to the GPU being used.

- Your GPU has poorly applied or aged thermal paste, causing the cooler to be unable to properly cool the GPU. This is a possible cause in situations where even maxed out fans cannot keep the card out of thermal throttling or if throttling is very heavy. Such cases also tend to have higher-than-usual GPU temperatures at idle. Thermal paste and thermal pads (for VRMs and memory) can be replaced for a fix, but this is not exactly a trivial operation, and there is a risk that you could damage the GPU. You'd also need exactly correct thermal pads that vary by card model. If this is the reason and the throttling is heavy (more than what was seen in your case), one could argue the card should be serviced under warranty. As to what the card manufacturer will say may vary. But in your case (just 5% difference and assuming stock fan curve) I doubt this is the issue.

- Your system has inadequate case ventilation - under heavy load the GPU cooler cannot push out hot air well enough and that causes thermal throttling. Bit rarer reason, but could happen if the case is not a good fit for the parts inside and airflow is poor. One way to check for this is simply to open the case side panel and set up a desk fan to blow cool air on the GPU. If this magically improves things, it could imply case ventilation is not good enough.

- There is some kind of performance hit from background programs that causes the performance to fluctuate during the run. This would be the suspect if the "worst" run is not the last one during a 20-loop run. Check the FPS graph on the UI and see if it degrades smoothly (as temperature rises) before hitting some plateau. If instead you see just one or two runs randomly dipping while later runs are again faster, then it is unlikely this is a thermal issue, and instead you have something running on the system that can degrade 3D graphics performance. Try to minimize the number of background applications when benchmarking.

But ultimately... we're talking about 2-4% difference in performance vs. completely non-throttling result, which is probably going to be a single digit fps difference in game. You are hard pressed to notice such a difference in anything other than a benchmark score.

[Akela]

Okay, thanks for the tips. I understand then that it is nothing to be alarmed then. Summer is starting in my region and strong heat is coming. I figure it is time for a preventive cleaning and change of thermal paste.