GTX 970 experiences extreme throttling under high voltage and clock conditions.
GTX 970 experiences extreme throttling under high voltage and clock conditions.
Hello everyone,
Under different loads, my MSI GTX 970 Gaming shows a noticeable drop in core clock and voltage.
When I play any game or run stress tests beyond Furmark, my card performs at its peak and stays stable, even with over 1500 cores. Unigine Heaven runs smoothly at maximum settings without dropping, even after a three-hour session.
However, when I run Furmark, the voltage drops from 1234mV to a much lower 1060mV (what?), and the clock also decreases noticeably. While increasing core overclocking does raise the max clock, it never reaches "boost speed."
I suspect the power supply unit might be the issue—it’s a 400-watt unit with high-quality specs and 36A double 12-rail, yet it still manages Heaven at full settings for hours.
[image link] (top image looks good, bottom image is worse)
On your taskbar, is the green icon "97" indicating your GPU usage? You can install GPU-Z to check the Perfcap Reason.
I've just observed that the clocks decrease when the power percentage reaches 112%, since your Power Limit is at 110% and your card requires more, which can't be handled in the standard BIOS (Modifying the GPU BIOS will void your warranty).
Consider disabling Memory Overclock completely and observe if power throttling continues, as it shouldn't.
Thanks for your input. I've turned off all overclocking and kept the power limit and voltage at their maximum settings. The voltage is now only slightly higher, though still well under the 1.2+ it should reach. (refer to screenshot)
Current stock configuration includes power limit and voltage settings.
P.s.: I've heard rumors that Furmark has previously bricked many GPUs, and Nvidia has updated its drivers accordingly. It might be exaggerated, but it could explain the situation.
I just observed this issue with other benchmarks too, influenced by the level of stress on the GPU. The "furry" benchmarks consistently reduce voltage and core clock speeds. The greatest reduction is seen only with Furmark, significantly more than others. Here are some examples:
http://i.imgur.com/C3rdoF7.png
non furry (+171 clock, 1512 effective, 1243 voltage, stable)
http://i.imgur.com/TdyzUI5.png
FURRY (+171 clock, 1366 effective, 1068 voltage, stable)
http://i.imgur.com/7b7lCHz.png
non furry (+171 clock, 1512 effective, 1243 voltage, stable)
I’m really puzzled by these variations.
Would you be okay with downloading the 3DMark Demo from Steam and running it? I want to check your graphs and compare them with mine to spot any differences.
Edit: I've just realized that like my card, yours is going over 63c which causes some minor throttling. Try removing the side panel or increasing the GPU fans to keep the temperature under 60c and see if the clock drops stop.
I've already tried adjusting the fan speed to 100% and it only lowered the temperature to a cool 50 degrees under load, without any improvement. The issue isn't the temperature itself. I plan to replace my PSU later this month to test its effect. If nothing changes, I'll share the graphs. The consistent finding is that power limits are hit during throttling, which matches what I see in the screenshots. It seems that might be the root cause, though a new PSU could help. Thanks!
Josef_ :
I've tried setting fan speed at 100% and it cooled down to a cool 50 degrees at load, no change. Temp is not the problem. I'm going to replace my PSU later this month and see what it does. If there's no change, I'll upload the graphs. The one consistency is that when it throttles, the power limit is reached, as can be seen on the graphs on my screenshots. Surely that is the issue, not so sure a new PSU will fix that, though.
Thanks!
Does it still do it at stock/underclocked?
Sergeant_Sneaky shared his observations with Josef. He noted that setting the fan speed to 100% didn't improve cooling, and the temperature remained stable at 50 degrees during load. He plans to replace his PSU later this month and compare the results. He mentioned that when throttling, the power limit is reached, which he confirmed from his screenshots. He remains unsure if a new PSU would resolve the issue, though he expects it to help. He also confirmed that at stock/underclocked settings, the performance is consistent and doesn't significantly affect scores. He found it annoying but not a major concern.
Josef_ tried adjusting the fan speed to 100% but didn't notice any improvement in cooling, even under load. The temperature stayed steady at 50 degrees, indicating the issue isn't with cooling. He plans to replace the PSU later and will share data if needed. The consistent behavior is that power limits are hit during throttling, as shown in his screenshots. It's unclear if a new PSU would resolve this. Thanks!