Throttling Gtx 1660 super (oc/uv)

[Oliveros]

Hello everyone,

I'm experiencing an issue with my NVIDIA GeForce GTX 1660 Super graphics card. Despite the settings in MSI Afterburner that allow me to raise the power limit to 120% and increase the memory clock by 800 MHz, the card downclocks to 2010 MHz after long gaming sessions or intensive work. It's quite irritating, especially since the temperatures under load reach about 70°C. It seems to be caused by the thermal throttling mechanism, but I'm hoping there's a way around it.

I've already attempted undervolting, but it seems I've reached the limit beyond which I can't go without losing stability. Editing the card's BIOS seems out of my league, and I wouldn't want to damage anything. The issue might also be related to the power limit, which I believe is around 125 W, but I'm not certain how to verify and potentially modify this.

Has anyone else encountered a similar issue and found a solution? Is there a way to adjust the temperature threshold at which the card downclocks without risking damage? Or is there another method to deal with this downclocking?

I would be very grateful for any advice. I should mention that I'm planning to purchase a new graphics card soon, so I'm not too worried about the future overheating of the current one, but I would like to resolve this issue before upgrading.

Thank you in advance and best regards,
Oliveros

 

[droopyRO]

load reach about 70°C.

What about vRAM or hot spot temperature ? Di you try to see if it downclocks with the fans at 100% ?

 

[RJARRRPCGP]

The Supers max out at 135W, IIRC. Unless you got one where the VBIOS don't even let you set it to 135W.

 

[pressing on]

This downclocking is a feature of the NVIDIA Boost Algorithm for these cards, they all do it. It is based on temperature sensors. The temperature thresholds as you have discovered are lower than you might hope for. The Algorithm is baked into the firmware (BIOS) for the card, I don't think there is any way of altering it.

 

[Oliveros]

What about vRAM or hot spot temperature ? Di you try to see if it downclocks with the fans at 100% ?

The core temperature of the graphics card is 76.3 °C, and the hot spot reached 89.5 °C. FurMark was running for 10 minutes with the settings visible in the screenshot. Although the maximum clocks in MSI Afterburner are set to 2040 MHz, under the FurMark load, the GPU clock dropped to 1815 MHz. Could replacing thermal pads help with this?

 

[:D:D]

@Oliveros that drop with furmark is due to power limiting, not thermal limiting. Card can be set up to 150W, defualt is usually 125W. Use cmd and "nvidia-smi -q -d power" to check.

 

[AnotherReader]

You shouldn't be using Furmark to judge the ability of your GPU to hold its clocks. As @:D:D said, it will not allow your GPU to reach its maximum clocks, because its power limit would be reached well before that. Run a test like 3D Mark or better yet see what clocks are in actual games that you play.

 

[Lew Zealand]

Do not use Furmark as it doesn't simulate a gaming load and reputedly some cards specifically clock down more if they detect it. Put Time Spy on a Graphics Test 2 loop and see what your card does at stock and then at +135 MHz on the cores and post that. I do this testing all the time and this is my preferred way to start assessing the clock speed capabilities of a GPU.

FYI Nvidia GPUs start at a relatively high clock when cold and slowly drop down speed bins as they warm up to around 83°C. Above that they drop speed precipitously to protect the GPU. So your starting Core clock may be 2010 MHz at 35°C and be 1920 MHz by 75°C, which is normal behavior.

 

[:D:D]

@Oliveros I have the same card

Different VBIOS and VRAM manufacturer, also runs hot and it doesn't help with room temps over 30C. Seems like you got a good one. If the drops are power then not much to do other than lower clocks/voltage if it's a worry. If it's boost/temperature drop then your only losing a very small percentage, not something to find annoying when gaming IMO as it probably won't make a lot of difference. I bought mine second hand for testing some stuff and it's so far turned out an interesting project. Unlike gaming testing is best done if clocks can remain the same and it does seem able to do that.

Some heaven with no clock change from 44C to 80C.

Even more interesting and dangerous is power level and temp limit can be exceeded without throttling/dropping clocks. A quick test with heaven showed that although this card has a max limit of 150W it claimed to draw 190W. Certainly wouldn't want furmark running. Not sure if it's just bugged with this card or newer ones too, don't have anything newer to check with.

A quick run at 200W, quick because card heats up too fast.