The fan settings in the AWCC that you changed from Balanced to Performance only control the bottom front intake fan and the top exhaust fan. The GPU fans are not controlled by AWCC and that is not clear to anyone from the way the AWCC software looks. Default out of the box, the GPU fans are controlled by the GPU and will run automatically and the throttling you had is a normal reaction of the GPU when it gets hot, you didn't damage anything. If, however, you want to have the GPU fans run at a more aggressive speed more of the time, you can download various GPU fan control apps (some free, some payware). Many folks use MSI Afterburner and in there you can set up a custom fan curve for the 3090 so it runs at a faster speed more often to keep the temps down a bit. Know that the Aurora cases in general are warmer due to their smaller size and only a few fans as compared to larger cases that have 4, 6, or even more fans. I have an i-9 and 3090 equipped R11 and have not had thermal throttling you describe and I made no change to the GPU fan control speeds and usually run microsoft flight simulator in 4k or VR. I used MSI Afterburner to log the temperatures of the CPU, GPU, memory, and fan speeds and usually the temps of the GPU are around 72C while gaming (similar to the mid 70s you saw but I do run the AWCC system fans in Performance mode and not Balanced). I think your system is fine from what you describe but it's worth filing a Dell Support ticket while still inside your return window and then you also run a hardware check that will stress the GPU and give you a report on it to be sure (google that and there is a hardware test/check you can run from the Bios screen I believe). From what I have read the GPU ram will be in the 90s and that's normal for this card running at nearly 100% usage especially in a smaller/warmer case like the R10/R11/R12, but it makes sense to run that hardware check and file a ticket to be on the record. On a return, I have done that inside of 30days from invoice and it was no problem but I had to wait another 6 weeks then to get another PC as I wanted to change up from a 3060ti to a 3090. Good luck and research some more on reddit, here, etc. on the GPU memory temps to get a sense for that but I don't think it's unusual to be near 100 as long as the overall GPU temp is low 70s. To be safe you could also return it for a full refund and just order a new one and wait out the lead times but have that peace of mind.
There's 3 thermal read outs you should keep an eye on for the RTX 3000 series.
A free utility like GPU-Z will display them, and log them so you can always see min/max/average temps.
The 3 readouts are:
- GPU temperature
- GPU hotspot temperature
- Memory Tjunction temperature (sometimes just called Vram temperature)
Each have their own sensing device.
Nvidia considers the memory Tjunction to be ok up to 110 Celsius on the GDDR6 x memory cells.
Personally I don't like any temperature to ever be higher than 90 Celsius on electronic components. 100 Celsius is where water boils and turns into steam.
I found by error and trial that the best way to keep all 3 temperatures in check is by using a free aftermarket utility called Afterburner for the video card fans, and the DELL supplied control center for the case fans.
By adjusting the fan curves on the 2 video card fans using afterburner, and the fan curves on the 2 case fans from the AWWC, the temperatures come way down without generating excessive fan noise.
I was able to bring my Tjunction down from mid 90's to 82 maximum. It will require some error and trial to figure out the best fan configuration. Also note that with the RTX 3000 cards you can control each fan on the video card separately. Running 1 faster than the other might prove more beneficial than running them both at the same speed. Error and trial is best method to balance them out.
r72019
6 Professor
•
5.3K Posts
0
April 21st, 2021 21:00
"The return window is 30 days, correct? I don't want to be stuck waiting for months on another 3090 or system issues..."
30 days from invoice (ship) date, not 30 days from receipt like amazon for example.
Caustik
2 Posts
0
April 22nd, 2021 03:00
ok, so that would be the 24th for me. i have 2 days to figure this out. ultimately, i want a computer that works with an i9 and a 3090.
tempestornado23
37 Posts
0
April 22nd, 2021 05:00
The fan settings in the AWCC that you changed from Balanced to Performance only control the bottom front intake fan and the top exhaust fan. The GPU fans are not controlled by AWCC and that is not clear to anyone from the way the AWCC software looks. Default out of the box, the GPU fans are controlled by the GPU and will run automatically and the throttling you had is a normal reaction of the GPU when it gets hot, you didn't damage anything. If, however, you want to have the GPU fans run at a more aggressive speed more of the time, you can download various GPU fan control apps (some free, some payware). Many folks use MSI Afterburner and in there you can set up a custom fan curve for the 3090 so it runs at a faster speed more often to keep the temps down a bit. Know that the Aurora cases in general are warmer due to their smaller size and only a few fans as compared to larger cases that have 4, 6, or even more fans. I have an i-9 and 3090 equipped R11 and have not had thermal throttling you describe and I made no change to the GPU fan control speeds and usually run microsoft flight simulator in 4k or VR. I used MSI Afterburner to log the temperatures of the CPU, GPU, memory, and fan speeds and usually the temps of the GPU are around 72C while gaming (similar to the mid 70s you saw but I do run the AWCC system fans in Performance mode and not Balanced). I think your system is fine from what you describe but it's worth filing a Dell Support ticket while still inside your return window and then you also run a hardware check that will stress the GPU and give you a report on it to be sure (google that and there is a hardware test/check you can run from the Bios screen I believe). From what I have read the GPU ram will be in the 90s and that's normal for this card running at nearly 100% usage especially in a smaller/warmer case like the R10/R11/R12, but it makes sense to run that hardware check and file a ticket to be on the record. On a return, I have done that inside of 30days from invoice and it was no problem but I had to wait another 6 weeks then to get another PC as I wanted to change up from a 3060ti to a 3090. Good luck and research some more on reddit, here, etc. on the GPU memory temps to get a sense for that but I don't think it's unusual to be near 100 as long as the overall GPU temp is low 70s. To be safe you could also return it for a full refund and just order a new one and wait out the lead times but have that peace of mind.
Vanadiel
6 Professor
•
7.1K Posts
0
April 22nd, 2021 06:00
There's 3 thermal read outs you should keep an eye on for the RTX 3000 series.
A free utility like GPU-Z will display them, and log them so you can always see min/max/average temps.
The 3 readouts are:
- GPU temperature
- GPU hotspot temperature
- Memory Tjunction temperature (sometimes just called Vram temperature)
Each have their own sensing device.
Nvidia considers the memory Tjunction to be ok up to 110 Celsius on the GDDR6 x memory cells.
Personally I don't like any temperature to ever be higher than 90 Celsius on electronic components. 100 Celsius is where water boils and turns into steam.
I found by error and trial that the best way to keep all 3 temperatures in check is by using a free aftermarket utility called Afterburner for the video card fans, and the DELL supplied control center for the case fans.
By adjusting the fan curves on the 2 video card fans using afterburner, and the fan curves on the 2 case fans from the AWWC, the temperatures come way down without generating excessive fan noise.
I was able to bring my Tjunction down from mid 90's to 82 maximum. It will require some error and trial to figure out the best fan configuration. Also note that with the RTX 3000 cards you can control each fan on the video card separately. Running 1 faster than the other might prove more beneficial than running them both at the same speed. Error and trial is best method to balance them out.