- Joined
- Sep 24, 2020
- Messages
- 145 (0.08/day)
System Name | Room Heater Pro |
---|---|
Processor | i9-13900KF |
Motherboard | ASUS ROG STRIX Z790-F GAMING WIFI |
Cooling | Corsair iCUE H170i ELITE CAPELLIX 420mm |
Memory | Corsair Vengeance Std PMIC, XMP 3.0 Black Heat spreader, 64GB (2x32GB), DDR5, 6600MT/s, CL 32, RGB |
Video Card(s) | Palit GeForce RTX 4090 GameRock OC 24GB |
Storage | Kingston FURY Renegade Gen.4, 4TB, NVMe, M.2. |
Display(s) | ASUS ROG Swift OLED PG48UQ, 47.5", 4K, OLED, 138Hz, 0.1 ms, G-SYNC |
Case | Thermaltake View 51 TG ARGB |
Power Supply | Asus ROG Thor, 1200W Platinum |
Mouse | Logitech Pro X Superlight 2 |
Keyboard | Logitech G213 RGB |
VR HMD | Oculus Quest 2 |
Software | Windows 11 23H2 |
I recently received a Gigabyte 3080 Vision OC as a birthday present, and it replaced my Asus Dual RTX 2080. The card is very nice, I'm very happy with it, except for one thing: it uses a lot of power and generates a lot of heat when idle, at the Windows desktop, compared to the 2080. As a result, the fan almost never stops, as it has to work almost all the time to keep the temperature under 50 degrees Celsius.
According to the Gigabyte tuning/monitoring tools the memory remains at full speed, 19,004 MHz (1188 MHz x 16), even when idle, and the power usage when idle is 21% (22% with the fan running). I'm not entirely sure, but assuming this 21-22% refers to its 350W TDP, it means it wastes about 75W of power when idle, which seems a bit extreme.
After noticing the card power usage actually goes down to just a couple of percent when the displays are turned off by Windows I did some testing this morning, and I discovered the high power usage is caused by having 3 displays connected to my computer:
Display 1: 3440 x 1440 @ 100 Hz - DisplayPort
Display 2: 3840 x 2160 @ 60 Hz - DisplayPort
Display 3: 3840 x 2160 @ 60 Hz HDR - HDMI (OLED TV), it's usually turned off.
After disconnecting some of the displays, the memory speed started to drop as low as 810 MHz (51 MHz x 16) when idle, and the power usage dropped to 2-5%, so probably somewhere around 12W, which is fine, the number is similar to what I see in the 3080 reviews from TPU. This was surprising to me. I'm aware that video cards tend to use more power in multi-monitor setups, but I didn't expect it to use 6 times more power when idle, 75W, just because I have 3 displays, one of which is turned off.
The TPU reviews were also confusing in this regard, as I didn't realize that the multi-monitor power usage and the memory clock from the Clock Profiles section only applies to the specific configuration used during the TPU multi-monitor test (2 monitors, 1920x1080 and 1280x1024), especially since all the numbers looked the same when not gaming, on all 3080 cards:
And, in my case, those clocks were higher when idle, the memory actually running at full speed. Anyway, @W1zzard clarified this to me in 6900 XT thread, here: https://www.techpowerup.com/forums/threads/amd-radeon-rx-6900-xt.275652/page-6#post-4410461
Since then, I did a lot of testing to determine exactly what exactly about my display configuration causes the high power usage when idle. The conclusion is that the high resolutions, high refresh rates, and having HDR enabled all contribute to the power usage.
I was sure that the 3rd display, the TV, didn't contribute to this, since I keep it off. I was wrong. Even though the other displays flicker when I turn the TV on or off, it's not actually ignored by the video card when off, in fact it doesn't matter if it's on or off, the power usage is the same, unless I disconnect the HDMI cable physically.
Anyway, these are some of the test results:
The first thing I attempted was to disable HDR on the TV. That reduced the power usage of the video card from 21% to 8%, more than 2 times. Also the GPU and memory frequencies started to match those from TPU reviews.
After re-enabling HDR on the TV, I reduced the refresh rate on my primary display from 100 Hz to something lower, to see if it would help. It did. Reducing the refresh rate on my primary monitor to 50Hz had the same effect as disabling HDR on the TV, the power usage dropped to 8%.
I then set my primary display back to 100Hz, kept the TV with HDR enabled as well, and I reduced the refresh rate on my secondary 4K monitor from 60Hz to 30Hz. This worked even better, power usage got as low as 5%.
I then started experimenting with reducing resolutions on the TV and monitors, while keeping the maximum refresh rates, and HDR enabled on the TV. This helped as well, so clearly the resolution is also a factor, not just refresh rates and HDR.
Then I physically disconnected my secondary display and the TV, keeping just my primary display, and the power usage got even lower, oscillating between 2% and 5%.
Then I reduced the refresh rate and resolution on the remaining monitor as low as 800x600@60Hz. The power usage kept oscillating between 2% and 5%, but it was more frequently 2%, at the lower resolutions. When it was indicating 2%, the GPU voltage indicated 0, and the GPU frequency was also 0.
It seems after a certain point, after reducing all the GPU and memory frequencies at the minimum, it actually started to fully power down the GPU, intermittently, probably between two screen refreshes, to reduce power usage further. Only the memory remains powered, for obvious reasons, and when that happens the power usage is only 2%, which would probably be about 7W. This is how it looks in the monitoring tool when it's powering down the GPU periodically:
So, to conclude, I need to disconnect the HDMI cable from the TV when I don't need it, or at least disable HDR, to make the power usage reasonable.
According to the Gigabyte tuning/monitoring tools the memory remains at full speed, 19,004 MHz (1188 MHz x 16), even when idle, and the power usage when idle is 21% (22% with the fan running). I'm not entirely sure, but assuming this 21-22% refers to its 350W TDP, it means it wastes about 75W of power when idle, which seems a bit extreme.
After noticing the card power usage actually goes down to just a couple of percent when the displays are turned off by Windows I did some testing this morning, and I discovered the high power usage is caused by having 3 displays connected to my computer:
Display 1: 3440 x 1440 @ 100 Hz - DisplayPort
Display 2: 3840 x 2160 @ 60 Hz - DisplayPort
Display 3: 3840 x 2160 @ 60 Hz HDR - HDMI (OLED TV), it's usually turned off.
After disconnecting some of the displays, the memory speed started to drop as low as 810 MHz (51 MHz x 16) when idle, and the power usage dropped to 2-5%, so probably somewhere around 12W, which is fine, the number is similar to what I see in the 3080 reviews from TPU. This was surprising to me. I'm aware that video cards tend to use more power in multi-monitor setups, but I didn't expect it to use 6 times more power when idle, 75W, just because I have 3 displays, one of which is turned off.
The TPU reviews were also confusing in this regard, as I didn't realize that the multi-monitor power usage and the memory clock from the Clock Profiles section only applies to the specific configuration used during the TPU multi-monitor test (2 monitors, 1920x1080 and 1280x1024), especially since all the numbers looked the same when not gaming, on all 3080 cards:
And, in my case, those clocks were higher when idle, the memory actually running at full speed. Anyway, @W1zzard clarified this to me in 6900 XT thread, here: https://www.techpowerup.com/forums/threads/amd-radeon-rx-6900-xt.275652/page-6#post-4410461
Since then, I did a lot of testing to determine exactly what exactly about my display configuration causes the high power usage when idle. The conclusion is that the high resolutions, high refresh rates, and having HDR enabled all contribute to the power usage.
I was sure that the 3rd display, the TV, didn't contribute to this, since I keep it off. I was wrong. Even though the other displays flicker when I turn the TV on or off, it's not actually ignored by the video card when off, in fact it doesn't matter if it's on or off, the power usage is the same, unless I disconnect the HDMI cable physically.
Anyway, these are some of the test results:
Code:
GPU Memory Power
MHz MHz Used Display 1 Display 2 Display 3
225 19,004 21% 3440 x 1440 @ 100Hz 3840 x 2160 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
210 810 8% 3440 x 1440 @ 100Hz 3840 x 2160 @ 60Hz 3840 x 2160 @ 60Hz HDR=off
225 19,004 21% 3440 x 1440 @ 95Hz 3840 x 2160 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 90Hz 3840 x 2160 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 85Hz 3840 x 2160 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 80Hz 3840 x 2160 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 60Hz 3840 x 2160 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
210 1,820 8% 3440 x 1440 @ 50Hz 3840 x 2160 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
210 1,820 5% 3440 x 1440 @ 100Hz 3840 x 2160 @ 30Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 3840 x 2160 @ 60Hz 3440 x 1440 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 3840 x 2160 @ 60Hz 2560 x 1600 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 3840 x 2160 @ 60Hz 2560 x 1440 @ 60Hz HDR=on
210 1,820 8% 3440 x 1440 @ 100Hz 3840 x 2160 @ 60Hz 2048 x 1536 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 3440 x 1440 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 2560 x 1600 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 2560 x 1440 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 2048 x 1536 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 1920 x 1440 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 1920 x 1200 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 1680 x 1050 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 1600 x 1200 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 1600 x 1024 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 19,004 21% 3440 x 1440 @ 100Hz 1600 x 900 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 18,504 21% 3440 x 1440 @ 100Hz 1440 x 900 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 10,004 18% 3440 x 1440 @ 100Hz 1366 x 768 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
225 10,004 18% 3440 x 1440 @ 100Hz 1280 x 768 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
210 1,820 8% 3440 x 1440 @ 100Hz 1176 x 664 @ 60Hz 3840 x 2160 @ 60Hz HDR=on
210 810 5% 3440 x 1440 @ 100Hz 800 x 600 @ 56Hz 3840 x 2160 @ 60Hz HDR=on
210 810 5% 3440 x 1440 @ 100Hz disconnected 3840 x 2160 @ 60Hz HDR=on
210 810 2-5% 3440 x 1440 @ 100Hz disconnected disconnected
210 810 2-5% 3440 x 1440 @ 50Hz disconnected disconnected
210 810 2-5% 2560 x 1440 @ 50Hz disconnected disconnected
210 810 2-5% 1920 x 1440 @ 50Hz disconnected disconnected
210 810 2-5% 1920 x 1080 @ 50Hz disconnected disconnected
210 810 2-5% 1280 x 768 @ 50Hz disconnected disconnected
210 810 2-5% 800 x 600 @ 60Hz disconnected disconnected
The first thing I attempted was to disable HDR on the TV. That reduced the power usage of the video card from 21% to 8%, more than 2 times. Also the GPU and memory frequencies started to match those from TPU reviews.
After re-enabling HDR on the TV, I reduced the refresh rate on my primary display from 100 Hz to something lower, to see if it would help. It did. Reducing the refresh rate on my primary monitor to 50Hz had the same effect as disabling HDR on the TV, the power usage dropped to 8%.
I then set my primary display back to 100Hz, kept the TV with HDR enabled as well, and I reduced the refresh rate on my secondary 4K monitor from 60Hz to 30Hz. This worked even better, power usage got as low as 5%.
I then started experimenting with reducing resolutions on the TV and monitors, while keeping the maximum refresh rates, and HDR enabled on the TV. This helped as well, so clearly the resolution is also a factor, not just refresh rates and HDR.
Then I physically disconnected my secondary display and the TV, keeping just my primary display, and the power usage got even lower, oscillating between 2% and 5%.
Then I reduced the refresh rate and resolution on the remaining monitor as low as 800x600@60Hz. The power usage kept oscillating between 2% and 5%, but it was more frequently 2%, at the lower resolutions. When it was indicating 2%, the GPU voltage indicated 0, and the GPU frequency was also 0.
It seems after a certain point, after reducing all the GPU and memory frequencies at the minimum, it actually started to fully power down the GPU, intermittently, probably between two screen refreshes, to reduce power usage further. Only the memory remains powered, for obvious reasons, and when that happens the power usage is only 2%, which would probably be about 7W. This is how it looks in the monitoring tool when it's powering down the GPU periodically:
So, to conclude, I need to disconnect the HDMI cable from the TV when I don't need it, or at least disable HDR, to make the power usage reasonable.