Intel Baseline Profile Yields Odd Power/Performance On Linux

Written by Michael Larabel in Motherboards on 1 May 2024 at 08:14 AM EDT. Page 2 of 4. 31 Comments.
AOM AV1 benchmark with settings of Encoder Mode: Speed 11 Realtime, Input: Bosphorus 1080p. Default was the fastest.

Right from the start the performance of the Core i9 14900K when operating in the Intel Baseline Profile mode was lower than the ASUS defaults...

AOM AV1 benchmark with settings of Encoder Mode: Speed 11 Realtime, Input: Bosphorus 1080p. Default was the fastest.

But what caught me by surprise was the Intel Baseline Profile CPU power consumption peaking higher than the ASUS defaults.... Huh?

AOM AV1 benchmark with settings of Encoder Mode: Speed 6 Two-Pass, Input: Bosphorus 1080p. Default was the fastest.
AOM AV1 benchmark with settings of Encoder Mode: Speed 6 Two-Pass, Input: Bosphorus 1080p. Default was the fastest.
AOM AV1 benchmark with settings of Encoder Mode: Speed 6 Two-Pass, Input: Bosphorus 1080p. Default was the fastest.

The Intel Baseline Profile performance was slightly lower during AV1 video encoding but with the Intel Baseline Profile CPU power consumption actually hitting higher than with the ASUS default configuration... In turn just a wreck for performance-per-Watt.

AOM AV1 benchmark with settings of Encoder Mode: Speed 6 Two-Pass, Input: Bosphorus 1080p. Default was the fastest.

At first I was thinking that perhaps the Intel Run-Time Average Power Limiting (RAPL) / PowerCap counters were somehow being skewed when in the Intel Baseline Profile mode for yielding incorrect CPU power consumption numbers. But the CPU temperature was also trending higher when running with the Intel Baseline Profile settings engaged.

C-Ray benchmark with settings of Resolution: 5K, Rays Per Pixel: 16. Default was the fastest.
C-Ray benchmark with settings of Resolution: 5K, Rays Per Pixel: 16. Default was the fastest.
C-Ray benchmark with settings of Resolution: 5K, Rays Per Pixel: 16. Default was the fastest.
C-Ray benchmark with settings of Resolution: 5K, Rays Per Pixel: 16. Default was the fastest.

With the basic C-Ray benchmark, the Intel Baseline Profile mode was slower as expected but again with higher CPU power consumption than at the default (more performant) mode. As you can see with the power graph here the CPU power consumption ends up fluctuating wildly where as in the default mode was more consistent in this Linux test. In turn the CPU peak frequency was fluctuating wildly in the Intel Baseline Profile mode and this also led to big swings for the CPU temperature.

uvg266 benchmark with settings of Video Input: Bosphorus 1080p, Video Preset: Ultra Fast. Default was the fastest.
uvg266 benchmark with settings of Video Input: Bosphorus 1080p, Video Preset: Ultra Fast. Default was the fastest.
uvg266 benchmark with settings of Video Input: Bosphorus 1080p, Video Preset: Ultra Fast. Default was the fastest.
uvg266 benchmark with settings of Video Input: Bosphorus 1080p, Video Preset: Ultra Fast. Default was the fastest.

The frequent fluctuations when using the Intel Baseline Profile mode led to a higher CPU power consumption on average and worse power efficiency than the sustained performance in the default mode.


Related Articles