F5F Stay Refreshed Hardware Desktop Driver power state issue during idle after balancing power plan and overclocking

Driver power state issue during idle after balancing power plan and overclocking

Driver power state issue during idle after balancing power plan and overclocking

A
Athame_
Senior Member
734
07-26-2017, 12:03 PM
#1
Details: CPU i7-7700k @ 4.20 GHz (boost: 4.5GHz) [delidded]
Motherboard Asus Maximus IX Extreme
GPU Gigabyte GTX 1080 Ti Waterforce WB Xtreme Edition 11G RAM: 32GB G.Skill Trident Z RGB DDR4 3600MHz - 32GB (8x4) - 16-16-16-36 PSU Seasonic PRIME 850W Titanium Custom Cooling x2 radiators (520x30x60mm) + 8 Noctua fans, EK-XRES 140 Revo D5 pump
After several trials, I tested these settings: Vcore 1.430V, cache ratio 46%, AVX offset 0, XMP enabled, DDR4 3600MHz, 32GB RAM (8x4), 16-16-16-36 PSU.
I ran a 12-hour stress test with Prime95 in blend mode; idle performance was flawless.
Later I switched Windows power plan to balanced and observed unstable behavior: CPU clock fluctuated wildly (800MHz–5GHz), cache clock between 800MHz–4.6GHz, Vcore steady. Cache always lagged behind CPU by about 300 MHz.
When idle in balanced mode for 5–10 minutes, the system crashed repeatedly with BSOD and DRIVER_POWER_STATE_FAILURE.
It appears unstable clocks cause instability even though voltages remain stable. Any advice?
A
Athame_
07-26-2017, 12:03 PM #1

Details: CPU i7-7700k @ 4.20 GHz (boost: 4.5GHz) [delidded]
Motherboard Asus Maximus IX Extreme
GPU Gigabyte GTX 1080 Ti Waterforce WB Xtreme Edition 11G RAM: 32GB G.Skill Trident Z RGB DDR4 3600MHz - 32GB (8x4) - 16-16-16-36 PSU Seasonic PRIME 850W Titanium Custom Cooling x2 radiators (520x30x60mm) + 8 Noctua fans, EK-XRES 140 Revo D5 pump
After several trials, I tested these settings: Vcore 1.430V, cache ratio 46%, AVX offset 0, XMP enabled, DDR4 3600MHz, 32GB RAM (8x4), 16-16-16-36 PSU.
I ran a 12-hour stress test with Prime95 in blend mode; idle performance was flawless.
Later I switched Windows power plan to balanced and observed unstable behavior: CPU clock fluctuated wildly (800MHz–5GHz), cache clock between 800MHz–4.6GHz, Vcore steady. Cache always lagged behind CPU by about 300 MHz.
When idle in balanced mode for 5–10 minutes, the system crashed repeatedly with BSOD and DRIVER_POWER_STATE_FAILURE.
It appears unstable clocks cause instability even though voltages remain stable. Any advice?

P
PinkSky_xoxo
Member
64
07-26-2017, 12:03 PM
#2
Using the Balanced power plan with random CPU speed adjustments usually requires higher voltage for stability. Locking the CPU to a fixed frequency and activating core C states provides better performance, especially during overclocking. This method works well compared to the Balanced plan when pushing speeds. An example shows cores running at 5000 MHz but with significant idle power savings. Enabling C states keeps cores idle at low voltage and clock speed, reducing heat and power use. The Ring Down Bin feature helps maintain cache speed without compromising stability.
P
PinkSky_xoxo
07-26-2017, 12:03 PM #2

Using the Balanced power plan with random CPU speed adjustments usually requires higher voltage for stability. Locking the CPU to a fixed frequency and activating core C states provides better performance, especially during overclocking. This method works well compared to the Balanced plan when pushing speeds. An example shows cores running at 5000 MHz but with significant idle power savings. Enabling C states keeps cores idle at low voltage and clock speed, reducing heat and power use. The Ring Down Bin feature helps maintain cache speed without compromising stability.

P
178
07-26-2017, 12:03 PM
#3
What method are you employing to interpret the C-states?
P
ProSkillsNinja
07-26-2017, 12:03 PM #3

What method are you employing to interpret the C-states?

D
DL_Zatlin
Junior Member
3
07-26-2017, 12:03 PM
#4
Shorten the speed or pause operation.
D
DL_Zatlin
07-26-2017, 12:03 PM #4

Shorten the speed or pause operation.

S
Solestial
Junior Member
10
07-26-2017, 12:03 PM
#5
I conducted additional tests and experimented with Windows power plan configurations. My initial thought was that the BSOD (DRIVER_POWER_STATE_FAILURE) occurred because the CPU clock fluctuated wildly between 800MHz and 5GHz when the balanced power plan was active. To test this, I adjusted the Minimum and Maximum processor states, gradually increasing the CPU frequency by 100MHz increments. Initially, at 900MHz idle, the system still triggered the BSOD. Then, I tried running the PC in idle at higher frequencies—900MHz, 1GHz, up to 2GHz—but the issue persisted. Eventually, I tested at 4.9GHz through to 2.1GHz and observed stability without BSOD. This led me to believe that the instability likely arises at lower clock speeds. I hypothesized that as the CPU clock stabilized around a certain point, the system would become more reliable. To explore further, I increased the CPU frequency in steps of 100MHz and monitored performance. At 2.1GHz running Prime95, the PC remained stable for several hours. I also noted specific technical details: the iGPU clock ranges from 350MHz to 1.15GHz, while the GT/OC limits are set at x23 (double 1.15GHz). The Uncore and ring/LLC ratios consistently doubled the iGPU frequency, suggesting a minimum ratio requirement. I observed that when the iGPU clock was 450MHz, the ring ratio equaled x9; at 850MHz it reached x17. This indicated the iGPU needed an uncore ratio and ring/llc clock at least double its CPU frequency. When I adjusted the CPU to 2.2GHz and ran FurMark, the ring ratio stabilized at x23, matching the iGPU’s actual clock. I also experimented with lowering the CPU clock to 800MHz and observed a similar ring ratio (x23). This reinforced my suspicion that the iGPU was the core issue. I also disabled the iGPU, kept the CPU at 800MHz idle for 30 minutes, and found stability without BSOD. In all cases, reverting BIOS settings to defaults helped—CPU clock fluctuated normally, VID values aligned with expectations, and no instability appeared. Notably, after resetting BIOS, HWiNFO64 no longer displayed the OC ratio limits for iGPU. This suggests a possible conflict or misconfiguration in the power plan or driver settings. Overall, it seems the problem is likely tied to the iGPU’s clock requirements rather than the CPU itself.
S
Solestial
07-26-2017, 12:03 PM #5

I conducted additional tests and experimented with Windows power plan configurations. My initial thought was that the BSOD (DRIVER_POWER_STATE_FAILURE) occurred because the CPU clock fluctuated wildly between 800MHz and 5GHz when the balanced power plan was active. To test this, I adjusted the Minimum and Maximum processor states, gradually increasing the CPU frequency by 100MHz increments. Initially, at 900MHz idle, the system still triggered the BSOD. Then, I tried running the PC in idle at higher frequencies—900MHz, 1GHz, up to 2GHz—but the issue persisted. Eventually, I tested at 4.9GHz through to 2.1GHz and observed stability without BSOD. This led me to believe that the instability likely arises at lower clock speeds. I hypothesized that as the CPU clock stabilized around a certain point, the system would become more reliable. To explore further, I increased the CPU frequency in steps of 100MHz and monitored performance. At 2.1GHz running Prime95, the PC remained stable for several hours. I also noted specific technical details: the iGPU clock ranges from 350MHz to 1.15GHz, while the GT/OC limits are set at x23 (double 1.15GHz). The Uncore and ring/LLC ratios consistently doubled the iGPU frequency, suggesting a minimum ratio requirement. I observed that when the iGPU clock was 450MHz, the ring ratio equaled x9; at 850MHz it reached x17. This indicated the iGPU needed an uncore ratio and ring/llc clock at least double its CPU frequency. When I adjusted the CPU to 2.2GHz and ran FurMark, the ring ratio stabilized at x23, matching the iGPU’s actual clock. I also experimented with lowering the CPU clock to 800MHz and observed a similar ring ratio (x23). This reinforced my suspicion that the iGPU was the core issue. I also disabled the iGPU, kept the CPU at 800MHz idle for 30 minutes, and found stability without BSOD. In all cases, reverting BIOS settings to defaults helped—CPU clock fluctuated normally, VID values aligned with expectations, and no instability appeared. Notably, after resetting BIOS, HWiNFO64 no longer displayed the OC ratio limits for iGPU. This suggests a possible conflict or misconfiguration in the power plan or driver settings. Overall, it seems the problem is likely tied to the iGPU’s clock requirements rather than the CPU itself.

P
PaigePlays
Member
173
07-26-2017, 12:03 PM
#6
The Ring Down Bin configuration in BIOS manages this behavior. Check if Ring Down Bin is activated in the BIOS settings. The VID voltage rises when the CPU is being loaded, regardless of whether it's capped at 800 MHz. A CPU under load requires higher voltage than when idle. I performed some tests and found that the setting fails when the CPU runs at 800 MHz with the iGPU enabled. Using GPU-Z Render Test confirmed the iGPU operates at 1200 MHz, boosting the cache ratio to 24, which raises the VID voltage from 0.6895 V to 0.7930 V. This could explain your instability at lower clock speeds. Speed Shift might offer a solution; locking the CPU to its maximum frequency seems preferable.
P
PaigePlays
07-26-2017, 12:03 PM #6

The Ring Down Bin configuration in BIOS manages this behavior. Check if Ring Down Bin is activated in the BIOS settings. The VID voltage rises when the CPU is being loaded, regardless of whether it's capped at 800 MHz. A CPU under load requires higher voltage than when idle. I performed some tests and found that the setting fails when the CPU runs at 800 MHz with the iGPU enabled. Using GPU-Z Render Test confirmed the iGPU operates at 1200 MHz, boosting the cache ratio to 24, which raises the VID voltage from 0.6895 V to 0.7930 V. This could explain your instability at lower clock speeds. Speed Shift might offer a solution; locking the CPU to its maximum frequency seems preferable.