Problems with low voltage – throttle stops occurring.
Problems with low voltage – throttle stops occurring.
I appreciate your feedback. Your advice is really useful. I plan to clean my laptop now following your tips. I've ordered fresh thermal paste, so I should be able to reapply it tomorrow. Right now I'll just do a quick clean. I also intend to set IccMax for the core and cache to their maximum values and check the outcomes after cleaning. Would you think a cooling stand could help with this machine? Any recommendations?
If XTU isn't reliable, your observation is accurate. The issue lies in how XTU reacts when another program alters the voltage register—XTU doesn’t notice these changes. The ThrottleStop monitoring table checks the CPU every second, so any software adjustments to the offset voltage will appear instantly. XTU has long been a source of inconsistencies. By keeping an eye on the voltage register, you’ll soon realize its behavior isn’t dependable. After a sleep resume cycle, it sometimes completely skips applying the offset voltages, no matter what it says. Consider using HWINFO to verify.
@Jurrunio Also check HWiNFO. I had to restart HWiNFO so it would update the voltage readings. HWiNFO doesn’t seem to refresh properly. The ThrottleStop table works without problems, updating right away. If HWMonitor shows values a bit off versus ThrottleStop, it’s likely due to an incorrect conversion method. It should divide by 1024, not 1000. Not sure why that happens. The ThrottleStop table appears to be the only one reporting offset voltages accurately and live.
Restarting Hwinfo changed the offset values it displayed, but the actual VID stays the same when adjusting the offset. It might be that your BIOS is interfering. The Clamp function in the power section appears to cause the core clocks to fluctuate excessively, which negatively impacts performance.
I wouldn’t rely on the stated VID voltage to check if an undervoltage issue existed. Instead, run Cinebench R23 for a proper evaluation. You’ll notice distinct performance changes when comparing 0 offset voltage with a -125 mV offset for both the core and cache. The core and cache readings don’t have to match exactly. When you set the cache to -125 mV, raise the core to -150 mV, then -175 mV and -200 mV, then test Cinebench at each level. Other users with 8750H devices have found improved results when the core voltage is higher than the cache and the CPU remains stable. Generally, a 50 mV to 75 mV variation yields optimal outcomes. Intel XTU doesn’t let you adjust separate voltage settings for core and cache. Here are some Cinebench findings shared by a previous user that support this approach. https://drive.google.com/file/d/1B2HZjwl...sp=sharing If you still doubt the voltage controls, keep adjusting both core and cache by -25 mV and perform a quick TS Bench each time. If the benchmark shows errors or a BSOD, the voltages are being applied correctly. When an Intel CPU hits power limits, the Clamp feature decides whether it can run slower than its base speed of 2.20 GHz for an 8750H. This setting is especially useful for low-power models. For testing, consider using ThrottleStop to reduce turbo limits to 15W and run a stress test. You can toggle Clamp on or off while the CPU is busy. If throttling occurs, monitor the MHz changes to see how enabling Clamp helps. Many HP laptops let you adjust power sliders, but you can’t exceed the internal power cap set by HP. There are three turbo limits: ThrottleStop and XTU offer two, while a third is managed by an embedded controller (EC). The CPU selects the lowest of these values. If MSR is set to 60W and EC to 45W, the EC takes control. Therefore, you should never surpass the device’s rated TDP during extended stress tests on many HP laptops. If you lower the MSR limit below the EC’s, the CPU will default to the EC’s setting.
What kind of thermal paste were you using? Monitor your temps closely over the coming weeks. Certain laptops, based on the paste you chose, can rapidly expel heat, leaving minimal gap between the heatsink and CPU. Some people have encountered problems with Arctic MX-4. Here’s an example of a Noctua NT-H1 that didn’t last long. Nice to hear your old machine is now better. Simple upkeep often resolves many issues.
I tried the Thermal Grizzly Cooling Paste. After nearly three hours of gaming, the temperature stayed below 75°F instead of reaching 95°F, which helped prevent thermal throttling.