Best throttle adjustments for Predator Helios PH315-53 i7 10870h with RTX 3060 and 32 GB RAM.
Best throttle adjustments for Predator Helios PH315-53 i7 10870h with RTX 3060 and 32 GB RAM.
Hey there, friend. It's been a while since we last chatted. I just got my new laptop and have been exploring it for nearly a year now. I've taken it apart, swapped out the old stock VRM paste sheets, and added thermal pads for both the CPU and GPU. Temperatures have stayed around 50-55°C, though they spike to about 100°C when I render in Blender or run a tough game. It seems like the hardware is handling it, but I'm curious about some settings.
I'm asking for your advice—should I keep the Thermal Velocity boost enabled? Some folks say it helps if you have a good cooler, while others think it can hurt performance as temps rise. I'm also wondering about VRM stress and ring-down. Can I safely bump the turbo limits while undervolting the CPU? And are there any tweaks in the TPL window that might help?
I'm thinking about adjusting the speedshift min-max to squeeze more performance out for my projects. Thanks a lot for your guidance and insights!
In the FIVR settings, ensure Thermal Velocity Boost is enabled if you wish your CPU to reduce speed to 100 MHz around 70°C. This idea originated from the marketing team. Merging "Velocity" and "Boost" gives the impression of a strong capability, but in practice it offers minimal impact. When activated, it leads to throttling instead of boosting, which isn’t very impressive. V-Max Stress can also bring the CPU down to 100 MHz when voltage rises too high. Your processor doesn’t allow overclocking, and undervolting it won’t change much. Clearing this option won’t help much. Opting to keep Ring Down Bin enabled is advised. It guarantees the ring bus (cache) stays at least 300 MHz slower than the CPU cores. Attempting to stabilize a 10th Gen CPU with matching core and core speeds is nearly impossible. For better reliability, keep Ring Down Bin checked. You don’t have a processor that supports overclocking. Raising turbo ratios beyond the default limits ThrottleStop from improving performance. The CPU will disregard any overclock attempts. You can adjust a few ratios to higher numbers than the defaults, but Intel no longer shares turbo data publicly, so some boards might misconfigure them. ThrottleStop reads the standard turbo values from the CPU, so trust those numbers. The figures under Turbo are accurate for your 10870H; another unit might vary slightly. Not releasing turbo data lets Intel tweak default values on the manufacturing side whenever they wish. Your opinions don’t matter. The top CPU boost multiplier for your 10870H is 50. Going beyond 50 won’t matter. Setting Speed Shift Max higher than 50 is unnecessary. Using the default Windows High Performance plan to run at peak speed will set Speed Shift EPP to 0 automatically. You can view the EPP settings on ThrottleStop’s main screen and adjust to 0 if desired, but it’s not essential. Windows usually handles this automatically. The rest of your configurations are solid. Core and cache voltage offsets aren’t required to match each other. Some users notice better Cinebench scores or cooler temps by adjusting core voltage higher than cache. Try the TS Bench 960M test. If it shows issues, lower cache undervoltage. If -85 mV triggers errors, reduce to -75 or -70 mV. Ignore any reported problems from the test—they’re close to instability. A tiny margin above stability is better than chasing perfection. Don’t invest time in finding the absolute ideal voltage settings. A small buffer helps the CPU perform consistently, regardless of workload. One crash is too many.
Thank you for your continued support @unclewebb! Your feedback is really helpful. I’ve kept all your suggestions in mind and will remember them. I’ve tested the 960M multiple times before adjusting undervolt settings, and the most recent run was successful. A long render of a complex scene in Blender often helps identify if my settings are too low. If you’re referring to glitches, I haven’t experienced any so far. Overall, the settings appear stable, but I’ll remember your advice about maintaining core and cache ratios.
If there are any errors in the calculations that the TS Bench does, instead of showing elapsed time at the top of the TS Bench, it will show 5 Errors or however many errors it finds. Here is an example. https://www.techpowerup.com/forums/threa...rs.278482/
Noted! I didn't realize TS would cause those issues until recently. Running the TS bench often led to screen glitches, BSODs (very rarely), or complete freezes. No such problems have occurred in a while. Thanks!
The problem can be resolved by turning off the virtual machine platform feature in Windows. This should fix the inability to use throttle stop.
Here’s a revised version of your message:
Thanks for the undervolt adjustments. I have a similar setup to yours and have also encountered heating problems. Following your recommendations, I reduced the cache voltage to -75 and achieved around 70°C during idle and up to 100°C under load. After reviewing your TS Bench test, I noticed my maximum CPU temperature is actually higher than yours, and I’m still unfamiliar with the factors affecting it. Could this be impacting performance? Additionally, I’ve been experiencing GPU power issues despite having a TDP of 105W (95W base + 10W boost). After pushing power limits, my GPU spiked but eventually settled around 50W. Is this related to thermal throttling? I recently applied Noctua NT-H2 paste to CPU, GPU, and VRAM, but it didn’t resolve the issue. Could you assist in finding a solution or suggest improvements?
Your machine is getting too hot and is throttling performance despite low power draw. To boost speed, you should address the cooling issue. I haven't tried it before. The Noctua NT-H2 works well, but the earlier paste NT-H1 caused overheating and temperature spikes. Honeywell PTM-7950 is a popular alternative with strong user feedback.
I rebuilt my CPU and GPU with Artic MX 6. Everything seems okay so far on the CPU, but I didn’t see any issues during the TS Bench. There were some lights in the limit window, but nothing critical. @unclewebb I should check if leaving the PL enabled causes problems. If I turn it off, the power warning lights go off. Should I adjust the power limits? Also, the CPU temperature improved, while the GPU stayed stable between 60-70°C. Once I start a program, it quickly hits 85°C, throttles, and the GPU power drops to 50W from 90W. Is there anything I can do? I’ve already swapped the thermal pads in the VRAMs with generic ones.