What's new

Surface Pro 4 i7-6650U power throttling issue?

See my post #12. It's not Microsoft but Intel rather. Just download intel power gadget, max the CPU with handbrake transcoding and you should see that the whole 15 W TDP is gobbled up by CPU alone. Throw in the GPU then you've got the power throttling mess.

Given the issue causing this as you have described, the hardware (Intel) design has limitations on load that impacts performance. Its hard for me to understand how this could not have been identified during product testing prior to releasing the chipset. Intel must have rolled the dice on the type i7 users knowing the potential risks caused by heavy load.
 
Given the issue causing this as you have described, the hardware (Intel) design has limitations on load that impacts performance. Its hard for me to understand how this could not have been identified during product testing prior to releasing the chipset. Intel must have rolled the dice on the type i7 users knowing the potential risks caused by heavy load.

It's just that the Skylake architecture is not efficient enough to sustain full CPU and GPU load to fit the 15W TDP. However , if Intel specified this with a 28 W indefinite TDP like those on the CPUs with iris 550 GPU, there wouldn't be power throttling at all.
 
If I were to limit the i7's max frequency to 1.7Ghz, is it possible that would open up enough room to run the IGP at max frequency without encountering throttling?
 
Last edited:
I would think that's the case. Since it's skylake you can go 2 GHz or even its base clock 2.2 GHz limit without power issues.
 
This is a philosophical issue.
In Intel's view... providing components that exceed the (15W, 28W, ...) thermal envelope when used at max capability is more efficient and meets more use case scenarios than the alternative of only providing a CPU and GPU that together would never exceed xW of power. I happen to agree with that philosophy.

Additionally you could say the same for any given tablet computer design. using a SoC or combination of CPU GPU and other components that all together at max load may exceed the thermal envelope of the case/heat transfer capacity yields a system that is more powerful and meets more use cases than a system that is built with the constraint to always live under a specific thermal ceiling.

This is the current state of modern computer design. Every Intel laptop has the same parts to work with and if you really look you'll find most if not all throttle because they know the math and tradeoffs as well. To make a device which doesn't throttle it has to have a certain size and airflow to dissipate the heat but in the thin and light era they couldn't sell a device that's thick enough or with high velocity fans to move enough air to prevent throttling. Oh by the way same goes for ARM based units.

{trips and falls from soapbox, sprains wrist and ankle, limps to the exit}
 
Have there been more benchmark reports about the i7 performance with the external fan? If that is all it takes to prevent significant CPU throttling then it is (to me) a pretty moot concern so long as there isn't damage being done to components when using the fan for long periods of time. Seems like folks are saying the CPU will throttle due to not enough power but other reports say that so long as the temperature sensor is cooled the CPU/GPU runs well.

I wonder if MS was overly sensitive to criticism of heat in the SP3 and set that sensor up to prevent the SP4 from getting "too hot" from a subjective POV even if it limits performance that is otherwise acceptable.
 
Have there been more benchmark reports about the i7 performance with the external fan? If that is all it takes to prevent significant CPU throttling then it is (to me) a pretty moot concern so long as there isn't damage being done to components when using the fan for long periods of time. Seems like folks are saying the CPU will throttle due to not enough power but other reports say that so long as the temperature sensor is cooled the CPU/GPU runs well.

I wonder if MS was overly sensitive to criticism of heat in the SP3 and set that sensor up to prevent the SP4 from getting "too hot" from a subjective POV even if it limits performance that is otherwise acceptable.

I'm not even sure external fans make any difference at all.

I just ran a POVRay rendering test. The file I use for rendering has resulted in CPU temps maxing out at around 75 degrees C.

Then I ran it with a laptop cooler under the entire bottom of the case, running full blast. It made no difference at all.

What's more, according to HWinfo, there was no thermal throttling evidenced either way. Only power throttling (and the CPU frequency still held steady at ~3100 Mhz).

Granted, this was a CPU-only test, not a CPU plus GPU test. But considering the Tjmax is 100 degrees C, and considering that the SP4 design already seems superb from a cooling standpoint, I have trouble seeing the impact of external cooling.
 
The youtube video linked in post #24 suggests an external fan keeps framerates from dropping on the i7. Who knows what it is actually doing though. Maybe there are two throttle sources, one power and one thermal, and they manifest under different circumstances. My concern is for gaming, where the fan seems to have a substantial benefit, at least short term.
 
It may very well be that the situation is different for GPU-heavy tasks - though since they're on the same die and should benefit overall from the same cooling system, that seems odd to me. Would like to know more about what's going on, as well.
 
I downloaded and installed a "turbo boost" monitor and my i7 performed flawlessly. Sadly enough, it seems hoards of people aren't having a good experience with the issued chipset / performance. If time and patience isn't a concern - I suggest a reload.... Took care of all my initial outta box issues
 
Anyone try turning hyperthreading off, and downclocking the CPU to like 1.7GHz or 1.5GHz to give the GPU more power room to breathe?
 
Back
Top