Asaggynoodle
Long Term Lurker
Alright, so a bit of a joke on the second half!
I spent a few hours last night pondering some cooling improvements to the Pyra's CPU board to mitigate throttling and improve device performance.
Few things I wanted to nip at the bud are:
-Throttling after about 10 minutes of heavy use. > Pushes clock into a downclock decay of 250Mhz, then cools, but immediately hits trip temperature back down to 250Mhz making system crippled.
-Aggregate heat buildup inside the device (thermal sink) which leads to heating up the battery and making the device quite warm. > Reduced battery life, likely bloating long term and overall just heat buildup throughout.
-Open up the POSSIBILITY to squeeze more performance out of the CPU both in regards to CPU Clock, and investigate other areas.. Such as DDR timings, and frequency, as well as SGX clock frequency (which go hand and hand as SGX uses system memory).
-Evaluate possibility of lowering core voltage through the device tree source file across (hopefully) all clock domains. As well as add more intermediate stepping (125Mhz) for the On Demand perf level.
While yet no radical redesigns have been implemented YET, I have started taking some measures to pave the path to future cooling versions and designs.
-Conformally coated the CPU board components to mitigate any sort of electrical shorting from components due to clearance of heatsink, and more aggressive (potentially) electrically conductive thermal pastes/interface materials.
-Soldered the existing heatsink (copper plate) to the main thermal element (the serrated square piece of copper). The goal here is to maximize the performance of the stock solution to get a baseline as to what is needed aftermarket to achieve the above goals.
-Removed the stock thermal glue/material from the heatsink. There seems to be an insulating adhesive coating the contact side of the heatsink, which likely significantly impacts thermal transfer from the CPU/PMC to the heatsink.
-Applied extremely high end thermal compound to heatsink and RAM/CPU/PMC chips. I considered using a graphite thermal pad here, but the tolerance/spacing of the board needs to be filled to do so to support a forceful contract on CPU/RAM/PMC.
All of this so far has yielded about a 20-23*C temperature delta from stock. Which now makes 1.5Ghz essentially perfectly usable outside of long term heavy load when the device becomes saturated.
This is a hurtle I plan on tackling this next week or so potentially, the problem here is that there is little air being circulated from the open (outside) system, to the closed (inside the Pyra) system. This should exponentially amplify the cooling performance of even the stock system when implemented.
I spent a few hours last night pondering some cooling improvements to the Pyra's CPU board to mitigate throttling and improve device performance.
Few things I wanted to nip at the bud are:
-Throttling after about 10 minutes of heavy use. > Pushes clock into a downclock decay of 250Mhz, then cools, but immediately hits trip temperature back down to 250Mhz making system crippled.
-Aggregate heat buildup inside the device (thermal sink) which leads to heating up the battery and making the device quite warm. > Reduced battery life, likely bloating long term and overall just heat buildup throughout.
-Open up the POSSIBILITY to squeeze more performance out of the CPU both in regards to CPU Clock, and investigate other areas.. Such as DDR timings, and frequency, as well as SGX clock frequency (which go hand and hand as SGX uses system memory).
-Evaluate possibility of lowering core voltage through the device tree source file across (hopefully) all clock domains. As well as add more intermediate stepping (125Mhz) for the On Demand perf level.
While yet no radical redesigns have been implemented YET, I have started taking some measures to pave the path to future cooling versions and designs.
-Conformally coated the CPU board components to mitigate any sort of electrical shorting from components due to clearance of heatsink, and more aggressive (potentially) electrically conductive thermal pastes/interface materials.
-Soldered the existing heatsink (copper plate) to the main thermal element (the serrated square piece of copper). The goal here is to maximize the performance of the stock solution to get a baseline as to what is needed aftermarket to achieve the above goals.
-Removed the stock thermal glue/material from the heatsink. There seems to be an insulating adhesive coating the contact side of the heatsink, which likely significantly impacts thermal transfer from the CPU/PMC to the heatsink.
-Applied extremely high end thermal compound to heatsink and RAM/CPU/PMC chips. I considered using a graphite thermal pad here, but the tolerance/spacing of the board needs to be filled to do so to support a forceful contract on CPU/RAM/PMC.
All of this so far has yielded about a 20-23*C temperature delta from stock. Which now makes 1.5Ghz essentially perfectly usable outside of long term heavy load when the device becomes saturated.
This is a hurtle I plan on tackling this next week or so potentially, the problem here is that there is little air being circulated from the open (outside) system, to the closed (inside the Pyra) system. This should exponentially amplify the cooling performance of even the stock system when implemented.
Last edited:
