Post by garden4 on Oct 12, 2020 12:26:49 GMT
2020 Macbook Pro VRM Mod Results - Graph of power draw
I have performed several iterations of the VRM thermal pad mod. The first started with only partial contact to the black thermal insulation tape of the bottom case, and the most recent one has up to four layers of pads for better contact with the bare (tape-stripped) bottom case.
Test: Test duration is 3.8 minutes
My thoughts on test: Due to prewarmup, time to throttling seesaw is shorter than it would be for a cold boot. However, I think this test run and resulting data is a good representation of the usual throttling I see when gaming.
Machine: 2020 Macbook Pro, i9 2.3GHz, 5600M, 32GB RAM
Hardware setup: Clamshell, resting on a laptop cooler with a fan (pushes air up to bottom of case)
Software setup: Bootcamp
Ambient temps: Lowest at 31C on battery sensor 1 (it's 4am)
Mod info: Fujipoly 17W/mK 1.5mm and .5mm. Only padded VRM parts, and not CPU or GPU. Bought a second bottom case on eBay for $80 to preserve warranty after removing tape.
Conditions: Machine warmed up until power draw throttling seesaw occurs, which is after 10 minutes watching a CS GO replay at 5K res + highish settings. Then, machine is idled for a few minutes to configure and set up logging, and tabbed back to the game for full active load for the start of the log data.
Software: Not running any modifications like ThrottleStop etc.
Logging software: XTU, GPU-Z, (Afterburner- data not used)
Result
----
Averages taken from ~229 one-second-interval data points (229 seconds = 3.8 minutes)
Average CPU draw (W): 18.45946378
Average GPU draw (W): 45.71491228
Combined average CPU+GPU draw (W): 64.17437606
Analysis
----
- During the throttled period, combined power draw is around 57-58W. The up periods seem to average around 70W.
- Up periods last 4-6 seconds, down periods are 8-16 seconds.
Thoughts
----
Even with my best attempts to perform the mod, this machine still experiences throttling such that it is unable to sustain 70W. In CS GO (at 5K), this means dipping to an unacceptable 50 fps at 58W versus the 60+ fps at 70W. While I believe there is an improvement in performance (more on this later), this mod is not sufficient to sustain 70W+. As 70W seems to be the stable point where the GPU maxes its load at 45W and the CPU boosts to an acceptable 4GHz+ at 25-35W, this mod by itself is not a definitive solution to fixing performance throttling.
If you believe me when I say I can run games at higher settings stably where I was unable to before, and considering that the combined average draw is now at ~64W versus the 57W (IIRC) max that I was limited to before, then there is still a decent improvement to be had with this mod.
I think a solution to throttling could be a combination of this mod and limiting the max power draw of the CPU to 19W, to stay within the thermal envelope (terminology?) of the VRM parts. Any recommendations on software to do this?
I have performed several iterations of the VRM thermal pad mod. The first started with only partial contact to the black thermal insulation tape of the bottom case, and the most recent one has up to four layers of pads for better contact with the bare (tape-stripped) bottom case.
Test: Test duration is 3.8 minutes
My thoughts on test: Due to prewarmup, time to throttling seesaw is shorter than it would be for a cold boot. However, I think this test run and resulting data is a good representation of the usual throttling I see when gaming.
Machine: 2020 Macbook Pro, i9 2.3GHz, 5600M, 32GB RAM
Hardware setup: Clamshell, resting on a laptop cooler with a fan (pushes air up to bottom of case)
Software setup: Bootcamp
Ambient temps: Lowest at 31C on battery sensor 1 (it's 4am)
Mod info: Fujipoly 17W/mK 1.5mm and .5mm. Only padded VRM parts, and not CPU or GPU. Bought a second bottom case on eBay for $80 to preserve warranty after removing tape.
Conditions: Machine warmed up until power draw throttling seesaw occurs, which is after 10 minutes watching a CS GO replay at 5K res + highish settings. Then, machine is idled for a few minutes to configure and set up logging, and tabbed back to the game for full active load for the start of the log data.
Software: Not running any modifications like ThrottleStop etc.
Logging software: XTU, GPU-Z, (Afterburner- data not used)
Result
----
Averages taken from ~229 one-second-interval data points (229 seconds = 3.8 minutes)
Average CPU draw (W): 18.45946378
Average GPU draw (W): 45.71491228
Combined average CPU+GPU draw (W): 64.17437606
Analysis
----
- During the throttled period, combined power draw is around 57-58W. The up periods seem to average around 70W.
- Up periods last 4-6 seconds, down periods are 8-16 seconds.
Thoughts
----
Even with my best attempts to perform the mod, this machine still experiences throttling such that it is unable to sustain 70W. In CS GO (at 5K), this means dipping to an unacceptable 50 fps at 58W versus the 60+ fps at 70W. While I believe there is an improvement in performance (more on this later), this mod is not sufficient to sustain 70W+. As 70W seems to be the stable point where the GPU maxes its load at 45W and the CPU boosts to an acceptable 4GHz+ at 25-35W, this mod by itself is not a definitive solution to fixing performance throttling.
If you believe me when I say I can run games at higher settings stably where I was unable to before, and considering that the combined average draw is now at ~64W versus the 57W (IIRC) max that I was limited to before, then there is still a decent improvement to be had with this mod.
I think a solution to throttling could be a combination of this mod and limiting the max power draw of the CPU to 19W, to stay within the thermal envelope (terminology?) of the VRM parts. Any recommendations on software to do this?