ryunokokoro wrote:
As a quick clarification, simplifying the overheating issue to simply "it's a more powerful GPU" is not a valid assessment of the issue. Any metaphors involving sports cars are entirely off base.
Actually, you kind of make my point.
The problem is not as simple as "more pixels = more energy" or "faster refreshes = higher energy" or even "newer GPU generation = more energy". It is far more detailed and far more nuanced.
It is, but the combination of high clock speed coupled with GDDR6 VRAM is very power hungry is one of the reasons HBM as used with the Radeon Pro 5600M was developed.
As some in this thread (and on other forums) have shown playing with the settings outlined in that quote (resolution, refresh rate, display adapter, etc.) have had dramatic effects on the GPU's energy usage. It is entirely possible to run a 4K monitor on one of these systems and see 5W GPU power consumption while an old 2K monitor draws 19W - the difference lies somewhere in the bag of settings/connections between the monitor and desktop.
It's possible - with software tweaks anything is possible; AMD's statement is that they run it at the speed (and power consumption level) they do to eliminate issues like flicker and graphical corruption.
Tweaking parameters lowered power consumption, yet there may be circumstances under which the effects AMD was not willing to risk occur.
As this is a known issue and expected behavior for Apple, it would be nice to have some guidance as to what settings/adapters/monitors (or combinations thereof) are known to work best.
Apple has never really done that, leaving it entirely up to the consumer as "best" will vary from person to person. For example, I want maximum graphics performance and don't care how fast fans would need to run to keep the machine cool; I always ran old MacBook Pros in "Higher Performance" graphics mode and heard fans often back then.
To those who would simplify the issue to "hot-rod GPU = high energy usage", I think a better approach to helping people would be to ask after their monitor setup details and perhaps suggest some adjustments to see if that helps them.
Which others have indeed determined experimentally, though not without the risk of other issues; in particular turning off Turbo Boost may result in lower fan speeds but it will result in lower performance.
To reiterate, that's how Intel Turbo Boost is designed to work: it runs the CPU at a higher than rated clock speed until heat/power requirements require the OS to throttle it down, giving peak performance for smaller tasks, which most tasks are. (It also helps with perception, as there are studies that indicate if a progress bar moves faster at the beginning rather than equally throughout, users will perceive a task as completing faster as a whole.)
Please stop oversimplifying this. It doesn't help anyone.
I'm not oversimplifying at all; it is a complex issue but the reasons I've given are why what is being seen is happening.
The question is what you and others are willing to give up for the sake of silent fans.
[Side Note: I have a 2018 15" MBP with two external monitors - 1@4K, 1@2K. With my standard setup, the Radeon High Side draws ~15W of power. If I simply rotate one of the monitors (I usually keep one in Landscape mode), that power usage changes (I've seen 19-20W usage depending upon settings as well). The fact of the matter is that this power draw/heat generation is not limited to the GPUs in the 16" machines - it may simply be that they are less good (or less quiet) about getting rid of the excess heat. Given that the GPU in the 16" is 'far more powerful' than those in the 2018 15" and yet have the same exact power draw, I think we can agree that the car metaphor should be put to bed.]
Actually teardowns and numbers released by Apple show the MBP 16 is much more effective at thermal dissipation than preceding MBPs, allowing the CPU to run at higher speeds before it needs to be throttled for thermal reasons.
For example, from third-party site 9to5Mac:
Initial 16-inch MacBook Pro thermal and graphics tests highlight performance gains
You are correct that heat production, and thus the need to spin up fans, is due to total thermal load, which means CPU heat + GPU heat.
It's kind of interesting to note that AMD specifies the Radeon Pro 5300M, 5500M and 5600M with identical TGP, 50w, meaning they all can consume up to 50w of power, but tests show HBM2 VRAM allows the 5600M to handle typical loads at a lower power level.
