The original study is from AMD's forum and it is a response to a hung up query.
Noctua, AMD and GN's technology journalists have attended to the question of why AMD runs hot. I'm going to prove it on a third party case study.
You can access individual company takes on the question in the hereby links:
You can go look at individual threads to go into the specifics, but the crux of the matter is that cpu heat density is limited from sufficient heat conduction by the thermal paste performance.
Our envoy takes us into the mystical journey of why 250w-spec tower air cooler is necessitated to run at high heat in order to cool the cpu properly. Gamersnexus is the go to guide here. As ironic as it is, how TDP is calculated is totally different to what we have been used to;
As it stands, the under recognized limit to cooling performance, the heat conductance through the IHS, is the dominant overclocking performance determinant.
AMD keeps reference to '°C/W', thermal resistivity, while 'W/°C'(thermal conductivity) - how much temperature gradient you can maintain between the ihs and heatsink plates - maintains how much heat capacity you can specify for the TDP of their cpus. This is inverse to what AMD has been saying, you get improved cooling by first maintaining a high gradient, not a lower temperature.
7nm dense architectures can only extract enough heat from the cpu die integrated heatsink, once temperature reaches safety limits.
The difference of liquid metal is, it provides the best conductance to the heatsink plate under normal operating temperatures.
Nonconductive low tier pastes can only operate within the same conductance at the highest TjMax. The cpu is making the best of available thermal gradient when liquid metal is applied, jumping from 80w>120w while cooling below 80°C>65°C.
Since AMD, GN, or Noctua hasn't mentioned in anyway how the Zen 2 architecture would be served best either with nickel plated air coolers that work with liquid metal in stock settings, or aio coolers that work cooler than 90°C needs the same liquid metal application, I think this requires a reevaluation. We don't see any observable difference to overclocking these cpus under normal conditions because the ihs is the conduction limit and the sole solution being upgrading to liquid metal interface materials.
Noctua, AMD and GN's technology journalists have attended to the question of why AMD runs hot. I'm going to prove it on a third party case study.
You can access individual company takes on the question in the hereby links:
You can go look at individual threads to go into the specifics, but the crux of the matter is that cpu heat density is limited from sufficient heat conduction by the thermal paste performance.
After applying metal liquid between the cooler and IHS things improved by almost 20°C in the range of 65-80°, that was when the CPU draws from 80 to 120watts, (80° became 60 at 120watts!) but improved only by 3°C when the wattage jumps over 150watts, (90° instead of 93-94°), also frequencies improved accordingly.
- Under 55°C it sits at 4350MHz all cores and occasionally some cores jump to 4.4
- @65°C it draws 80Watts and sits at 4150MHz all cores
- @75°C it draws 90Watts and sits at 4100MHz all cores
- @85°C it draws 100-110Watts and sits at 4050MHz all cores
- @90°C it draws 150-160Watts and sits at 3980MHz all cores
- @94°C+ almost 170watts, throttling rules and it sits at 3900MHz all cores.
AMD defines HSF θca (°C/W) as: The minimum °C per Watt rating of the heatsink to achieve rated performance.
Its internal definition, for comparison, says “the minimum required heatsink resistance necessary to maintain the case temperature within specification for the thermal design power (TDP) and assumptions for the external ambient temperature and system temperature rise (Tsys).”
“Theta CA”
The HSF is what stands between the CPU and the surrounding air, and θca is the thermal resistance between the CPU and the air, so HSF θca is the thermal resistance of the heatsink. Lower is actually better here, not higher, so AMD’s phrasing has some interpretive gray areas. AMD’s reviewer document should instead read “maximum” instead of “minimum,” so it should be the, quote, “maximum *C per Watt rating of the heatsink,” as lower is better and so maximum would be the last value permissible for rated performance before becoming insufficient for rated performance. When we reached out, AMD clarified that “you can interpret the original copy to mean ‘the [minimum standard]’ where lower values produce superior results.”
AMD keeps reference to '°C/W', thermal resistivity, while 'W/°C'(thermal conductivity) - how much temperature gradient you can maintain between the ihs and heatsink plates - maintains how much heat capacity you can specify for the TDP of their cpus. This is inverse to what AMD has been saying, you get improved cooling by first maintaining a high gradient, not a lower temperature.
From Noctua:
Due to the small size of the CPU-die, the heat density (W/mm²) of this chip is very high. For example, a 120W heatload at a chip-size of 74mm² results in a heat-density of 1.62W/mm², whereas the same heatload on an older Ryzen processor with a chip-size of 212mm² gives a heat-density of just 0.57W/mm².
I'm pretty sure AMD did enough stress test to determine the CPUs are going to be fine.
TL;DR: It's just harder to cool because the heat builds up so fast, the heat spreader can't extract the heat from the die fast enough.
The difference of liquid metal is, it provides the best conductance to the heatsink plate under normal operating temperatures.
Nonconductive low tier pastes can only operate within the same conductance at the highest TjMax. The cpu is making the best of available thermal gradient when liquid metal is applied, jumping from 80w>120w while cooling below 80°C>65°C.
Since AMD, GN, or Noctua hasn't mentioned in anyway how the Zen 2 architecture would be served best either with nickel plated air coolers that work with liquid metal in stock settings, or aio coolers that work cooler than 90°C needs the same liquid metal application, I think this requires a reevaluation. We don't see any observable difference to overclocking these cpus under normal conditions because the ihs is the conduction limit and the sole solution being upgrading to liquid metal interface materials.