5800X3D Thermal Performance on AMD Wraith Prism

PowerSpec_MikeW

Greetings,

Summary:

This is a test on the thermal performance of the 5800X3D with the AMD Wraith Prism. The goal is to illustrate how modern Ryzen processors from the 5000 series and beyond handle high temperatures. To do we'll compare a stock run of Cinebench R23 against a small undervolt and see how the system handles the change.

Platform:

Roughly configured as a G517/G518.

Processor: 5800X3D

Motherboard: ASRock B550M-C

RAM: GSkill 2x16GB DDR4-3200. Single rank Hynix CJR/DJR.

Storage: 1TB PowerSpec NVMe.

GPU: RX550

Environment:

Open bench, no additional fans directed onto the platform. Ambient was 72f.

Stock:

Undervolt:

Overall Scores:

Results:

As you can see above we failed to hit the TJ Max with this configuration. When we applied the undervolt frequency and power consumption increased. This is to be expected as doing so will allow for a more aggressive boost under most conditions. Frequency increased from 4.17GHz to 4.3GHz and power consumption increased by roughly 2W. We also recorded a 1C increase in CPU temperature. And it netted 400 points in the test. This was a quick and dirty undervolt for the purpose of this test, auto scalar, -15 all core. It is not dialed in, there's room to play here.

Conclusion:

Lets go over thermal throttling and how the definition needs to be updated. In the past processors hard throttled at TJ Max, usually down to or below 1GHz. This resulted in a massive slow down on the PC and was very obvious. This is not how modern processors operate. This applied to Ryzen 5000 and newer on the AMD side, and the Alder Lake and later architectures on the Intel side. Lets review what AMD's (formerly) Robert Hallock said on the issue.

"Yes. I want to be clear with everyone that AMD views temps up to 90C (5800X/5900X/5950X) and 95C (5600X) as typical and by design for full load conditions. Having a higher maximum temperature supported by the silicon and firmware allows the CPU to pursue higher and longer boost performance before the algorithm pulls back for thermal reasons," Hallock said.

"Is it the same as Zen 2 or our competitor? No. But that doesn't mean something is 'wrong.' These parts are running exactly as-designed, producing the performance results we intend," Hallock continued.

Source: https://www.pcgamer.com/amd-views-ryzen-5000-cpu-temperatures-up-to-95c-as-typical-and-by-design/

The processor is meant to balance on the edge of the TJ Max and maximize performance. If you undervolt it, you give it more headroom and it overclocks more aggressively. Improved cooling can have the same effect. Do keep in mind there is a 142W socket limit, though this is going to be more of an issue with high end CPU's with high core counts like the 5900/5950. On those processors you tend to hit the power limit rather than the thermal limit.

In that context we expect a cooler to be able offer a substantial boost well above the base clock speed (3.4GHz on this CPU, 4.5GHz max boost) on a max core load. Keep in mind Cinebench is a very heavy multicore load, the load is considerably lighter on games. However, maxing all cores is our standard for selecting which coolers can be used on which processors. Temperatures are not as much of a factor, as what the processor can boost to and the power it's able to maintain over an extended period of time.