Verify temperature readings against expected values
Verify temperature readings against expected values
The differences between 3D v-cache sizes like 5000x3D and 7000x3D appear negligible in practice. There’s no standard way to test a non-3D CPU against a 3D one on the same die, as there isn’t a direct equivalent like the R9 5000x3D. A feasible approach would be simulating identical thermal environments and comparing a 5800x versus 5800x3D setup with consistent power, voltage, and frequency settings. In contrast, a 7950x3D with direct cooling offers more control over temperature variations, making it easier to isolate draw differences between CCD modules. My observations suggest that under similar conditions, the gap between CCD0 and CCD1 would be relatively small. I used a rough estimate based on my +25°C rule, but for precise results you’d need more detailed core temperature data. The screenshot shows the process quickly, with the dataset reset right after starting.
At 143w PPT on my X3D I observe less than 80%. I haven't started yet, but I can connect it later at home. R23 is a relatively low-demand unit.
The observation that it runs in the 20s makes it difficult to accept either way. Cooling a processor at ambient temperature is nearly impossible. Your experience doesn’t even approach ambient conditions. Achieving high power consumption like 143W PPT seems unusual. This CPU appears strictly limited to around 105W as far as I know. The highest PPT I recorded during a 10-minute Cinebench run was just over 106W. After another remount, using the pea method for thermal paste improved things slightly. Now the CPU warms up more gradually—reaching about 80°C instead of jumping instantly to 90°C—and the CinebenchR23 score reached 14700. It’s now functioning within acceptable limits, though the -30 voltage offset is still in effect.