HWiNFO v6.04 available. Changes: Full AMD Radeon VII monitoring support. Added Intel Cooper Lake and Snow Ridge. Improved configuration of tray icons. Improved support of multiple devices behind Intel VROC. Added monitoring of Corsair Platinum series (H100i, H115i) coolers. Improved AMD Zen2 support. Added reporting of TDC and EDC limits for AMD Zen CPUs. Improved detection of serial ports. Fixed reporting on system temperature on some MSI mainboards. Fixed occasional 0 RPM fan values on some systems with ITE SIOs. Added preliminary support of Intel Emmitsburg. Added preliminary support of AMD Navi. Added reporting of DIMM temperatures via IMC on Sandy Bridge and later server CPUs. Added several new CPU and GPU models. Minor improvements to UI and bug fixes.
Ran v6.04 with Z97 + i7-4980HQ and HWinfo seems to be reporting only 80% of CPU / DRAM power. In fact this 80% reporting seems to go way back to at least version 543 Version 502 gives the expected values so don't know why newer versions are different.
What is your actual BCLK and what PCIe clock is shown in HWiNFO, 125 MHz ? It would be great if you could attach the HWiNFO Debug File.
Hi Martin, in that particular SS using 4:5 strap, BCLK 125, PCIe 100. RAPL appears tied to BCLK, ie proportionally more counts per millisecond the higher BCLK is. For instance using a forced power limit of 40W with fixed voltage supply shows the same power delta form mains supply, 12V Rail supply and VRM IOUT for straps of 5:5, 4:5 and 3:5 and same throttled clocks at 100MHz, 125MHz and 167MHz BCLK. HWiNFO shows 40W for package power for all straps while the others including latest XTU to date show 40W, 50W and 66W respectively, each proportional to the strap used. So in this case HWiNFO is correct and others wrong. I'm assuming you switched to using the strap ratio for RAPL estimations at some point but have maybe forgotten with all that has happened since then. Your results don't seem to be based solely on BCLK as changing BCLK to 95MHz and PCIe at 95MHz results in 38W being reported with 40W limit. I'm a little surprised this isn't mentioned in the Intel SDM or specification update unless I missed it and even more surprised that XTU is incorrect.
Indeed, there's a RAPL power adjustment done by HWiNFO, but it is tied only to those cases where a different PCIE:BCLK strap is used. So in case of 95 MHz BCLK this isn't applied. Problem is that this is not well documented by Intel, not even in special non-public documents. So besides the case of 95 MHz, in which other cases you see the issue? I'd really need to see the HWiNFO Debug File for those so I can check why it's not correct.
